Dwarkesh Podcast

Sergey Levine, one of the world’s top robotics researchers and co-founder of Physical Intelligence, thinks we’re on the cusp of a “self-improvement flywheel” for general-purpose robots. His median estimate for when robots will be able to run households entirely autonomously? 2030.

If Sergey’s right, the world 5 years from now will be an insanely different place than it is today. This conversation focuses on understanding how we get there: we dive into foundation models for robotics, and how we scale both the data and the hardware necessary to enable a full-blown robotics explosion.

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Labelbox provides high-quality robotics training data across a wide range of platforms and tasks. From simple object handling to complex workflows, Labelbox can get you the data you need to scale your robotics research. Learn more at labelbox.com/dwarkesh

• Hudson River Trading uses cutting-edge ML and terabytes of historical market data to predict future prices. I got to try my hand at this fascinating prediction problem with help from one of HRT’s senior researchers. If you’re curious about how it all works, go to hudson-trading.com/dwarkesh

• Gemini 2.5 Flash Image (aka nano banana) isn’t just for generating fun images — it’s also a powerful tool for restoring old photos and digitizing documents. Test it yourself in the Gemini App or in Google’s AI Studio: ai.studio/banana

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – Timeline to widely deployed autonomous robots

(00:17:25) – Why robotics will scale faster than self-driving cars

(00:27:28) – How vision-language-action models work

(00:45:37) – Changes needed for brainlike efficiency in robots

(00:57:59) – Learning from simulation

(01:09:18) – How much will robots speed up AI buildouts?

(01:18:01) – If hardware’s the bottleneck, does China win by default?

Dwarkesh Patel

Today I'm chatting with Sergey Levine, who is a co-founder of Physical Intelligence, which is a robotics foundation model company, and also a professor at UC Berkeley and just generally one of the world's leading researchers in robotics, RL, and AI. Sergey, thank you for coming on the podcast.

Sergey Levine

Thank you, and thank you for the kind introduction.

Dwarkesh Patel

Let's talk about robotics. Before I pepper you with questions, I'm wondering if you can give the audience a summary of where Physical Intelligence is at right now. You guys started a year ago. What does the progress look like? What are you guys working on?

Sergey Levine

Physical Intelligence aims to build robotic foundation models. That basically means general-purpose models that could in principle control any robot to perform any task. We care about this because we see this as a very fundamental aspect of the AI problem. The robot is essentially encompassing all AI technology. If you can get a robot that's truly general, then you can do, hopefully, a large chunk of what people can do.

Where we're at right now is that we've kind of gotten to the point where we've built out a lot of the basics. Those basics actually are pretty cool. They work pretty well. We can get a robot that will fold laundry and that will go into a new home and try to clean up the kitchen. But in my mind, what we're doing at Physical Intelligence right now is really the very, very early beginning. It's just putting in place the basic building blocks, on top of which we can then tackle all these really tough problems.

Dwarkesh Patel

What's a year-by-year vision? One year in, I got a chance to watch some of the robots, they can do pretty dexterous tasks like folding a box using grippers. It's pretty hard to fold the box even with my hands.

If you had to go year by year until we get to the full robotics explosion, what is happening every single year? What is the thing that needs to be unlocked, et cetera?

Sergey Levine

There are a few things that we need to get right. Dexterity obviously is one of them. In the beginning we really want to make sure that we understand whether the methods that we're developing have the ability to tackle the kind of intricate tasks that people can do. As you mentioned, folding a box, folding different articles of laundry, cleaning up a table, making a coffee, that sort of thing. That's good, that works. The results we've been able to show are pretty cool, but the end goal of this is not to fold a nice T-shirt. The end goal is to just confirm our initial hypothesis that the basics are solid.

From there, there are a number of really major challenges. Sometimes when results get abstracted to the level of a three-minute video, someone can look at this video and it's like, "Oh, that's cool. That's what they're doing." But it's not. It's a very simple and basic version of what I think is to come. What you really want from a robot is not to tell it like, "Hey, please fold my T-shirt." What you want from a robot is to tell it like, "Hey, robot, you're now doing all sorts of home tasks for me. I like to have dinner made at 6:00 p.m. I wake up and go to work at 7:00 a.m. I like to do my laundry on Saturday, so make sure that it's ready. This and this and this. By the way, check in with me every Monday to see what I want you to pick up when you do the shopping.”

That's the prompt. Then the robot should go and do this for six months, a year. That's the duration of the task. Ultimately if this stuff is successful, it should be a lot bigger. It should have that ability to learn continuously. It should have the understanding of the physical world, the common sense, the ability to go in and pull in more information if it needs it. Let’s say I ask it, "Hey, tonight, can you make me this type of salad?" It should figure out what that entails, look it up, go and buy the ingredients. There's a lot that goes into this.

It requires common sense. It requires understanding that there are certain edge cases that you need to handle intelligently, cases where you need to think harder. It requires the ability to improve continuously. It requires understanding safety, being reliable at the right time, being able to fix your mistakes when you do make those mistakes. There's a lot more that goes into this. But the principles there are: you need to leverage prior knowledge and you need to have the right representations.

Dwarkesh Patel

This grand vision, what year? If you had to give an estimate. 25 percentile, 50, 75?

Sergey Levine

I think it's something where it's not going to be a case where we develop everything in the laboratory and then it's done and then come 2030-something, you get a robot in a box. Again, it'll be the same as what we've seen with AI assistants. Once we reach some basic level of competence where the robot is delivering something useful, it'll go out there in the world. The cool thing is that once it's out there in the world, they can collect experience and leverage that experience to get better. To me, what I tend to think about in terms of timelines is not the date when it will be done, but the date when the flywheel starts basically.

Dwarkesh Patel

When does the flywheel start?

Sergey Levine

That could be very soon. There's some decisions to be made. The trade-off there is that the more narrowly you scope the thing, the earlier you can get it out into the real world. But this is something we're already exploring. We're already trying to figure out what are the real things this thing can do that could allow us to start spinning the flywheel. But in terms of stuff that you would actually care about, that you would want to see… I don't know but single-digit years is very realistic. I'm really hoping it'll be more like one or two before something is actually out there, but it's hard to say.

Dwarkesh Patel

Something being out there means what? What is out there?

Sergey Levine

It means that there is a robot that does a thing that you actually care about, that you want done. It does so competently enough to actually do it for real, for real people that want it done.

Dwarkesh Patel

We already have LLMs which are broadly deployed. That hasn't resulted in some sort of flywheel, at least not some obvious flywheel for the model companies where now Claude is learning how to do every single job in the economy or GPT's learning how to do every single job in the economy. So, why doesn’t that flywheel work for LLMs?

Sergey Levine

Well, I think it's actually very close to working and I am 100% certain that many organizations are working on exactly this. In fact, arguably there is already a flywheel. It’s not an automated flywheel but a human-in-the-loop flywheel. Everybody who's deploying an LLM is of course going to look at what it's doing and it's going to use that to then modify its behavior.

It's complex because it comes back to this question of representations and figuring out the right way to derive supervision signals and ground those supervision signals in the behavior of the system so that it improves on what you want. I don't think that's a profoundly impossible problem. It's just something where the details get pretty gnarly and challenges with algorithms and stability become pretty complex. It's something that's taken a while for the community collectively to get their hands on.

Dwarkesh Patel

Do you think it'll be easier for robotics? Or do you think that with these kinds of techniques to label data that you collect out in the world and use it as a reward, the whole wave will rise and robotics will rise as well? Or is there some reason robotics will benefit more from this?

Sergey Levine

I don't think there's a profound reason why robotics is that different. There are a few small differences that make things a little bit more manageable. Especially if you have a robot that's doing something in cooperation with people, whether it's a person that's supervising it or directing it, there are very natural sources of supervision. There's a big incentive for the person to provide the assistance that will make things succeed. There are a lot of dynamics where you can make mistakes and recover from those mistakes and then reflect back on what happened and avoid that mistake in the future.

When you're doing physical things in the real world, that stuff just happens more often than it does if you're an AI assistant answering a question. If you answer a question and just answer it wrong, it's not like you can just go back and tweak a few things. The person you told the answer to might not even know that it's wrong. Whereas if you're folding the T-shirt and you messed up a little bit, it's pretty obvious. You can reflect on that, figure out what happened, and do it better next time.

Dwarkesh Patel

Okay, in one year we have robots which are doing some useful things. Maybe if you have some relatively simple loopy process, they can do it for you, like keep folding thousands of boxes or something.

But then there's some flywheel… and there's some machine which will just run my house for me as well as a human housekeeper would. What is the gap between this thing which will be deployed in a year that starts the flywheel and this thing which is like a fully autonomous housekeeper?

Sergey Levine

It's actually not that different from what we've seen with LLMs in some ways. It's a matter of scope. Think about coding assistants. Initially the best tools for coding, they could do a little bit of completion. You give them a function signature and they'll try their best to type out the whole function and they'll maybe get half of it right. As that stuff progresses, then you're willing to give these things a lot more agency.

The very best coding assistance now—if you're doing something relatively formulaic, maybe it can put together most of a PR for you for something fairly accessible. It'll be the same thing. We'll see an increase in the scope that we're willing to give to the robots as they get better and better. Initially the scope might be a particular thing you do. You're making the coffee or something. As they get more capable, as their ability to have common sense and a broader repertoire of tasks increases, then we'll give them greater scope. Now you're running the whole coffee shop.

Dwarkesh Patel

I get that there's a spectrum. I get that there won't be a specific moment that feels like we've achieved it but if you had to give a year for your median estimate of when that happens?

Sergey Levine

My sense there too is that this is probably a single-digit thing rather than a double-digit thing. The reason it's hard to really pin down is because, as with all research, it does depend on figuring out a few question marks. My answer in terms of the nature of those question marks is that I don't think these are things that require profoundly, deeply different ideas but it does require the right synthesis of the kinds of things that we already know. Sometimes synthesis, to be clear, is just as difficult as coming up with profoundly new stuff.

It's intellectually a very deep and profound problem. Figuring that out is going to be very exciting. But I think we kind of know roughly the puzzle pieces and it's something that we need to work on. If we work on it and we're a bit lucky and everything kind of goes as planned, single-digit is reasonable.

Dwarkesh Patel

I'm just going to do binary search until I get a year. It's less than 10 years, so more than five years, your median estimate? I know there's a range.

Sergey Levine

I think five is a good median.

Dwarkesh Patel

Okay, five years. If you can fully autonomously run a house, then you can fully autonomously do most blue-collar work. Your estimate is that in five years it should be able to do most blue-collar work in the economy.

Sergey Levine

There's a nuance here. It becomes more obvious if we consider the analogy to coding assistants. It's not like the nature of coding assistants today is that there's a switch that flips and instead of writing software, suddenly all software engineers get fired and everyone's using LLMs for everything. It actually makes a lot of sense that the biggest gain in productivity comes from experts, which is software engineers, whose productivity is now augmented by these really powerful tools.

Dwarkesh Patel

Separate from the question of whether people will get fired or not, a different question is, what will the economic impact be in five years? The reason I'm curious about this is because with LLMs, the relationship between the revenues for these models to their seeming capability has been sort of mysterious. You have something which feels like AGI. You can have a conversation where it really passes the Turing test. It really feels like it can do all this knowledge work. It's obviously doing a bunch of coding, et cetera. But the revenues from these AI companies are cumulatively on the order of $20-30 billion per year and that's much less than all knowledge work, which is $30-40 trillion.

In five years are we in a similar situation to what LLMs are in now, or is it more like we have robots deployed everywhere and they're actually doing a whole bunch of real work, et cetera?

Sergey Levine

It's a very subtle question. What it probably will come down to is this question of scope. The reason that LLMs aren't doing all software engineering is because they're good within a certain scope, but there's limits to that. Those limits are increasing, to be clear, every year. I think that there's no reason that we wouldn't see the same kind of thing with robots.

The scope will have to start out small because there will be certain things that these systems can do very well and certain other things where more human oversight is really important. The scope will grow. What that will translate into is increased productivity. Some of that productivity will come from the robots themselves being valuable. Some of it will come from the people using the robots are now more productive in their work.

Dwarkesh Patel

But there's so many things which increase productivity. Like wearing gloves increases productivity or I don't know. You want to understand something which increases productivity a hundredfold versus something which has a small increase. Robots already increase productivity for workers.

Where LLMs are right now in terms of the share of knowledge work they can do, is I guess like 1/1000th of the knowledge work that happens in the economy, at least in terms of revenue. Are you saying that fraction will be possible for robots, but for physical work, in five years?

Sergey Levine

That's a very hard question to answer. I'm probably not prepared to tell you what percentage of all labor work can be done by robots, because I don't think right now, off the cuff, I have a sufficient understanding of what's involved in that big of a cross-section of all physical labor. What I can tell you is this. It's much easier to get effective systems rolled out gradually in a human-in-the-loop setup.

Again, this is exactly what we've seen with coding systems. I think we'll see the same thing with automation, where basically robot plus human is much better than just human or just robot. That just makes total sense. It also makes it much easier to get all the technology bootstrapped. Because when it's robot plus human now, there's a lot more potential for the robot to actually learn on the job, acquire new skills.

Dwarkesh Patel

Because a human can label what's happening?

Sergey Levine

Also because the human can help, the human can give hints. Let me tell you this story. When we were working on the π0.5 project, the paper that we released last April, we initially controlled our robots with teleoperation in a variety of different settings. At some point we actually realized that we can actually make significant headway, once the model was good enough, by supervising it not just with low-level actions but actually literally instructing it through language.

Now you need a certain level of competence before you can do that, but once you have that level of competence, just standing there and telling the robot, "Okay, now pick up the cup, put the cup in the sink, put the dish in the sink," just with words already, actually gives the robot information that it can use to get better. Now imagine what this implies for the human plus robot dynamic.

Now basically, learning for these systems is not just learning from raw actions, it's also learning from words. Eventually it’ll be learning from observing what people do from the kind of natural feedback that you receive when you're doing a job together with somebody else. This is also the kind of stuff where the prior knowledge that comes from these big models is tremendously valuable, because that lets you understand that interaction dynamic. There's a lot of potential for these kinds of human plus robot deployments to make the model better.

Dwarkesh Patel

In terms of robotics progress, why won't it be like self-driving cars. It's been more than 10 years since Google launched its… Wasn't it in 2009 that they launched the self-driving car initiative? I remember when I was a teenager, watching demos where we would go buy a Taco Bell and drive back. Only now do we have them actually deployed. Even then they may make mistakes, etc. Maybe it'll be many more years before most of the cars are self-driving.

You're saying five years to this quite robust thing, but actually will it just feel like 20 years? Once we get the cool demo in five years, then it'll be another 10 years before we have the Waymo and the Tesla FSD working.

Sergey Levine

That's a really good question. One of the big things that is different now than it was in 2009 has to do with the technology for machine learning systems that understand the world around them. Principally for autonomous driving, this is perception. For robots, it can mean a few other things as well. Perception certainly was not in a good place in 2009. The trouble with perception is that it's one of those things where you can nail a really good demo with a somewhat engineered system, but hit a brick wall when you try to generalize it.

Now at this point in 2025, we have much better technology for generalizable and robust perception systems and, more generally, generalizable and robust systems for understanding the world around us. When you say that the system is scalable, in machine learning scalable really means generalizable. That gives us a much better starting point today. That's not an argument about robotics being easier than autonomous driving. It's just an argument for 2025 being a better year than 2009.

But there's also other things about robotics that are a bit different than driving. In some ways, robotic manipulation is a much, much harder problem. But in other ways, it's a problem space where it's easier to get rolling, to start that flywheel with a more limited scope. To give you an example, if you're learning how to drive, you would probably be pretty crazy to learn how to drive on your own without somebody helping you. You would not trust your teenage child to learn to drive just on their own, just drop them in the car and say, "Go for it." That's also a 16-year-old who's had a significant amount of time to learn about the world. You would never even dream of putting a five-year-old in a car and telling him to get started. But if you want somebody to clean the dishes, dishes can break too. But you would probably be okay with a child trying to do the dishes without somebody constantly sitting next to them with a brake, so to speak.

For a lot of tasks that we want to do with robotic manipulation, there's potential to make mistakes and correct those mistakes. When you make a mistake and correct it, well first you've achieved the task because you've corrected, but you've also gained knowledge that allows you to avoid that mistake in the future. With driving, because of the dynamics of how it's set up, it's very hard to make a mistake, correct it and then learn from it because the mistakes themselves have significant ramifications.

Not all manipulation tasks are that. There are truly some very safety-critical stuff. This is where the next thing comes in, which is common sense. Common sense, meaning the ability to make inferences about what might happen that are reasonable guesses, but that do not require you to experience that mistake and learn from it in advance. That's tremendously important. That's something that we basically had no idea how to do about five years ago. But now we can use LLMs and VLMs and ask them questions and they will make reasonable guesses. They will not give you expert behavior, but you can say, "Hey, there's a sign that says slippery floor. What's going to happen when I walk up over that?" It's pretty obvious, right? No autonomous car in 2009 would have been able to answer that question.

Common sense plus the ability to make mistakes and correct those mistakes, that's sounding an awful lot what a person does when they're trying to learn something. All of that doesn't make robotic manipulation easy necessarily, but it allows us to get started with a smaller scope and then grow from there.

Dwarkesh Patel

So for years, I mean not since 2009, but we've had lots of video data, language data, and transformers for 5-8 years. Lots of companies have tried to build transformer-based robots with lots of training data, including Google, Meta, et cetera. What is the reason that they've been hitting roadblocks? What has changed now?

Sergey Levine

That's a really good question. I'll start out with a slight modification to your comment. They've made a lot of progress. In some ways, a lot of the work that we're doing now at Physical Intelligence is built on the backs of lots of other great work that was done, for example, at Google. Many of us were at Google before. We were involved in some of that work. Some of it is work that we're drawing on that others did. There's definitely been a lot of progress there.

But to make robotic foundation models really work, it's not just a laboratory science experiment. It also requires industrial scale building effort. It's more like the Apollo program than it is a science experiment. The excellent research that was done in the past industrial research labs, and I was involved in much of that, was very much framed as a fundamental research effort. That's good. The fundamental research is really important, but it's not enough by itself. You need the fundamental research and you also need the impetus to make it real.

Making it real means actually putting the robots out there, getting data that is representative, the tasks that they need to do in the real world, getting that data at scale, building out the systems, and getting all that stuff right. That requires a degree of focus, a singular focus on really nailing the robotic foundation model for its own sake, not just as a way to do more science, not just as a way to publish a paper, and not just as a way to have a research lab.

Dwarkesh Patel

What is preventing you now from scaling that data even more? If data is a big bottleneck, why can't you just increase the size of your office 100x, have 100x more operators operating these robots and collecting more data. Why not ramp it up immediately 100x more?

Sergey Levine

That's a really good question. The challenge here is understanding which axes of scale contribute to which axes of capability. If we want to expand capability horizontally—meaning the robot knows how to do 10 things now and I'd like it to do 100 things later—that can be addressed by just directly horizontally scaling what we already have. But we want to get robots to a level of capability where they can do practically useful things in the real world.

That requires expanding along other axes too. It requires, for example, getting to very high robustness. It requires getting them to perform tasks very efficiently, quickly. It requires them to recognize edge cases and respond intelligently. Those things can also be addressed with scaling. But we have to identify the right axes for that, which means figuring out what data to collect, what settings to collect it in, what methods consume that data, and how those methods work.

Answering those questions more thoroughly will give us greater clarity on the axes, on those dependent variables, on the things that we need to scale. We don't fully know right now what that will look like. I think we'll figure it out pretty soon. It's something we're working on actively. We want to really get that right so that when we do scale it up, it'll directly translate into capabilities that are very relevant to practical use.

Dwarkesh Patel

Just to give an order of magnitude, how does the amount of data you have collected compare to internet-scale pre-training data? I know it's hard to do a token-by-token count, because how does video information compare to internet information, et cetera. But using your reasonable estimates, what fraction?

Sergey Levine

It's very hard to do because robotic experience consists of time steps that are very correlated with each other. The raw byte representation is enormous, but probably the information density is comparatively low. Maybe a better comparison is to the datasets that are used for multimodal training. And there, I believe last time we did that count, it was between one and two orders of magnitude.

Dwarkesh Patel

The vision you have of robotics, will it not be possible until you collect what, 100x, 1000x more data?

Sergey Levine

That's the thing, we don't know that. It's certainly very reasonable to infer that robotics is a tough problem. Probably it requires as much experience as the language stuff. But because we don't know the answer to that, to me a much more useful way to think about it is not how much data do we need to get before we're fully done, but how much data do we need to get before we can get started. That means before we can get a data flywheel that represents a self-sustaining and ever-growing data-collection recipe.

Dwarkesh Patel

When you say self-sustaining, is it just learning on the job or do you have something else in mind?

Sergey Levine

Learning on the job or acquiring data in a way such that the process of acquisition of that data itself is useful and valuable.

Dwarkesh Patel

I see. Some kind of RL.

Sergey Levine

Doing something actually real. Ideally I would like it to be RL, because with RL you can get away with the robot acting autonomously which is easier. But it's not out of the question that you can have mixed autonomy. As I mentioned before, robots can learn from all sorts of other signals. I described how we can have a robot that learns from a person talking to it. There's a lot of middle ground in between fully teleoperated robots and fully autonomous robots.

Dwarkesh Patel

How does the π0 model work?

Sergey Levine

The current model that we have basically is a vision-language model that has been adapted for motor control. To give you a little bit of a fanciful brain analogy, a VLM, a vision-language model, is basically an LLM that has had a little pseudo visual cortex grafted to it, a vision encoder. Our models, they have a vision encoder, but they also have an action expert, an action decoder essentially. It has a little visual cortex and notionally a little motor cortex.

The way that the model makes decisions is it reads in the sensory information from the robot. It does some internal processing. That could involve outputting intermediate steps. You might tell it, "Clean up the kitchen." It might think to itself, "Hey, to clean up the kitchen, I need to pick up the dish and I need to pick up the sponge and I need to put this and this." Eventually it works its way through that chain-of-thought generation down to the action expert, which produces continuous actions.

That has to be a different module because the actions are continuous, they're high frequency. They have a different data format than text tokens. But structurally it's still an end-to-end transformer. Roughly speaking, technically, it corresponds to a mixture-of-experts architecture.

Dwarkesh Patel

And what is actually happening is that it's predicting "I should do X thing." Then there's an image token, then some action tokens –what it actually ends up doing– and then more image, more text description, more action tokens. Basically I'm looking at what stream is going on.

Sergey Levine

That's right, with the exception that the actions are not represented as discrete tokens. It actually uses flow matching and diffusion because they're continuous and you need to be very precise with your actions for dexterous control.

Dwarkesh Patel

I find it super interesting that you're using the open-source Gemma model, which is Google's LLM that they released open source, and then adding this action expert on top. I find it super interesting that the progress in different areas of AI is based on not only the same techniques, but literally the same model. You can just use an open-source LLM and add this action expert on top.

You naively might think that, “Oh, there's a separate area of research which is robotics, and there's a separate area of research called LLMs and natural language processing.” No, it's literally the same. The considerations are the same, the architectures are the same, even the weights are the same. I know you do more training on top of these open-source models, but I find that super interesting.

Sergey Levine

One theme here that is important to keep in mind is that the reason that those building blocks are so valuable is because the AI community has gotten a lot better at leveraging prior knowledge. A lot of what we're getting from the pre-trained LLMs and VLMs is prior knowledge about the world. It's a little bit abstracted knowledge. You can identify objects, you can figure out roughly where things are in image, that sort of thing.

But if I had to summarize in one sentence, the big benefit that recent innovations in AI give to robotics is the ability to leverage prior knowledge. The fact that the model is the same model, that's always been the case in deep learning. But it's that ability to pull in that prior knowledge, that abstract knowledge that can come from many different sources that's really powerful.

Dwarkesh Patel

I was talking to this researcher, Sander at GDM, and he works on video and audio models. He made the point that the reason, in his view, we aren't seeing that much transfer learning between different modalities. That is to say, training a language model on video and images doesn't seem to necessarily make it that much better at textual questions and tasks because images are represented at a different semantic level than text.

His argument is that text has this high-level semantic representation within the model, whereas images and videos are just compressed pixels. When they're embedded, they don't represent some high-level semantic information. They're just compressed pixels. Therefore there's no transfer learning at the level at which they're going through the model.

Obviously this is super relevant to the work you're doing. Your hope is that by training the model on the visual data that the robot sees, visual data generally maybe even from YouTube or whatever eventually, plus language information, plus action information from the robot itself, all of this together will make it generally robust. You had a really interesting blog post about why video models aren't as robust as language models. Sorry, this is not a super well-formed question. I just wanted to get a reaction.

Sergey Levine

Yeah, what’s up with that? I have maybe two things I can say there. I have some bad news and some good news. The bad news is what you're saying is really getting at the core of a long-running challenge with video and image generation models. In some ways, the idea of getting intelligent systems by predicting video is even older than the idea of getting intelligent systems by predicting text. The text stuff turned into practically useful things earlier than the video stuff did. I mean, the video stuff is great. You can generate cool videos. The work that's been done there recently is amazing. But it's not like just generating videos and images has already resulted in systems that have this deep understanding of the world where you can ask them to do stuff beyond just generating more images and videos. Whereas with language, clearly it has.

This point about representations is really key to it. One way we can think about it is this. Imagine pointing a camera outside this building, there's the sky, the clouds are moving around, the water, cars driving around, people. If you want to predict everything that'll happen in the future, you can do so in many different ways. You can say, "Okay, there's people around. Let me get really good at understanding the psychology of how people behave in crowds and predict the pedestrians." But you could also say, "Well, there's clouds moving around. Let me understand everything about water molecules and ice particles in the air." You could go super deep on that. If you want to fully understand down to the subatomic level everything that's going on, as a person you could spend decades just thinking about that and you'll never even get to the pedestrians or the water.

If you want to really predict everything that's going on in that scene, there's just so much stuff that even if you're doing a really great job and capturing 100% of something, by the time you get to everything else, ages will have passed. Whereas with text, it's already been abstracted into those bits that we as humans care about. The representations are already there. They're not just good representations, they focus on what really matters.

That's the bad news. Here's the good news. The good news is that we don't have to just get everything out of pointing a camera outside this building. When you have a robot, that robot is trying to do a job. It has a purpose, and its perception is in service to fulfilling that purpose. That is a really great focusing factor.

We know that for people, this really matters. Literally what you see is affected by what you're trying to do. There's been no shortage of psychology experiments showing that people have almost a shocking degree of tunnel vision where they will literally not see things right in front of their eyes if it's not relevant to what they're trying to achieve. That is tremendously powerful. There must be a reason why people do that. Certainly if you're out in the jungle, seeing more is better than seeing less. If you have that powerful focusing mechanism, it must be darn important for getting you to achieve your goal. Robots will have that focusing mechanism because they're trying to achieve a goal.

Dwarkesh Patel

The fact that video models aren't as robust, is that bearish for robotics? So much of the data you will have to use… I guess you're saying a lot of it will be labeled. Ideally, you just want to be able to throw everything on YouTube, every video we've ever recorded, and have it learn how the physical world works and how to move about. Just see humans performing tasks and learn from that. I guess you're saying it's hard to learn just from that and it needs to practice the task itself.

Sergey Levine

Let me put it this way. Let's say that I gave you lots of videotapes or lots of recordings of different sporting events and gave you a year to just watch sports. After that year, I told you, "Okay, now your job, you're going to be playing tennis." Okay, that's pretty dumb right? Whereas if I told you first you're going to be playing tennis and then I let you study up, now you really know what you're looking for.

There's a very real challenge here. I don't want to understate the challenge. But there's also a lot of potential for foundation models that are embodied, that learn from interaction, from controlling robotic systems, to be better at absorbing the other data sources because they know what they're trying to do. I don't think that by itself is a silver bullet. I don't think it solves everything, but it does help a lot. We've already seen the beginnings of that where we can see that including web data in training for robots really does help with generalization. I have the suspicion that in the long run, it'll make it easier to use those sources of data that have been tricky to use up until now.

Dwarkesh Patel

Famously, LLMs have all these emergent capabilities that were never engineered in, because somewhere in internet text is the data to train and to be able to give it the knowledge to do a certain kind of thing. With robots, it seems like you are collecting all the data manually. So there won't be this mysterious new capability that is somewhere in the dataset that you haven't purposefully collected.

Which seems like it should make it even harder to then have robust, out-of-distribution capabilities. I wonder if the trek over the next 5-10 years will be like this: Each subtask, you have to give it thousands of episodes. Then it's very hard to actually automate much work just by doing subtasks. If you think about what a barista does, what a waiter does, what a chef does, very little of it involves just sitting at one station and doing stuff. You got to move around, you got to restock, you got to fix the machine, et cetera, go between the counter and the cashier and the machine, etc.

Will there just be this long tail of things and skills that you have to keep adding episodes for manually and labeling and seeing how well they did? Or is there some reason to think that it will progress more generally than that?

Sergey Levine

There's a subtlety here. Emergent capabilities don't just come from the fact that internet data has a lot of stuff in it. They also come from the fact that generalization, once it reaches a certain level, becomes compositional. There was a cute example that one of my students really liked to use in some of his presentations. You know what the International Phonetic Alphabet (IPA) is?

Dwarkesh Patel

No.

Sergey Levine

If you look in a dictionary, they'll have the pronunciation of a word written in funny letters. That's basically International Phonetic Alphabet. It's an alphabet that is pretty much exclusively used for writing down pronunciations of individual words and dictionaries. You can ask an LLM to write you a recipe for making some meal in International Phonetic Alphabet, and it will do it. That's like, holy crap. That is definitely not something that it has ever seen because IPA is only ever used for writing down pronunciations of individual words.

That's compositional generalization. It's putting together things you've seen in new ways. Arguably there's nothing profoundly new here because yes, you've seen different words written that way, but you've figured out that now you can compose the words in this other language the same way that you've composed words in English. That's actually where the emergent capabilities come from. Because of this, in principle, if we have a sufficient diversity of behaviors, the model should figure out that those behaviors can be composed in new ways as the situation calls for it.

We've actually seen things even with our current models. In the grand scheme of things, looking back five years from now, we'll probably think that these are tiny in scale. But we've already seen what I would call emerging capabilities. When we were playing around with some of our laundry folding policies, we actually discovered this by accident. The robot accidentally picked up two T-shirts out of the bin instead of one. It starts folding the first one, the other one gets in the way, picks up the other one, throws it back in the bin. We didn't know it would do that. Holy crap.

Then we tried to play around with it, and yep, it does that every time. It's doing its work. Drop something else on the table, it just picks it up and puts it back. Okay, that's cool. It starts putting things in a shopping bag. The shopping bag tips over, it picks it back up, and stands it upright. We didn't tell anybody to collect data for that. I'm sure somebody accidentally at some point, or maybe intentionally picked up the shopping bag.

You just have this kind of compositionality that emerges when you do learning at scale. That's really where all these remarkable capabilities come from. Now you put that together with language. You put that together with all sorts of chain-of-thought reasoning, and there's a lot of potential for the model to compose things in new ways.

Dwarkesh Patel

Right. I had an example like this when I got a tour of the robots at your office. It was folding shorts. I don't know if there was an episode like this in the training set, but just for fun I took one of the shorts and turned it inside out. Then it was able to understand that it first needed to get… First of all, the grippers are just like this, two opposable finger and thumb-like things. It's actually shocking how much you can do with just that. But it understood that it first needed to fold it inside out before folding it correctly.

What's especially surprising about that is it seems like this model only has one second of context. Language models can often see the entire codebase. They're observing hundreds of thousands of tokens and thinking about them before outputting. They're observing their own chain of thought for thousands of tokens before making a plan about how to code something up. Your model is seeing one image, what happened in the last second, and it vaguely knows it's supposed to fold this short. It's seeing the image of what happened in the last second. I guess it works.

It's crazy that it will just see the last thing that happened and then keep executing on the plan. Fold it inside out, then fold it correctly. But it's shocking that a second of context is enough to execute on a minute-long task. Yeah. I'm curious why you made that choice in the first place and why it's possible to actually do tasks… If a human only had a second of memory and had to do physical work, I feel like that would just be impossible.

Sergey Levine

It's not that there's something good about having less memory, to be clear. Adding memory, adding longer context, all that stuff, adding higher resolution images, those things will make the model better. But the reason why it's not the most important thing for the kind of skills that you saw when you visited us, at some level, comes back to Moravec's paradox.

Moravec's paradox basically, if you want to know one thing about robotics, that's the thing. Moravec's paradox says that in AI the easy things are hard and the hard things are easy. Meaning the things that we take for granted—like picking up objects, seeing, perceiving the world, all that stuff—those are all the hard problems in AI. The things that we find challenging, like playing chess and doing calculus, actually are often the easier problems.

I think this memory stuff is actually Moravec’s paradox in disguise. We think that the cognitively demanding tasks that we do that we find hard, that cause us to think, "Oh man, I'm sweating. I'm working hard." Those are the ones that require us to keep lots of stuff in memory, lots of stuff in our minds. If you're solving some big math problem, if you're having a complicated technical conversation on a podcast, those are things where you have to keep all those puzzle pieces in your head.

If you're doing a well-rehearsed task—if you are an Olympic swimmer and you're swimming with perfect form—and you're right there in the zone, people even say it's "in the moment." It's in the moment. It's like you've practiced it so much you've baked it into your neural network in your brain. You don't have to think carefully about keeping all that context. It really is just Moravec's paradox manifesting itself.

That doesn't mean that we don't need the memory. It just means that if we want to match the level of dexterity and physical proficiency that people have, there's other things we should get right first and then gradually go up that stack into the more cognitively demanding areas, into reasoning, into context, into planning, all that kind of stuff. That stuff will be important too.

Dwarkesh Patel

You have this trilemma. You have three different things which all take more compute during inference that you want to increase at the same time. You have the inference speed. Humans are processing 24 frames a second or whatever it is. We can react to things extremely fast. Then you have the context length. For the kind of robot which is just cleaning up your house, I think it has to be aware of things that happened minutes ago or hours ago and how that influences its plan about the next task it's doing. Then you have the model size. At least with LLMs, we've seen that there's gains from increasing the amount of parameters.

I think currently you have 100 millisecond inference speeds. You have a second-long context and then the model is a couple billion parameters? Each of these, at least two of them, are many orders of magnitude smaller than what seems to be the human equivalent. A human brain has trillions of parameters and this has like 2 billion parameters. Humans are processing at least as fast as this model, actually a decent bit faster, and we have hours of context. It depends on how you define human context, but hours of context, minutes of context.

Sergey Levine

Sometimes decades of context.

Dwarkesh Patel

Exactly. You have to have many order-of-magnitude improvements across all of these three things which seem to oppose each other. Increasing one reduces the amount of compute you can dedicate towards the other one in inference. How are we going to solve this?

Sergey Levine

That's a very big question. Let's try to unpack this a little bit. There's a lot going on in there. One thing is a really interesting technical problem. It's something where we'll see perhaps a lot of really interesting innovation over the next few years. It’s the question of representation for context.

You gave some of the examples, like if you have a home robot that's doing something then it needs to keep track. As a person, there are certainly some things where you keep track of them very symbolically, almost in language. I have my checklist. I'm going shopping. At least for me, I can literally visualize in my mind my checklist. Pick up the yogurt, pick up the milk, pick up whatever. I'm not picturing the milk shelf with the milk sitting there. I'm just thinking, “milk.”

But then there's other things that are much more spatial, almost visual. When I was trying to get to your studio, I was thinking, "Okay, here's what the street looks like. Here's what that street looks like. Here's what I expect the doorway to look like."

Representing your context in the right form, that captures what you really need to achieve your goal—and otherwise discards all the unnecessary stuff—I think that's a really important thing. We're seeing the beginnings of that with multimodal models. But I think that multimodality has much more to it than just image plus text. That's a place where there's a lot of room for really exciting innovation.

Dwarkesh Patel

Do you mean in terms of how we represent?

Sergey Levine

How we represent both context, both what happened in the past, and also plans or reasoning, as you call it in the LLM world, which is what we would like to happen in the future or intermediate processing stages in solving a task. Doing that in a variety of modalities, including potentially learned modalities that are suitable for the job, is something that has enormous potential to overcome some of these challenges.

Dwarkesh Patel

Interesting. Another question I have as we're discussing these tough trade-offs in terms of inference is comparing it to the human brain. The human brain is able to have hours, decades of context while being able to act on the order of 10 milliseconds, while having 100 trillion parameters or however you want to count it. I wonder if the best way to understand what's happening here is that human brain hardware is just way more advanced than the hardware we have with GPUs, or that the algorithms for encoding video information are way more efficient. Maybe it's some crazy mixture of experts where the active parameters are also on the order of billions, low billions. Or it’s some mixture of the two.

If you had to think about why we have these models that are, across many dimensions, orders of magnitude less efficient compared to the brain, is it hardware or algorithms?

Sergey Levine

That's a really good question. I definitely don't know the answer to this. I am not by any means well-versed in neuroscience. If I had to guess and also provide an answer that leans more on things I know, it's something like this. The brain is extremely parallel. It has to be just because of the biophysics, but it's even more parallel than your GPU.

If you think about how a modern multimodal language model processes the input, if you give it some images and some text, first it reads in the images, then it reads in the text, and then proceeds one token at a time to generate the output. It makes a lot more sense to me for an embodied system to have parallel processes. Now mathematically you can make close equivalences between parallel and sequential stuff. Transformers aren't fundamentally sequential. You make them sequential by putting in position embeddings. Transformers are fundamentally very parallelizable things. That's what makes them so great.

I don't think that mathematically this highly parallel thing—where you're doing perception and proprioception and planning all at the same time—necessarily needs to look that different from a transformer, although its practical implementation will be different. You could imagine that the system will in parallel think about, "Okay, here's my long-term memory, here's what I've seen a decade ago, here's my short-term spatial stuff, here's my semantic stuff, here's what I'm seeing now, here's what I'm planning." All of that can be implemented in a way that there's some very familiar attentional mechanism, but in practice all running in parallel, maybe at different rates, maybe with the more complex things running slower, the faster reactive stuff running faster.

Dwarkesh Patel

If in five years we have a system which is as robust as a human in terms of interacting with the world, then what has happened that makes it physically possible to be able to run those models? To have video information that is streaming at real time, or hours of prior video information is somehow being encoded and considered while decoding in a millisecond scale, and with many more parameters. Is it just that Nvidia has shipped much better GPUs or that you guys have come up with much better encoders and stuff? What's happened in the five years?

Sergey Levine

There are a lot of things to this question. Certainly there's a really fascinating systems problem. I'm by no means a systems expert. I would imagine that the right architecture in practice, especially if you want an affordable low-cost system, would be to externalize at least part of the thinking. You could imagine in the future you'll have a robot where, if your Internet connection is not very good, the robot is in a dumber reactive mode. But if you have a good Internet connection then it can be a little smarter. It's pretty cool.

There is also research and algorithms stuff that can help here, figuring out the right representations, concisely representing both your past observations but also changes in observation. Your sensory stream is extremely temporally correlated. The marginal information gained from each additional observation is not the same as the entirety of that observation. The image that I'm seeing now is very correlated to the image I saw before. In principle, I want to represent it concisely. I could get away with a much more compressed representation than if I represent the images independently. There's a lot that can be done on the algorithm side to get this right. That's really interesting algorithms work.

There's also a really fascinating systems problem. To be truthful, I haven't gotten to the systems problem because you want to implement the system once you know the shape of the machine learning solution. But there's a lot of cool stuff to do there.

Dwarkesh Patel

Maybe you guys just need to hire the people who run the YouTube data centers because they know how to encode video information.

This raises an interesting question. With LLMs, theoretically you could run your own model on this laptop or whatever. Realistically what happens is that the largest, most effective models are being run in batches of thousands and millions of users at the same time, not locally. Will the same thing happen in robotics because of the inherent efficiencies of batching, plus the fact that we have to do this incredibly compute-intensive inference task? You don't want to be carrying around $50,000 GPUs per robot or something. You just want that to happen somewhere else.

In this robotics world, should we just be anticipating something where you need connectivity everywhere? You need robots that are super fast. You're streaming video information back and forth, or at least video information one way. Does that have interesting implications about how this deployment of robots will be instantiated?

Sergey Levine

I don't know. But if I were to guess, I would guess that we'll see both. That we'll see low-cost systems with off-board inference and more reliable systems. For example, in settings where you have an outdoor robot or something where you can't rely on connectivity, those will be costlier and have onboard inference.

I'll say a few things from a technical standpoint that might contribute to understanding this. While a real-time system obviously needs to be controlled in real time, often at high frequency, the amount of thinking you need to do for every time step might be surprisingly low. Again, we see this in humans and animals. When we plan out movements, there is definitely a real planning process that happens in the brain. If you record from a monkey brain, you will find neural correlates of planning. There is something that happens in advance of a movement. When that movement takes place, the shape of the movement correlates with what happened before the movement. That's planning.

That means that you put something in place and set the initial conditions of some process and then unroll that process, and that's the movement. That means that during that movement, you're doing less processing and you batch it up in advance. But you're not entirely an open loop. It's not that you're playing back a tape recorder. You are reacting as you go. You're just reacting at a different level of abstraction, a more basic level of abstraction.

Again, this comes back to representations. Figure out which representati ons are sufficient for planning in advance and then unrolling, and which representations require a tight feedback loop. For that tight feedback loop, what are you doing feedback on? If I'm driving a vehicle, maybe I'm doing feedback on the position of the lane marker so that I stay straight. At a lower frequency, I sort of gauge where I am in traffic.

Dwarkesh Patel

You have a couple of lectures from a few years back where you say that even for robotics, RL is in many cases better than imitation learning. But so far the models are exclusively doing imitation learning. I'm curious how your thinking on this has changed. Maybe it hasn’t changed. But then you need to do this for the RL. Why can't you do RL yet?

Sergey Levine

The key here is prior knowledge. In order to effectively learn from your own experience, it turns out that it's really, really important to already know something about what you're doing. Otherwise it takes far too long, just like it takes a person, when they're a child, a very long time to learn very basic things, to learn to write for the first time, for example. Once you already have some knowledge, then you can learn new things very quickly. The purpose of training the models with supervised learning now is to build out that foundation that provides the prior knowledge so they can figure things out much more quickly later.

Again, this is not a new idea. This is exactly what we've seen with LLMs. LLMs start off being trained purely with next token prediction. That provided an excellent starting point, first for all sorts of synthetic data generation and then for RL. It makes total sense that we would expect basically any foundation model effort to follow that same trajectory. We first build out the foundation essentially in a somewhat brute-force way. The stronger that foundation gets, the easier it is to then make it even better with much more accessible training.

Dwarkesh Patel

In 10 years, will the best model for knowledge work also be a robotics model or have an action expert attached to it? The reason I ask is, so far we've seen advantages from using more general models for things. Will robotics fall into this bucket? Will we just have the model which does everything, including physical work and knowledge work, or do you think they'll continue to stay separate?

Sergey Levine

I really hope that they will actually be the same. Obviously I'm extremely biased. I love robotics, I think it's very fundamental to AI. But optimistically, I hope it's actually the other way around, that the robotics element of the equation will make all the other stuff better.

There are two reasons for this that I can tell you about. One has to do with representations and focus. What I said before, with video prediction models if you just want to predict everything that happens, it's very hard to figure out what's relevant. If you have the focus that comes from trying to do a task now that acts to structure how you see the world in a way that allows you to more fruitfully utilize the other signals. That could be extremely powerful.

The second one is that understanding the physical world at a very deep, fundamental level, at a level that goes beyond just what we can articulate with language, can help you solve other problems. We experience this all the time. When we talk about abstract concepts, we say, “This company has a lot of momentum.” We'll use social metaphors to describe inanimate objects. “My computer hates me.” We experience the world in a particular way and our subjective experience shapes how we think about it in very profound ways. Then we use that as a hammer to basically hit all sorts of other nails that are far too abstract to handle any other way.

Dwarkesh Patel

There might be other considerations that are relevant to physical robots in terms of inference speed and model size, et cetera, which might be different from the considerations for knowledge work. Maybe it's still the same model, but then you can serve it in different ways. The advantages of co-training are high enough.

I'm wondering, in five years if I'm using a model to code for me, does it also know how to do robotics stuff? Maybe the advantages of code writing on robotics are high enough that it's worth it.

Sergey Levine

The coding is probably the pinnacle of abstract knowledge work in the sense that just by the mathematical nature of computer programming, it's an extremely abstract activity, which is why people struggle with it so much.

Dwarkesh Patel

I'm a bit confused about why simulation doesn't work better for robots. If I look at humans, smart humans do a good job of, if they're intentionally trying to learn, noticing what about the simulation is similar to real life and paying attention to that and learning from that.

If you have pilots who are learning in simulation or F1 drivers who are learning in simulation, should we expect it to be the case that as robots get smarter they will also be able to learn more things through simulation? Or is this cursed and we need real-world data forever?

Sergey Levine

This is a very subtle question. Your example with the airplane pilot using simulation is really interesting. But something to remember is that when a pilot is using a simulator to learn to fly an airplane, they're extremely goal-directed. Their goal in life is not to learn to use a simulator. Their goal in life is to learn to fly the airplane. They know there will be a test afterwards. They know that eventually they'll be in charge of a few hundred passengers and they really need to not crash that thing.

When we train models on data from multiple different domains, the models don't know that they're supposed to solve a particular task. They just see, "Hey, here's one thing I need to master. Here's another thing I need to master." Maybe a better analogy there is if you're playing a video game where you can fly an airplane and then eventually someone puts you in the cockpit of a real one. It's not that the video game is useless, but it's not the same thing. If you're trying to play that video game and your goal is to really master the video game, you're not going to go about it in quite the same way.

Dwarkesh Patel

Can you do some kind of meta-RL on this? There's this really interesting paper you wrote in 2017. Maybe the loss function is not how well it does at a particular video game or particular simulation. I'll let you explain it. But it was about how well being trained at different video games makes it better at some other downstream task. I did a terrible job at explaining but can you do a better job and try to explain what I was trying to say?

Sergey Levine

What you're trying to say is that maybe if we have a really smart model that's doing meta-learning, perhaps it can figure out that its performance on a downstream problem, a real-world problem, is increased by doing something in a simulator.

Dwarkesh Patel

And then specifically make that the loss function, right?

Sergey Levine

That's right. But here's the thing with this. There's a set of these ideas that are all going to be something like, “Train to make it better on the real thing by leveraging something else.” The key linchpin for all of that is the ability to train it to be better on the real thing. I suspect in reality we might not even need to do something quite so explicit. Meta learning is emergent, as you pointed out before. LLMs essentially do a kind of meta learning via in-context learning. We can debate how much that's learning or not, but the point is that large powerful models trained on the right objective and on real data, get much better at leveraging all the other stuff. I think that's actually the key.

Coming back to your airplane pilot, the airplane pilot is trained on a real world objective. Their objective is to be a good airplane pilot, to be successful, to have a good career. All of that kind of propagates back into the actions they take and leveraging all these other data sources. So what I think is actually the key here to leveraging auxiliary data sources including simulation, is to build the right foundation model that is really good and has those emergent abilities. To your point, to get really good like that, it has to have the right objective. Now we know how to get the right objective out of real world data, maybe we can get it out of other things, but that's harder right now.

Again, we can look to the examples of what happened in other fields. These days if someone trains an LLM for solving complex problems, they're using lots of synthetic data. The reason they're able to leverage that synthetic data effectively is because they have this starting point that is trained on lots of real data that gets it. Once it gets it, then it's more able to leverage all this other stuff. Perhaps ironically, the key to leveraging other data sources including simulation, is to get really good at using real data, understand what's up with the world, and then you can fruitfully use all this stuff.

Dwarkesh Patel

Once we have, in 2035 or 2030, basically this sci-fi world, are you optimistic about the ability of true AGIs to build simulations in which they are rehearsing skills that no human or AI has ever had a chance to practice before? They need to practice to be astronauts because we're building the Dyson sphere and they can just do that in simulation. Or will the issue with simulation continue to be one regardless of how smart the models get?

Sergey Levine

Here’s what I would say. Deep down at a very fundamental level, the synthetic experience that you create yourself doesn't allow you to learn more about the world. It allows you to rehearse things, it allows you to consider counterfactuals. But somehow information about the world needs to get injected into the system. The way you pose this question elucidates this very nicely. In robotics classically, people have often thought about simulation as a way to inject human knowledge. A person knows how to write down differential equations, they can code it up and that gives the robot more knowledge than it had before.

But increasingly what we're learning from experiences in other fields, from how the video generation stuff goes from synthetic data for LLMs, is that probably the most powerful way to create synthetic experience is from a really good model. The model probably knows more than a person does about those fine-grained details. But then of course, where does that model get the knowledge? From experiencing the world.

In a sense, what you said is quite right in that a very powerful AI system can simulate a lot of stuff. But also at that point it almost doesn't matter because, viewed as a black box, what's going on with that system is that information comes in and capability comes out. Whether the way to process that information is by imagining some stuff and simulating or by some model-free method is kind of irrelevant in our understanding of its capabilities.

Dwarkesh Patel

Do you have a sense of what the equivalent is in humans? Whatever we're doing when we're daydreaming or sleeping. I don't know if you have some sense of what this auxiliary thing we're doing is, but if you had to make an ML analogy, what is it?

Sergey Levine

Certainly when you sleep your brain does stuff that looks an awful lot like what it does when it's awake. It looks an awful lot like playing back experience or perhaps generating new statistically similar experience. It's very reasonable to guess that perhaps simulation through a learned model is part of how your brain figures out counterfactuals, basically.

Something that's even more fundamental than that is that optimal decision making at its core, regardless of how you do it, requires considering counterfactuals. You basically have to ask yourself, “If I did this instead of that, would it be better?” You have to answer that question somehow. Whether you answer that question by using a learned simulator, or whether you answer that question by using a value function or something, by using a reward model, in the end it's all the same. As long as you have some mechanism for considering counterfactuals and figuring out which counterfactual is better, you've got it. I like to think about it this way because it simplifies things. It tells us that the key is not necessarily to do really good simulations. The key is to figure out how to answer counterfactuals.

Dwarkesh Patel

Yeah, Interesting. Stepping into the big picture again. The reason I'm interested in getting a concrete understanding of when this robot economy will be deployed is because it's relevant to understanding how fast AGI will proceed in the sense that it's obviously about the data flywheel.

But also, if you just extrapolate out the capex for AI by 2030, people have different estimates, but many people have estimates in the hundreds of gigawatts – 100, 200, 300 gigawatts. You can just crunch numbers on having 100-200 gigawatts deployed by 2030. The marginal capex per year is in the trillions of dollars. It's $2-4 trillion dollars a year. That corresponds to actual data centers you have to build, actual chip foundries you have to build, actual solar panel factories you have to build.

I am very curious about whether by 2030, the big bottleneck is just the people to lay out the solar panels next to the data center or assemble the data center, or will the robot economy be mature enough to help significantly in that process.

Sergey Levine

That's cool. You're basically saying, how much concrete should I buy now to build the data center so that by 2030 I can power all the robots. That is a more ambitious way of thinking about it than has occurred to me, but it's a cool question. The good thing, of course, is that the robots can help you build that stuff.

Dwarkesh Patel

But will they be able to by that time? There's the non-robotic stuff, which will also mandate a lot of capex. Then there's robot stuff where you have to build robot factories, etc. There will be this industrial explosion across the whole stack. How much will robotics be able to speed that up or make it possible?

Sergey Levine

In principle, quite a lot. We have a tendency sometimes to think about robots as mechanical people, but that's not the case. People are people and robots are robots. The better analogy for the robot, it's like your car or a bulldozer. It has much lower maintenance requirements. You can put them into all sorts of weird places and they don't have to look like people at all. You can make a robot that's 100 feet tall. You can make a robot that's tiny.

If you have the intelligence to power very heterogeneous robotic systems, you can probably do a lot better than just having mechanical people, in effect. It can be a big productivity boost for real people and it can allow you to solve problems that are very difficult to solve. For example, I'm not an expert on data centers by any means, but you could build your data centers in a very remote location because the robots don't have to worry about whether there's a shopping center nearby.

Dwarkesh Patel

There's the question of where the software will be, and then there's the question of how many physical robots we will have. How many of the robots you're training in Physical Intelligence, these tabletop arms, are there physically in the world? How many will there be by 2030? These are tough questions, how many will be needed for the intelligence explosion.

Sergey Levine

These are very tough questions. Also, economies of scale in robotics so far have not functioned the same way that they probably would in the long term. Just to give you an example, when I started working in robotics in 2014, I used a very nice research robot called a PR2 that cost $400,000 to purchase. When I started my research lab at UC Berkeley, I bought robot arms that were $30,000. The robots that we are using now at Physical Intelligence, each arm costs about $3,000. We think they can be made for a small fraction of that.

Dwarkesh Patel

What is the cause of that learning rate?

Sergey Levine

There are a few things. One, of course, has to do with economies of scale. Custom-built, high-end research hardware, of course, is going to be much more expensive than more productionized hardware. Then of course, there's a technological element. As we get better at building actuated machines, they become cheaper.

There's also a software element. The smarter your AI system gets, the less you need the hardware to satisfy certain requirements. Traditional robots in factories need to make motions that are highly repeatable. Therefore it requires a degree of precision and robustness that you don't need if you can use cheap visual feedback. AI also makes robots more affordable and lowers the requirements on the hardware.

Dwarkesh Patel

Interesting. Do you think the learning rate will continue? Do you think it will cost hundreds of dollars by the end of the decade to buy mobile arms?

Sergey Levine

That is a great question for my co-founder, Adnan Esmail, who is probably the best person arguably in the world to ask that question. Certainly the drop in cost that I've seen has surprised me year after year.

Dwarkesh Patel

How many arms are there probably in the world? Is it more than a million? Less than a million?

Sergey Levine

I don't know the answer to that question, but it's also a tricky question to answer because not all arms are made equal. Arguably, the robots that are assembling cars in a factory are just not the right kind to think about.

Dwarkesh Patel

The kind you want to train on.

Sergey Levine

Very few because they are not currently commercially deployed as factory robots.

Dwarkesh Patel

Less than 100,000?

Sergey Levine

I don't know, but probably.

Dwarkesh Patel

Okay. And we want billions of robots, at least millions of robots. If you're just thinking about the industrial explosion that you need to get this explosive AI growth, not only do you need the arms, but you need something that can move around. Basically, I'm just trying to think whether that will be possible by the time that you need a lot more labor to power this AI boom?

Sergey Levine

Well, economies are very good at filling demand when there's a lot of demand. How many iPhones were in the world in 2001? There's definitely a challenge there. It's something that is worth thinking about. A particularly important question for researchers like myself is how can AI affect how we think about hardware?

There are some things that are going to be really, really important. You probably want your thing to not break all the time. There are some things that are firmly in that category of question marks. How many fingers do we need? You said yourself before that you were surprised that a robot with two fingers can do a lot. Maybe you still want more than that, but still finding the bare minimum that still lets you have good functionality, that's important. That's in the question mark box.

There are some things that we probably don't need. We probably don't need the robot to be super duper precise, because we know that feedback can compensate for that. My job, as I see it right now, is to figure out what's the minimal package we can get away with. I really think about robots in terms of minimal package because I don't think that we will have the one ultimate robot, the mechanical person basically.

What we will have is a bunch of things that good, effective robots need to satisfy. Just like good smartphones need to have a touchscreen. That's something that we all agreed on. Then they’ll need a bunch of other stuff that's optional, depending on the need, depending on the cost point, et cetera. There will be a lot of innovation where once we have very capable AI systems that can be plugged into any robot to endow it with some basic level of intelligence, then lots of different people can innovate on how to get the robot hardware to be optimal for each niche.

Dwarkesh Patel

In terms of manufacturers, is there some Nvidia of robotics?

Sergey Levine

Not right now. Maybe there will be someday. Maybe I'm being idealistic, but I would really like to see a world where there's a lot of heterogeneity in robots.

Dwarkesh Patel

What is the biggest bottleneck in the hardware today as somebody who's designing the algorithms that run on it?

Sergey Levine

It's a tough question to answer, mainly because things are changing so fast. To me, the things that I spend a significant amount of time thinking about on the hardware side is really more reliability and cost. It's not that I'm that worried about cost. It's just that cost translates to the number of robots, which translates to the amount of data. Being an ML person, I really like having lots of data. I really want to have robots that are low cost, because then I can have more of them and therefore more data. Reliability is important, more or less for the same reason.

It's something that we'll get more clarity on as things progress. Basically, the AI systems of today are not pushing the hardware to the limit. As the AI systems get better and better, the hardware will get pushed to the limit, and then we'll hopefully have a much better answer to your question.

Dwarkesh Patel

This is a question I've had for a lot of guests. If you go through any layer of this AI explosion, you find that a bunch of the actual source supply chain is being manufactured in China, other than chips obviously. You talk about data centers and you're like, “Oh, all the wafers for solar panels and a bunch of the cells and modules, et cetera, are manufactured in China.” You just go through the supply chain. Obviously robot arms are being manufactured in China.

You’ll live in this world where it’s just incredibly valuable to ramp up manufacturing of the hardware, because each robot can produce some fraction of the value that a human worker can produce. Not only is that true, but the value of human workers or any worker has tremendously skyrocketed because we need tons of bodies to lay out the tens of thousands of acres of solar farms and data centers and foundries and everything.

In this boom world, the big bottleneck there's just how many robots can you physically deploy? How many can you manufacture? Because you guys are going to come up with the algorithms now. We just need the hardware. This is a question I've asked many guests. If you look at the part of the chain that you are observing, what is the reason that China just doesn't win by default? If they're producing all the robots and you come up with the algorithms that make those robots super valuable, why don't they just win by default?

Sergey Levine

This is a very complex question. I'll start with the broader themes and then try to drill a little bit into the details. One broader theme here is that if you want to have an economy where you get ahead by having a highly educated workforce—by having people that have high productivity, meaning that for each person's hour of work, lots of stuff gets done—automation is really, really good. Automation is what multiplies the amount of productivity that each person has. Again, it’s the same as LLM coding tools. LLM coding tools amplify the productivity of a software engineer. Robots will amplify the productivity of basically everybody that is doing work.

Now that's a final state, a desirable final state. There's a lot of complexity in how you get to that state, how you make that an appealing journey to society, how you navigate the geopolitical dimension of that. All of that stuff is pretty complicated. It requires making a number of really good decisions. Good decisions about investing in a balanced robotics ecosystem, supporting both software innovation and hardware innovation. I don't think any of those are insurmountable problems. It just requires a degree of long-term vision and the right balance of investment.

What makes me really optimistic about this is the final state. We can all agree that in the United States we would like to have a society where people are highly productive, where we have highly educated people doing high-value work. Because that end state seems to me very compatible with automation, with robotics, at some level there should be a lot of incentive to get to that state.

Then from there we have to solve for all the details that will help us get there. That's not easy. There's a lot of complicated decisions that need to be made in terms of private industry, in terms of investment, in terms of the political dimension. But I'm very optimistic about it because it seems to me that the light at the end of the tunnel is in the right direction.

Dwarkesh Patel

I guess there's a different question. If the value is bottlenecked by hardware and you just need to produce more hardware, what is the path by which hundreds of millions of robots or billions of robots are being manufactured in the US or with allies? I don't know how to approach that question, but it seems like a different question than, “Well, what is the impact on human wages or something?”

Sergey Levine

For the specifics of how we make that happen, that's a very long conversation that I'm probably not the most qualified to speak to. But in terms of the ingredients, the ingredient here that is important is that robots help with physical things, physical work. If producing robots is itself physical work, then getting really good at robotics should help with that.

It's a little circular, of course, and as with all circular things, you have to bootstrap it and try to get that engine going. But it seems like it is an easier problem to address than, for example, the problem of digital devices. Work goes into creating computers, phones, et cetera. But the computers and phones don't themselves help with the work.

Dwarkesh Patel

Right. I guess feedback loops go both ways. They can help you or they can help others and it's a positive sum world. It's not necessarily bad that they help others. But to the extent that a lot of the things which would go into this feedback loop—the sub-component, manufacturing and supply chain, already exist in China—it seems like the stronger feedback loop would exist in China. Then there's a separate discussion. Maybe that's fine, maybe that's good, and maybe they'll continue exporting this to us. But I just find it notable that whenever I talk to guests about different things, it's just like, “Yeah, within a few years the key bottleneck to every single part of the supply chain here will be something that China is the 80% world supplier of.”

Sergey Levine

This is why I said before that something really important to get right here is a balanced robotics ecosystem. AI is tremendously exciting, but we should also recognize that getting AI right is not the only thing that we need to do. We need to think about how to balance our priorities, our investment, the kind of things that we spend our time on. Just as an example, at Physical Intelligence we do take hardware very seriously. We build a lot of our own things and we want to have a hardware roadmap alongside our AI roadmap. But that's just us.

For the United States, arguably for human civilization as a whole, we need to think about these problems very holistically. It is easy to get distracted sometimes when there's a lot of excitement, a lot of progress in one area like AI. We are tempted to lose track of other things, including things you've said. There's a hardware component. There's an infrastructure component with compute and things like that. In general it's good to have a more holistic view of these things. I wish we had more holistic conversations about that sometimes.

Dwarkesh Patel

From the perspective of society as a whole, how should they be thinking about the advances in robotics and knowledge work? Basically society should be planning for full automation. There will be a period in which people's work is way more valuable because there's this huge boom in the economy where we’re building all these data centers and factories. Eventually humans can do things with their body and we can do things with our mind. There's not some secret third thing.

What should society be planning for? It should be full automation of humans. Society will also be much wealthier. Presumably there are ways to do this such that everybody is much better off than they are today. But the end state, the light at the end of the tunnel, is the full automation but plus super wealthy society with some redistribution or whatever way to figure that out. I don't know if you disagree with that characterization.

Sergey Levine

At some level that's a very reasonable way to look at things. But if there's one thing that I've learned about technology, it's that it rarely evolves quite the way that people expect. Sometimes the journey is just as important as the destination. It's very difficult to plan ahead for an end state. Directionally, what you said makes a lot of sense. I do think that it's very important for us collectively to think about how to structure the world around us in a way that is amenable to greater and greater automation across all sectors. But we should really think about the journey just as much as the destination, because things evolve in all sorts of unpredictable ways. We'll find automation showing up in all sorts of places, probably not the places we expect first.

The constant here that is really important is that education is really, really valuable. Education is the best buffer somebody has against the negative effects of change. If there is one single lever that we can pull collectively as a society, it's more education.

Dwarkesh Patel

Is that true? Moravec's paradox is that the things which are most beneficial from education for humans might be the easiest to automate because it's really easy to educate AIs. You can throw the textbooks that would take you eight years of grad school to do at them in an afternoon.

Sergey Levine

What education gives you is flexibility. It's less about the particular facts you know, as it is about your ability to acquire skills, acquire understanding. It has to be a good education.

Dwarkesh Patel

Yeah. Okay, Sergey, thank you so much for coming on the podcast. Super fascinating.

Sergey Levine

Yeah, this was intense. Tough questions.

In this lecture, military historian Sarah Paine explains how Britain used sea control, peripheral campaigns, and alliances to defeat Nazi Germany during WWII. She then applies this framework to today, arguing that Russia and China are similarly constrained by their geography, making them vulnerable in any conflict with maritime powers (like the U.S. and its allies).

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Labelbox partners with researchers to scope, generate, and deliver the exact data frontier models need, no matter the domain. Whether that’s multi-turn audio, action data for robotics, advanced STEM problem sets, or even novel RL environments, Labelbox delivers high-quality data, fast. Learn more at labelbox.com/dwarkesh

• Warp is the best interface I’ve found for coding with agents. It makes building custom tools easy: Warp’s UI helps you understand agent behavior and its in-line text editor is great for making tweaks. You can try Warp for free, or, for a limited time, use code DWARKESH to get Warp’s Pro Plan for only $5. Go to warp.dev/dwarkesh

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – How WW1 shaped WW2

(00:15:10) – Hitler and Churchill’s battle to command the Atlantic

(00:30:10) – Peripheral theaters leading up to Normandy

(00:37:13) – The Eastern front

(00:48:04) – Russia’s & China’s geographic prisons

(01:00:28) – Hitler’s blunders & America’s industrial might

(01:15:03) – Bismarck’s limited wars vs Hitler’s total war

Sarah Paine 00:00:00

It turns out that the possibilities for maritime and continental powers are a little different. Basically, a small subset of countries can defend themselves primarily at sea. That opens certain possibilities. Others can't. That opens and closes certain possibilities. I'm going to talk about this story from Britain's point of view, the country with the 360° “you can't get me” moat. It's an instructive case for the United States, of the possibilities and the perils of having this sort of position. So that is my game plan today.

You can look at the great peninsula of Europe where Britain is located. You can see this northern coastline for Britain, where it's uncomfortably close to the continent. Its enemies are sitting there and it's an interesting neighborhood. So here's my plan. I'm going to talk about these continental problems that Britain has been dealing with. If you think about it, Britain was always fighting France. In 1871, Germany unifies and then the problem's Germany.

I'm going to pick up the story in 1939 when things are really bad for Britain. I'm first going to talk about these continental problems, and then I'm going to talk about how Britain tried to deal with it. First it has to do with getting sea control, and then once you can do that, finding some peripheral theaters where you might be able to fight and deal with the continental problem. You probably need allies.

So those are the first four topics. That was then and now is now. The continental problems now are China and Russia. We will see what this case study might reveal about the ongoing things.

So here's Britain uncomfortably close to the continent. If it wants to get to Russia, which is its big ally in World War I and World War II, it's either got to go way up north around the Norwegian coastline—you get up into places like Murmansk and Archangel—or it's got to go way around through this very narrow sea, the Mediterranean, through the choke point of choke points, which is the Dardanelles and the Bosporus, into the Black Sea. The main point port back in the day was Odessa.

If you compare French and British access to the high seas, France has got a pretty good coastline that just gets it right into the oceans. Germany, if it wants to send merchant traffic or naval traffic, it's got to go through these narrow seas. Then it's got to get by Britain, which is its big enemy in the two world wars, which is the dominant naval power. So that's complicated.

For Britain, if it wants to get to its empire back in the day, it wants to go through the Suez Canal. That requires the cooperation of Spain, France, Italy. If it wants to get to Russia, it requires Turkey as well. Well, Turkey didn't cooperate very well in either war. If you think about World War II—the Fall of France, fascist sympathies of Spain, Italy as part of the Axis—Britain is in real trouble. What do you do about this? Britain has this big empire that it wants to protect. It's got a massive basing system, more bases than anybody else does in order to protect this empire.

From this, which is a very resentful Germany. Doesn't much like the Versailles settlement of World War I. It's a divided country in that a Polish Corridor separates East Prussia. The Germans start trying to solve this problem. Initially they're taking Austria, the Sudeten German parts of Czechoslovakia in 1938. Then they take all of Czechoslovakia in 1939. They've already dealt with the Rhineland, which is supposed to be demilitarized per the terms of the Versailles Treaty. They ignore that and remilitarize in 1936.

This is important because there are a lot of industrial resources and factories and things there. It turns out that Hitler's plans require—it's not optional— it requires the resources not only of the Rhineland, but also Czechoslovakia and Poland and Romania, which is going to have the oil for them.

You get to 1939, when Russia and Germany are dividing up Poland between them. This is the part of the history that Russians don't like to talk about, but it's exactly what they were doing. This triggers World War II because the French and the British honor their alliance with the Poles to help deal with this.

So in 1940, Britain's in a world of hurt. It faces this massive blue problem. Then there's this growing green disaster. It's facing two continental powers, Germany and Russia, that both have these expansive empires they want to create. They want to divide Europe and then the rest of the world. Bad news for Britain.

But 1941, Hitler decides he needs Russia too. Then Russia decides, “Ah, the one that's attacking me is probably the major problem, not the other one.” So Stalin is going to swap sides and he's going to be coordinating with Britain. That's better from Britain's point of view than having two continental powers to deal with. So now it's down to one.

Once the United States has gotten into the war, this is what the world looks like. You have a big cancer in Europe from the British point of view. But it's all surrounded by oceans that Britain can get to. A lot of allies are neutral states. So Britain has access to those places. Then there's a separate cancer in Asia where Japan's trying to work its magic. I say separate because the Axis never coordinated these two theaters.

So this is looking at the world from a maritime perspective. What you're looking at are all the oceanic routes that connect everything. Britain's problem is how to leverage the miracle of sea transport that basically can access you the whole world, versus a logistical nightmare of land transport where you can only drive through countries that will let you drive through. The seas give you mobility, they give you access to theaters, markets, resources, allies, and. They also give you sanctuary at home. If you're surrounded by them, it makes it harder for people to invade. So Britain's trying to leverage all of that against the armies of the continental armies. It's going to try to strangle them economically, diplomatically and militarily.

Now, the generation that led World War II in Britain—not just Britain but elsewhere—were the conscripts of World War I, which was supposed to end all wars. They knew full well, in the midst of World War II, that it did not remotely achieve that promise. They learned a whole series of lessons. I'm going to do a comparison of what was done in World War I versus World War II. They are the greatest generation, not their children who claim the title.

Lesson number one is don't go beyond the culminating point of attack. What's that? The terminology comes from Carl von Clausewitz, who is the Western guru on conventional land warfare. It means if you're attacking in a battle, if you go too far you'll weaken yourself because the enemy will counterattack and send you further backwards than you would have otherwise been. In the case of World War I, you're sending young men over trenches into ongoing machine gun fire. What do you suppose is going to happen to them? This profligate waste of life in these assaults out of trenches… Maybe you took a little territory in the first two weeks, but after that, nothing. These offensives would go on for months and months, racking up hundreds of thousands of deaths.

No more doing that in World War II. You can look at the death figures for World War I and II. In World War I, the British army gets the multi-million man army that they had coveted. They deploy it on the main front from start to finish and chalk up twice as many deaths as they did in World War II when they have a peripheral strategy.

In World War II, they do make the mistake of landing the big army on the continent as an opening move, but it doesn't do well. Then they reassess and get that army off the continent immediately. This is what the Dunkirk evacuation is. The French are covering the British as they're decamping from the continent, saving the British army. This is why France has such large casualties, even though France isn't in the war that long. It's going to be a long wait before the British get back on the continent. It’s also a long wait also for the United States to get in the war. There's no more going beyond the culminating point of attack.

The way diplomacy is run is also completely different. In World War I there were exactly two conferences trying to coordinate things among the Entente powers—the December 1915 and November 1916 Chantilly Conferences. All they are are the military heads. Well, the Russian Romanov dynasty is overthrown in early 1917. It is too late. What happened in World War I is that the Germans focused on the Western Front in 1914, Eastern Front in 1915, back to the Western Front in 1916. In World War II the idea is you want to squeeze them simultaneously from all fronts so they can't divert people back and forth.

If you look at the coordination, it begins even before the United States is in the war with the ABC staff talks, then the Atlantic Conference which yields the Atlantic Charter, talking about what war objectives are: unconditional surrender of Germany and what the postwar situation is going to look like. There's coordination not only among military leaders, but among civil and military leaders as well. There's a combined command of US and British forces. We have offices in each other's capitals. We're also coordinating with the Russians so that you're setting up not only war termination but also what postwar institutions are going to be like to hold the peace. It's a completely different event from World War I when Russia falls out of the Entente because there are bread riots in St. Petersburg and Russia could not supply its troops with adequate armaments.

Well, that's not going to happen in World War II. Russia comes with a really large army. Germany has another large army. You've got to have a big army to deal with Germany's army. In World War I, there is no Lend-Lease aid. No one would have thought of giving that much stuff to each other. It's like everybody first and then it's like okay, everybody last in that one. .

The railways that had not been completed in World War I. The Trans-Siberian doesn't get completed until 1916, and the Murmansk railway isn't completed until early 1917. The Romanov dynasty is gone. It's too late. In World War II, three-quarters of Lend-Lease aid would go over those completed railway systems.

Now, to the British credit, they did try to break the blockade on Russia. The way to get into Russia in those days and hook in with the railway system would be to get into the Black Sea at Odessa. Russia had all the men, but they didn't have all the war materiel to fight. This is where the Gallipoli campaign comes in. You can argue about whether you think it’s good strategy or not, but it was miserably executed.

First of all, it wasn't a joint operation. What's joint? Joint means you're coordinating your different military services, in this case army and navy. What goes on? The British Navy tries to run the Dardanelles for two months. That does not work well. Do you suppose the Ottomans might think something was up? Yeah. When you get up there on the Dardanelles, it's a very steep place. The Ottomans are all busy sorting that all out, getting troops in place. Two months later, when the British, New Zealand, Australian and French troops all land on a given day, the Ottomans are there with a welcome party, essentially. That invasion stalemates in three days. But they keep at it for eight months, taking 190,000 casualties, 55,000 of whom are dead, totally miserably executed. It comes as collateral damage.

As the British are trying to run the Dardanelles with their navy, this is when the Ottomans are terrified of their Christian subjects, the Armenians. They're starting to round them up, pulling them out of the army. Then days before the landing, the Turkish massacre of the Armenians begins. Between 1915 and 1923, 1.5 million Armenians are killed. That's a lot of collateral damage.

The Normandy landings are a completely different event in World War II. This is another contested landing of trying to get armies in. First of all, the buildup of war materiel goes on for years to get all the landing vessels, the equipment, the forces all ready to go in Britain. Then the disinformation campaign kept the Germans completely misinformed. They were expecting the landing to be at the Pas-de-Calais, which is the shortest place. It's way off in Normandy. That worked. Everybody lands on a day and they're up and over and into places inland.

Another lesson learned. The Royal Navy did not think that convoy duty was the manly thing to do. They would convoy troop transports, but they wouldn't deal with the Merchant Marine until 1918. Well, the Germans almost sank a terminal quantity of that stuff. The Navy is not thinking about the economic dimensions of warfare. They're just focused on all the military things in World War I. In World War II, the British would be convoying even before they got in the war.

Another difference between World War II and World War I. At the end of World War I, if you look at the disposition of German troops, they're abroad. They're occupying Belgium, Luxembourg, and parts of France. Nobody's in Germany. Yes, the Germans had really lousy meals during the war, but German civilians did not feel the full brunt of what their government had done. Therefore Prime Minister Winston Churchill and President Franklin Roosevelt felt that it was really critical to have boots on the ground in Berlin to let the Germans know exactly what had happened to them and let them feel the war that they had inflicted on others in order to end it.

Even so, the Allies win this thing and they wreck the continental powers. But they almost wreck themselves in the process. It's a Pyrrhic victory for France and Britain. It really weakens them. World War II is going to be a different event. That's the lesson from being in this continental situation, the lessons learned from the last time around

Sarah Paine 00:15:10

Now for what the British did in World War II to deal with the continental problem. The opening move of a maritime power in a really high-stakes war like this is typically blockade. What you want to do is cut your enemy off from the oceans and force it to cannibalize its own resources and those of occupied areas. Because of the geographic position of a maritime power, you can quite often do this to a continental power on narrow seas.

The British were well aware that Germany's a trading country. Most of its trade goes by sea and it's also on these narrow seas. Geographically and economically it's really vulnerable to blockade. Germany gets alternate resources, but they come at a much higher cost. They're much more difficult. You're really putting stress on the German economy and causing inflation and other things.

But if you blockade, a continental power can't blockade you back. Why? They're on the narrow seas, so they can't deploy a surface fleet. You'll sink it. Also they can't easily blockade a coastline that faces the open seas, the high seas. You can do other things on narrow seas, but it's pretty tough. What do you do if you cannot blockade ships in port?

Well, then what you're going to try to do is commerce raiding to try and sink things when they're out and about. That was what Germany did in World War II and why its occupation of France was so important. Once it took the French coastline, it then set up U-boat pens in Brest, Lorient, Saint-Nazaire, La Rochelle, and Bordeaux. They're going to be using these to fight the Battle of the Atlantic. That's the game. Maritime powers blockade. The response of continental powers is commerce raiding. Then the maritime response to that is going to be convoying your merchant ships.

As the war begins in Europe, the United States is not in, but Roosevelt is coordinating with Latin America to set up a big neutrality zone all around the Americas, 300 nautical miles. Then the Lend-Lease Act, also before US belligerency, has pieces of it which say the United States is going to take over British bases in Iceland and set up bases in Greenland. Why do we care about those locations? They're really important if you're going to send convoys back and forth across the Atlantic both to attack and to defend. That's what we're up to.

Even so, the Axis visits a real nightmare in the Battle of the Atlantic. A lot of things are going down. How does that all work? Before the United States gets in, at the very beginning these U-boats turn the North Sea into a kill zone. Britain is losing an awful lot of stuff off of its shores and also off the shores of Africa. The Germans get really good at commerce raiding really fast. There's also the Fall of France which is a mess because prior to the Fall of France, Britain is only convoying just beyond Ireland. Once you get the Fall of France and all of those submarines on French territory, then the British have to convoy 400 miles further west. They've lost an awful lot of destroyers between the fall of Norway and then the Dunkirk evacuation. They haven't got enough ships to convoy properly.

Then the Germans are really creative. Admiral Dönitz, who runs the submarine service at this point, uses wolfpack tactics where you concentrate a whole bunch of submarines on a convoy, attack it at night, and bad things happen to the convoy. Also, the Germans have captured some of the British codes, so they have a sense of where the convoys are. They're sinking almost a terminal tonnage of this traffic. 850,000 tons of allied shipping is going down. This is Hitler's Happy Time when he's sinking an awful lot of stuff. Then there's a big Greenland gap. This is where there's a lack of air cover, and you'll see a lot of things are going down in this Greenland gap.

Meanwhile, the British have gotten pretty good at espionage and they've captured a lot of Enigma machines. Those are what the Germans are using to encrypt their messages. Well, the British capture some machines, some rotors, some code books in 1940 and 1941. By the summer of 1941 through February 1942, they can actually read the codes and decrypt them so that within 36 hours they can get the information out. This allows convoys to go, "Oh, wolfpack there. We're going to do evasive routing of the convoys somewhere else." That may have saved up to 2 million tons of Allied shipping.

Meanwhile, for the Germans, General Rommel is in North Africa. He's having troubles because he's supplied across the Mediterranean and the British and friends are sinking too many of his supplies. Admiral Dönitz is told to reroute some of the U-boats in the Atlantic to go help General Rommel up in North Africa. The United States isn't in the war, so all quiet on the eastern seaboard. It's looking like it might be okay for the British for a while, except Dönitz thinks something's up. They add a fourth rotor to the Enigma machine. The British are then blind again for most of 1942 until they can capture a four-rotor Enigma machine, all the rotors plus the code books. It takes a while.

They're in a world of hurt. The United States enters the war, which you think would be good for Britain, except it produces Hitler's Second Happy Time. Why? Because Admiral King, like his Royal Navy predecessors in the previous war, doesn't think convoying is the manly thing for naval officers to be up to. He's not for convoying. Also, Americans don't turn off the lights. Therefore as merchant ships are going up the east coast of the United States, the lighting is just highlighting their silhouettes, making them much easier to sink. By the way, in those days, Louisiana and Texas oil, which is supplying the east coast where a lot of American industry is, is coming up by ships on the eastern seaboard, particularly by the Cape Hatteras shoals which are like 30 miles wide. They become a total kill zone.

Admiral King rethinks it after losing more than a million tons of tonnage in the first three months of 1942 and goes, "Oh gee whiz, maybe we should do convoys." The United States does an interlocking convoy system by May of 1942. But then Dönitz just starts hunting things a little further south in the Caribbean.

The Brits get their four-rotor Enigma machine and they're able to decode things again. But there's another problem. The British think that there's something up with their Admiralty codes in August of 1942, but they don't change them out until June of 1943. There was something wrong. The Germans were reading them. You can see this back and forth in the Battle of the Atlantic. But eventually the air cover gap is closed. This makes a tremendous difference.

There are new technologies that are introduced that ruin Admiral Dönitz's operation. The United States had radar. Germans never did. American radar improved, so you can see through the fog. The United States adds hedgehogs. What are they? Not the cute little critters. Rather, if you have a ship and you have hedgehogs, they deliver an elliptical spray of depth charges. Anybody who's anywhere underneath you is in a world of hurt.

In addition, the United States introduces two new classes of ships. One, auxiliary aircraft carriers, little ones. That means you're going to have air cover for the entire journey. When you get beyond land-based air, then these folks will take over. In addition, small destroyer escorts were introduced instead of the big ones. These little ships have all sorts of fun things on board: sonar, radar, depth charges, hedgehogs. They transform commerce raiding into a low-life expectancy profession so that in May of 1943 the Germans lose 41 U-boats. That's unsustainable. That's a massive percentage of what they have. In one of those encounters—I think it's about 25 U-boats going after a convoy of 37 ships—they sink nothing, lose three U-boats plus another one gets damaged. On one of these U-boats is Admiral Dönitz's 19-year-old son Peter, who dies in all of this.

Dönitz, as a result, redeploys the U-boats out of the North Atlantic because it's unsustainable for Germany, south of the Azores. Yet there are problems going on with the Arctic convoy that takes a quarter of the Lend-Lease aid to Stalin, and it gets called off for much of 1942 and much of 1943. I'll get into it, but that problem gets solved and then you can see where all the kills are. The Germans are losing U-boats closer and closer to home shores.

The Battle of the Atlantic is won by the Allies in part by reducing merchantman losses through evasive convoys, and also increasing U-boat losses through all these different technologies and reading their mail, which helps to find them. But one could argue even more important was the civilian side of it: the United States' ability to just overwhelm Germany with the construction of US naval and merchantmen.

Here are the stats. Look what happens with naval strength. In 1943, US Navy hulls and personnel are tripling . That's quite a lot. Then naval hulls are going to double in the next year. That's a lot of ships and a lot of people in the navy.

Here are some more fun statistics. If you look at 1941 and go over, that’s a bumper year for U-boat construction. Then going into 1942. Really ugly if you’re a merchantman crewmember, because it's double the number of tonnage of merchant ships that are being sunk. But keep moving over. Look at Merchant hull construction, up by four times, and next year it’s going to double. Now let's look over 1943. Look at how many U-boats are being sunk. It goes way up with all those new technologies that I just told you about. Even though the Germans build a lot more U-boats, the kill rate is so high that there’s hardly any net gain.

Here’s more fun ways of looking at it. As soon as the United States gets in the war, new ship rates go up, up and away. But the losses are really high through mid-1943, but then it's in mid-1943 that you can see the big divergence between construction and what’s being destroyed. The Germans just can’t keep up with this. There’s just way too much stuff out there for them to sink.

Admiral Dönitz did get one thing right. His boss, Admiral Raeder, who was the head of the Navy, had said that pocket battleships were the way to do commerce raiding. Dönitz proved him wrong. U-boats were the way to go. Hitler agreed. He cashiered Raeder, made Dönitz the head of the Navy, and then Hitler went one step further. He scrapped his surface fleet, because it was useless to him in this war. You cannot deploy it in this kind of high-stakes war. It’s something countries like China, surrounded by narrow seas, ought to think about.

If you look at why the Battle of the Atlantic turned out the way it did, it’s that Germany and Britain had very different geographies. Arguably, Germany bought the wrong navy before the war. It should have bought a lot of U-boats and minimized the surface boats they were buying. Britain could do things Germany just couldn’t. That’s just part of the geography.

The effects of the blockade were really significant. You're really straining the German economy. But the German commerce raiding was also very effective. The Germans almost sank a terminal quantity of British trade. It's close because Britain is dependent not only on oil imports and resources, but about half of its food supply. The Germans tried and came close. Then you go, well, the counter-commerce raiding strategy did work, but it was touch and go, back and forth, and it required a lot of things that had to be coordinated. You needed the intelligence. It really helps reading other people's mail. You needed a whole set of new ship classes. You need to be able to construct adequate quantities of naval hulls and merchant hulls. You need to coordinate with allies. You got to get food and other things to Britain. There are a lot of things going on here. You need air cover and the planes that are capable of doing it. There are a lot of things going on. Remove any one of them and the outcome may have been different.

Sarah Paine 00:30:10

So that's it on the commerce raiding. Once you command the Atlantic—which means that a sustainable amount of traffic is going to get through from the Allied powers—then you can start thinking about peripheral theaters. What's a peripheral theater? The main theater is Russia in this war, not what you think. Why? Because between two-thirds and three-quarters of German ground forces were always fighting Russia. That means these other theaters, which are peripheral to the main theater, take up between one-quarter to one-third of German forces. That makes them peripheral. It doesn't make them unimportant, but they're not the main theater. By the way, who would want to fight in the main theater if there are alternatives? People die in droves in the main theater.

This man, Sir Julian Corbett, a very fine naval theorist of Britain, was heartbroken in World War I that Britain ignored his naval strategy, which was to not do the continental commitment and run the war through these peripheral operations. It wrecked his health. But according to Corbett, here are the prerequisites for a theater that makes a really good one for a peripheral operation.

One, it has to be overseas so the enemy can't invade you or wreck your productive base. Secondly, you need local sea control to get in there. But that local sea access has got to be better than the land access, because you want to have it easier for Britain to get in and out than it is for the enemy. Because then attrition rates will favor Britain. Britain also should deploy a disposal force. What's that? It means those forces you have who are in excess of what's necessary for homeland defense. It's not a disposable force. It's rather a force sized so if everything goes south, you won't ruin the homeland. You can afford to risk it on potentially very risky, but potentially war-changing operations.

In addition, they have joint operations. Joint means cooperation of land and sea in this era. You're going to come in by sea, you're going to supply by sea, but you'll be fighting on land. You better coordinate all that. It's got to be combined operations. What does that mean? Coordinating with allies. You need friendly locals who are going to help you do all of this. Then you need to command your own forces. Why? Because you want to determine how they come in, how they go out. If it all goes badly, you want to be able to leave, which is what they did at Dunkirk. As much as the continental power might want to play this game, they cannot because they don't have the requisite sea access to pull it off.

The war winds up with Britain and the United States going through four phases—well, for Britain it’s all four and with us it’s some of the later ones—of four peripheral theaters. First, the North Sea. That's essential for British homeland defense. There are three keys to the North Sea: Scapa Flow, Strait of Dover, coast of Norway. Britain always controlled the Strait of Dover and Scapa Flow. Germany went after Norway to try and open things up that way.

Norway is not an ideal peripheral theater for Britain because the Germans had better land and sea access to that theater. They take Norway and then they set up sub bases at Bergen and Trondheim in addition to the ones they had at Kiel, Hamburg and the Heligoland Archipelago. These are the things that are ruining those Arctic convoys. That one does not go well for Britain. It winds up facing a totally hostile continental Europe shoreline. The Axis has it all.

In the Mediterranean, there are certain keys that Britain controlled. Gibraltar that’s access to the Atlantic. The Suez Canal that’s access to the Red Sea. Crete that’s access to the Black Sea. The British had a fallback position, a midway point at Malta if things went really badly. The Germans attack the fallback position because they're trying to bail out the Italians fighting in North Africa. They're trying to take Malta because whoever's got planes there can cause convoys all kinds of problems, or they can protect their own things. Malta is blockaded by the Germans. It's not relieved until the end of 1942. That is not great for Britain. It's threatening its access to the empire.

The problem for Britain is Italian belligerency. Italians had been part of the Entente in World War I. Well, they're part of the Axis in World War II. Three of these keys—Malta, Crete and the Suez Canal—lie in the center of Italian ambitions because they want an empire where they're going to go down the Balkan coast, and then they're going to go deep into Libya and Ethiopia to unify their empire around the Red Sea. Mussolini kicks this thing off, except then he gets stuck in Greece.

The Germans have to come and bail him out. They chase the British out of Greece. The Britons are then on Crete, and then that falls to the Germans. The Britons are in a world of hurt. The Suez Canal is under threat. Think about it. You have the fall of Norway, fall of France, the blockade of Malta, the fall of Crete. It is really bad news.

However, when the United States gets into the war, the United States can help Britain retake some of these keys to the Mediterranean. This is what is going on in North Africa. Admiral Dönitz had saved it initially. But after the United States is in with more assets, 60% of the tankers supplying Dönitz's oil are getting sunk because Malta holds and Malta is supplied. These tankers are getting bombed out of existence.

Rommel loses in North Africa, not because he's the inferior general—he's a better general—he's just not supplied. This is the key. Once you take North Africa, that peripheral operation, it opens the opportunity to go into Sicily and then the rest of Italy. Once you get that going, there's a possibility to do the Normandy landing.

So there are four different peripheral operations. Some would argue that the air campaign over Germany was another peripheral campaign. You're bringing the war home to Germans, you're wrecking their productive base. Also it's a major help to the Russians. Why? Because once the British start bombing Berlin, Hitler calls off air squadrons from the Eastern front. These anti-aircraft guns can be used against tanks or aircraft. He pulls those back and that means the Germans no longer own the skies of Russia and you're no longer going to be taking hundreds of thousands of POWs.

Sarah Paine 00:37:13

That's how the peripheral strategies worked out, but they're coordinated through allies figuring out how to make all of them work. Here are the stats on allies. Alliances are additive, right? You should ideally add up all the complementary capabilities of you and your friends and then share them, divvy them out in an optimal way to deal with things. Well, the Axis alliance was also additive in the sense that the Germans added to their GDP the GDPs of all the places they occupied. That pretty much accounted for GDP growth in Germany. Except, that whole way of doing things involves big occupation forces which are big overhead for these conquered places. You've also damaged them in the conquest. Whereas the Allies are all game to help each other.

Also when Germany took over continental Europe, it's taking the big petroleum deficit zone because Europe doesn't produce petroleum. The North Sea oil, those things hadn't been discovered in those days. Yes, Romania has oil. Yes, Hitler takes Romania. But Romanian pipelines ran to prewar customers. It's very difficult to find the manpower and the steel to reroute all of your pipelines in wartime. That's a whole other part of Hitler's problems.

As you're putting together these complementary capabilities, the Russians have a huge army. Most of the fighting takes place between Russia and Germany. The Russians wreck the German army. You know, millions of Russians and Germans are dying here. But if you look at Operation Barbarossa, which is Germany's initial invasion of the Soviet Union, they suffer a nearly 30% casualty rate. That's called catastrophic success. You have too many successes like that, it'll be catastrophic.

Here's the mathematics of the main front. Only two countries have really big armies, Germany and Russia. Once the Germans maximize their territorial conquest, they've reduced the Russian population—because they've conquered all these areas—to less than that of the United States. Nevertheless, the Russians mobilize twice the army that the United States does. If you look at the mathematics of munitions, the Russians on their lonesome produce more munitions than did Germany. Once the United States gets in the war, we produce, I think like $100 billion in munitions. In that period the Germans produced less than $40 billion. Bad news for Germany in all of these numbers.

The trick is getting the munitions to the men. In World War I that wasn't feasible, but now these railway lines have been completed and it's pouring in. The Lend-Lease aid, you look at it and wow. Britain's getting a lot, Russia quite a bit, but less. China, nothing. What's going on there? You can only get aid in if there are ports and railways. Japan blockaded China. Proof of concept, it can be done. The Japanese did it and you could not get stuff in.

What the United States was trying to do... The British thought we were insane. They're probably correct. We were flying things over the Himalayas called the Hump. Great, you're going to fly in aviation fuel to land? That's going to be the same aviation fuel that's going to get you back. However many other bombers that you can deal with in China, it's not workable.

Russia, even though it got a lot less than Britain because it's not getting ships, is getting equally valuable useful things. For instance, Russia produces a lot of planes, but it doesn't produce adequate high octane aviation fuel. The United States had loads of that. The United States produced all kinds of vehicles, all kinds of rolling stock locomotives. This is what Russia used to transport everything.

The United States prevented Russia from going into a famine in the winter of 1942-1943. We fed them with this Spam in a can. This is Hormel Foods' contribution to the war. Everyone had much Spam during the war, they never wanted to see it again. But a little can of pork goes a long way and it doesn't spoil in a can. A quarter of this Lend-Lease aid goes through Murmansk. A lot of it gets sunk up there and not reliable. Quarter of it goes through Persia and then half of it goes over the Trans-Siberian. You go, “Well, what's this Axis alliance about? Why aren't the Japanese sinking any of this?” Talk about a dysfunctional alliance.

This is to tell you what happened with those Arctic convoys. You can see in 1942, a tremendous number of them were sunk. These are fully-laden ships with scarce war materiel. We and the British called them off for most of 1942 and most of 1943. Stalin was beside himself. In 1943, he's sending out feelers to Hitler trying to make a separate peace. Luckily Hitler was not interested. Hitler wasn't wasting his time with targeting empty ships on the home voyage. He was only going after the laden ones.

What is all this Lend-Lease aid support? What does it add up to? Well, in Barbarossa you got 3+ million Russians going after 3 million Germans. That's a lot of people to supply and tie up. As you watch that go, the Axis is killing Russians or making them POWs by the millions, but it's suffering 15% losses of its forces. It's going to add up. When you get to August 1941, that's when the siege of Leningrad, St. Petersburg begins. By December, the Axis is within 25 miles of Moscow. It’s not looking good at all. But once the United States is in the war, you can start doing peripheral theaters like in North Africa. Then these casualties, while they aren't as significant as what's going on on the main front, it's cumulative.

Stalingrad, the largest battle of the war, is going on. You got North Africa. Then when you have Kursk, which is the largest set piece battle of the war, a big tank battle. This is when Sicily's happening and moving up the Italian peninsula. You can see the cumulative effects of these sequential operations and how they’re adding up, both the main theater and the peripheral theater.

When you get to Normandy, the Russians are tying up 228 Axis divisions. There are only 58 Axis divisions in Western Europe, Italy and everywhere. That's what makes Normandy possible, the Russians really holding onto things. When you start looking at the main front, peripheral operations in Italy and then Normandy, and then you have additional fronts in France and then the Balkans, this continental cancer goes into remission. The Germans just can't sustain it.

To summarize, before Britain had allies in this war, it was in a world of hurt. It was just losing one thing after another. But once you get the big continental buddy—don't dismiss the importance of a big continental buddy—that is in the area where the fighting is taking place, not separated by the seas, but there with a big army, that was essential. Then once the United States gets in with its big productive base, then you can really start doing things. Because if you can command the seas—that's what the Battle of the Atlantic is about—then you can connect the world and connect allies, theaters, resources.

Here are the operational effects of these peripheral operations. You start with one where you can. If you win there, it'll open up a menu of more promising locations. All the while, you're attriting your enemy's forces. Also, if you're doing it right, you're relieving pressure on the main front for Russia, which is doing the heavy lifting. The strategic effects—if you can do this successfully—are that you're going to control resources for yourself, deny them from people you don't like. This will help put time on your side.

You're strengthening your alliance system because you're essential to each other's survival as you coordinate things and you're dividing your enemy's attention among multiple theaters, overextending them. You start by trying to contain the problem, and as things go on, you try to roll it back, and then you go for regime change. You're producing cumulative effects from these sequential operations. That's how it works.

Churchill, the great wit, talked about the hassles of dealing with allies. These are high stakes discussions. They don't always go pleasantly. But his idea is that there's only one thing worse than fighting with allies, and that's fighting without them, because you’ll be toast. You need these complementary capabilities, different locations, and coordinating it to gang up on your continental problem. Also, if you stick with it, you can help establish precedents, laws, institutions that'll hold the peace after the war. That's his take.

Sarah Paine 00:46:40

I've given you a big exposition on maritime solutions to continental problems. In a global war you go for blockade, and then counter their commerce raiding, peripheral operations, massive production, and then joint and combined operations.

Sarah Paine 00:48:04

So that was then. How about now?

This is what NATO looks like. You can look at the United States with our western and eastern coasts that are unencumbered. You wouldn't be able to blockade those. The two narrow seas are the Sea of Labrador and the Caribbean, but it’s basically hard to imagine that the United States Navy wouldn't be able to deploy in wartime. The same thing is true on the big peninsula of Europe. Turkey may not be able to deploy, but Spain, France, and Norway, there are these unencumbered coastlines. They'll probably get their navies out.

Eurasia is a completely different story. This is where China and Russia live. No one's done this to them. This is just the way it is. It's all of these narrow seas that become kill zones in wartime. I've described it to you how it works. I get it there's the Arctic up there, but there's no economic activity or population. Being able to run things in the Arctic doesn't do you very much good. Nature naturally contains both Russia and China. It's just the way it is.

Let's start with Putin here, Vlad the Bad on his little Mongol pony, a continental mode of transport. If you live there, what can you get out of it? It's a long horse ride if you're going to do it that way. Whom can he blockade? Well, he's got one aircraft carrier, the Admiral Kuznetsov, which has been under repair since 2018. Long time to be in the shop. It required the largest dry dock that Russia has to repair it, and then it flipped the dry dock, which then gashed the hull. There's been all sorts of charges of embezzlement, various fires on board. It's not getting repaired anytime fast. It's just like an occupational hazard problem. That one's not doing him any good.

Putin has got two liquid playgrounds that he likes. One's the Black Sea. Before he got involved with Ukraine, in theory he could have blockaded Bulgaria, Romania, Ukraine and Georgia if he wanted to. That was why he's interested in the Kerch Strait, which would blockade the Sea of Azov, because he wanted to take Ukraine from east to west. That was part of the plan. He used to have, maybe still does, it’s unknown, one overseas base, Tartus in Syria. Talk about a garden spot. I guess if you want to go bomb civilians, that would be a location for you. But now that Bashar al-Assad has moved to Moscow, it's unclear how that works. In any case, it's a useless base because in wartime, no one gets through the Dardanelles. That place can be shut down with mines. The Ukrainians have shown that you don't even need a navy to stop a navy in narrow seas. It should be a real wake-up call to anyone on a narrow sea. Just by using drones and shore ordnance and planes and things that you can wreck navies. In fact, the Russian naval base used to be at Sevastopol in Crimea. Putin's had to move it to Novorossiysk. Great, he can't do too much there. Have fun with that one.

There are many fewer possibilities for him nowadays. His second favorite liquid playground is the Baltic. Kaliningrad is sovereign Russian territory. Back in the day, I guess if they wanted to blockade the Baltic states, they could try it. But after the latest iteration of the Ukraine war, Sweden and Finland have ditched neutrality. They're part of NATO. Baltic states are also NATO. The Baltic is really a NATO lake. There's not too much Putin can do there. His latest gig is cutting undersea cables. That's about where he's at. I mean just look at it. Trying to leave the Baltic, it doesn't happen in wartime. You're stuck in there.

It turns out that Putin and Russia are much more vulnerable to blockade than the other way around. It's very easy to close up the Baltic and Black Seas. You look out on what's going on and they're hemmed in. It's all way up north. Russia has two really big naval bases, one on the Bering Sea and one in the Barents Sea. But the problem is when they deploy out of those bases, they have to go by NATO territory. You can see why Greenland and Iceland are strategic, because for Russia to deploy, it's got to go in between those places. Then on the Bering Sea, the more promising one for Russia, it has to go by the United States. The problem is how do you supply the thing on the Bering Sea? It's just a long way from any industrial base.

Another part of Putin's sad story has to do with NATO. If you look at NATO, you can look at its accession in arcs. Initially in the Cold War, the early Cold War, it's all these smaller European nations and everybody's smaller than Russia. It's by far the biggest country on the planet. Why they need more territory remains a mystery. Initially it's the smaller places joining NATO to protect themselves. Then at the implosion of the Soviet Union, it's all its former satellites fleeing at the first possible moment and saying, "NATO, NATO, let me in." Then with Ukraine in 2022, part of the Ukraine war, Sweden and Finland, which long had preferred neutrality, said, "Whoa, no, no, we're going to join NATO." Now the Russians look at this and they go, "Well, this is NATO coming at us in arcs." They're ignoring their complicity in all of this. If you occupy places and brutalize them for generations, this is what you get.

Here is Putin's muse. I don't know if it's his muse, but it's a guy who expresses a lot of ideas like Putin's: Aleksander Dugin. Here's his view of how the world should be. It's not a universal rules-based order where we all trade with each other. Rather we're going to divide it up into these spheres of influence where each is a world unto itself. Dugin's worked it out for all of us.

Here are the places that Russia's actually taken lately. At the end of the Cold War, it took Transnistria from Moldova. In 2008, it took Abkhazia and South Ossetia from Georgia. Then 2014, it's taking Crimea, Luhansk and Donetsk from Ukraine. Look how it really works. This is a continental view. You want to color it all in. That's what the 2022 invasion of Ukraine is. It's the way a continental power looks at territory.

The United States thinks that Putin is just worried about us and NATO. Actually, Putin's problem is China. NATO doesn't want Russian territory. Who would want it? It's full of Russians. Stay at home, please. Just don't leave. But China might want it.

Also, Putin has said that if you really feel deeply about a place, forget about boundary treaties, you should be able to take it. Well, Tsarist Russia took a lot of places from China that the Chinese might feel deeply about, that have precisely the resources that China needs. They have not only energy and all those sorts of things out in Siberia, but water. Lake Baikal holds 20% of the world's surface fresh water. China's been blowing through its water table in North China. That one, I think the topography works such that you could pipe it all in.

Putin is dumping his ordnance on Ukraine that never wanted to invade, while he's letting this problem metastasize. Speaking of the problem, here he is. What are Xi Jinping's possibilities? You like the little natty upgraded Mao suit? Here's his world, welcome to it. What he's got are 20 neighbors, 13 by land, seven by sea, many of which despise China for excellent reasons. Not all of them, but some have them. If you look at it, it's all of these narrow seas. If Xi Jinping wanted to blockade somebody I guess you could try Korea, because you got the Yellow Sea and then the Sea of Japan, except Japan sits out there. That's a complicating factor.

Better bets are in the South China Sea, because Vietnam, Brunei and Cambodia don't have alternate coastlines. If they want to reach the open oceans, they've got to transit the South China Sea. However, Thailand, Malaysia, Indonesia, the Philippines, Taiwan, all have other coastlines that face the high seas.

Remember, I've just given you a whole discussion about the Battle of the Atlantic. When people have high seas access, it's really difficult to shut them down because China would have to go all the way around to be on the far side to shut everything down on the other side. Maybe they're just going to obliterate it, I don't know. Anyway, the South China Sea is such a legal mess. No one knows who owns what. In any kind of war, you're going to have all these neutrals who might suddenly join sides if they don't like what's happening. But you can get a sense of what the possibilities are there. It's very difficult.

Usually in a war, we've already done narrow seas, right? The North Sea, the Baltic, the Black Sea, the Mediterranean, they shut down in wartime to commercial traffic. It simply can't make it through. Surface fleets have great difficulty. If the neighbors, as Ukraine has shown, just buy the right drones—and since the sea's a little bigger—buy some submarines, some planes, shore artillery, that might be enough to just close down China's merchant marine and also its navy.

If you want to go the other way of working, well, what about Xi Jinping doing a surprise visit on the west coast? There's a lot of open ocean. Hawaii would be the nearest big island and it's a long way away. Hawaii's dependent on imports for everything. It's not very useful. But for the United States going the other way, the islands get bigger and better the closer to China that you get.

I leave you with Alfred Thayer Mahan, captain and later admiral, by far the most famous person ever associated with the U.S. Naval War College where I spent my career. Here are his prerequisites for playing the maritime game. One, you need a moat. You've got to have insulation from attack if you want to play this game. You need a dense internal transportation grid to get the goods out in peacetime, reliable egress by sea to get the navy out in wartime, a dense coastal population that's going to be running all the trade, a commerce-driven economy. Then you need a government that's stable that is going to support funding a navy and supporting commerce.

Let's line up Russia and China with these prerequisites. Well, neither one's got a moat. They’ve got more neighbors than any other two countries on the planet. Definitely not that. Internal transportation grid, Russia's remains lamentable. China's is getting better. Neither one has reliable egress by sea with these narrow seas. Sure, Russia's got the Arctic, but there's nothing up there. There are polar bears. Great, go get bitten. As for dense coastal population, yes, China has a dense coastal population, but Russia doesn't way up north. Russia's never had a commerce-driven economy. China was more under Deng Xiaoping, but under Xi Jinping, he is privileging the crony sector over the private sector. Neither one has stable government institutions. The litmus test of that one is whether you have transparent, regular transfers of power, usually through elections. A dictator for life does not remotely qualify.

So sure, China and Russia remain continental problems, but I don't think they understand the maritime limitations of where they're at. Of course, peace would be the better thing. Keep compounding growth and war, as the Ukrainian war shows daily, what a waste of Russian assets. But sadly, the enemy gets a vote. That's what I had to say for you this evening. Thank you so much for coming and being such a wonderful audience. Thank you.

Dwarkesh Patel 01:00:28

First question. The lecture is framed as explaining Britain's strategic wins in how they prosecuted World War II. But it seems like what seems to keep happening is that Germany and Hitler specifically keep making mistakes. In 1940 and 1941, Halifax was saying, "Look, we need to come to a peace with Germany. We've lost the Battle of France. We've had to evacuate Dunkirk." Then Hitler could have done a blockade of Britain and could have just kept prosecuting the war. Instead, he decides to open up a second front against the other biggest army in the entire world, against the Soviet Union.

When Japan does Pearl Harbor, again an unforced error, he makes another error on top of that by declaring war on America, which he didn't have to do. It made it much easier for America to lend support in the war against Germany. This seems much more about Hitler continually bungling than about Britain getting the strategic picture right.

Sarah Paine 01:01:38

It's both. I don't know that Germany ever blockaded Britain. It was more that he was planning to invade Britain if he could. He lost so many naval assets doing the Norway campaign, which he really needed to do, that he doesn't have the naval assets that he had at the beginning of the war. That's unlike World War I, where the Germans basically had their whole fleet at the end of the war, and then it got scuttled at the end of the war. This time, Germany did the smart thing, which is to use that fleet to get something. They get Norway, but then they don't have a fleet. Invading is difficult.

Then because he's in the big petroleum deficit zone, he's gotta get moving. I'm not clear that he had tons of choices, but he's gotta go for resources. The oceans have been shut off for him. The whole proposition of what he's up to is questionable. But it doesn't mean that it isn't a really bitter struggle and difficult. Even though you go, "Well, Germany's GNP, if you add it up compared to the totality of the Allies vs. Axis, the Allies have much more stuff…" You can go, "I think the Allies are gonna win this thing." But after how many millions of losses? I showed you the Battle of the Atlantic, it goes this way, it goes that way, it goes this way, it goes that way. A lot of people are dying throughout all this.

Yeah, the Germans made errors. The first one is, why even go to war? If you want to dominate Europe and you have a really great growth rate, which in World War I they did, they would have taken over Europe years ago by just growing. Don't do either world war. Just keep growing your economy and then you're going to have all these good business relations. The error is even going to war for these things.

But it doesn't mean these wars couldn't have gone the other way. You're right, you look at blunders. But you call it a blunder now because it didn't work out. At the time, you don't know. A lot of people get away with really risky things and then you say they're brilliant. The ones who don't get away with the risky things, you say they're idiots. But it's dicey.

Dwarkesh Patel 01:03:59

I want to go back to the question of, when we're trying to explain who wins a big war like this, how much should we be looking at the specific strategies used in different battles versus just the total tonnage of industrial output that America especially contributed? By 1942, what is the way in which, even if they had a much better strategy, Germany would have won?

Sarah Paine 01:04:20

Define win. What is "win"? If Germany's win means controlling the entire continent plus Britain, plus Russia, you come up with one answer. If Hitler had just done the Anschluss, Austria joins, and maybe done the Sudeten number, which is a large German population in Czechoslovakia, and quit, he'd be called Bismarck II, a genius. No losses and he gets to keep that bigger area. But that's not who he is. You need to define what "win" is for both the Germans and for the Allies to determine what's feasible and what's not feasible.

Germany can be invaded overland potentially everywhere, whereas Britain can't. Germany just cannot deploy a navy the way Britain can. It's just geography. Therefore, Germany should have bought a completely different navy. Skip the surface fleet, buy a lot more U-boats. Maybe they would have zapped the British before the United States got its act together in either world war. Do it fast. They didn't do it.

Dwarkesh Patel 01:05:28

I found your thesis that Britain fought this very effective maritime strategy really interesting, but let me just try out a different theory for you.

Sarah Paine 01:05:36

Counter argument. Here we go. Yes.

Dwarkesh Patel 01:05:38

After the fall of France in June 1940, Germany has all these U-boats. They're actually themselves running a very effective maritime strategy. In the Atlantic, they're not in the narrow seas. They can actually go to the ocean and cut off British trade through the Atlantic. In October of 1940, they sink close to 400,000 to 500,000 tons of British shipping.

The Germans are running this very effective blockade of Britain. The reason it fails is just that the Americans eventually can produce many more ships than the Germans can even sink. Ultimately it wasn't a matter of who ran the better maritime strategy. It was who could keep up the industrial output to sustain this strategy. It comes back to industry and not strategy, even for these maritime operations.

Sarah Paine 01:06:34

A, they're not blockading Britain. Britain's getting all its stuff out, but then they get sunk somewhere else. Blockade is when you're actually keeping things in. Germany has exactly zero merchant marine going anywhere. The stuff that you're talking about being sunk, they're naval ships but a lot of it is merchant marine because the British are getting things out. Tremendous amounts of it are getting out. That's just a difference.

You're correct about the importance of productive bases. Britain's problem in World War II is different from the Napoleonic Wars. When you change to oil for what's powering your ships and things, Britain doesn't have that at home. Whereas coal, it does. It has some A-list coal. When you're doing age of sail and coal, the British are okay. But then when you're changing energy sources, they're definitely not okay. There are other factors that are going on with them. But you're right about the productive base.

There's another game to play which I think is worth playing. It's the game of takeaway. You're telling me that there's this factor which is economic size. It's decisive and this will determine the outcome of wars. You're absolutely right that it’s important. Air power enthusiasts will say, "No, no, no, that's nonsense. It's all about controlling the skies." That when you control the air and you get air cover for your convoys, that's what does it.

Let's talk about why the Battle of the Atlantic turns out the way it is. If I took away cryptography, would it have turned out the same way? Negative. If I take away not just the sheer amount of technology, but certain key pieces like radar and those things, what happens? Does it change? Yes, it does. What happens if the United States just doesn't like alliances and you don't coordinate things particularly well? Does it change things? Yes, it does.

You can go through a list of this. Different people are designing different types of ships. You are also determining that you're going to share things with Britain. Britain's providing a lot of free stuff. If you play remove any one of these things, that Battle of the Atlantic turns out differently. The story is, it's probably a package of many things. This whole game of takeaway, when you have people who will come up with a monocausal explanation for you and they'll go with one cause, they'll probably be right that their one cause is really important. But then they are just ignoring all these other things. That's the more important thing in strategy. Your thing is truly important.

Dwarkesh Patel 01:09:24

The question then is what does "changing things" mean? Does it mean that the war could have gone on longer, or shorter? Yes, anything would have changed that. There's another question of would it have changed who won the war.

I'd claim that cryptography, radar, even oil… Even if Germany had a huge reserve of oil, maybe it would have lengthened the length at which Germany could have sustained itself. But if you change the fact that the Axis had one-fourth of the GDP of the Allies combined, I think switching that genuinely changes who wins the war, whereas these other things actually wouldn't change who won the war.

Sarah Paine 01:10:02

It depends whether the war is a war for limited or unlimited objectives. You can play that game with the Russo-Japanese War. All the data you talk about, Japan’s is a fraction of Russia’s but Russia gets trounced in the Russo-Japanese War. There's something that's going on. Japan has a very limited objective and Russia has other problems at home and is willing to bail. That's one thing.

But there's another thing. If Germany had simply never bought a surface fleet, if they'd been sensible… Surface fleets are great, Britain's got one, but it can always deploy that thing. Germany will never deploy its surface fleet. It'll get it sunk, which is what happened in the Norwegian campaign. From the German point of view, they got Norway out of it.

If they had simply bought U-boats instead, I don't know whether it would have… You'd have to do the math and I haven't done that. What their goal would need to be is to knock Britain out before the United States ever gets into the war. Definitely do not declare war on the United States. If you can keep the commerce raiding going for another six months, maybe Britain falls. Maybe that's the pivotal error. It has nothing to do with what you're talking about. It's simply that Hitler shouldn't have declared war, I don't know.

There are many factors that go into these things. A reason why you should listen to people who disagree with you is that they may see some of these other factors. We all have blind spots and we rely on other people to find them. But you have to be receptive to listening.

Dwarkesh Patel 01:11:40

When Germany and the Soviet Union split up Poland, there's this 2000-kilometer border from the Baltics to the Black Sea that they've signed themselves up for, sharing it with the other largest army that has ever been assembled. Did the Soviet Union think that in the long run that this was a sustainable proposition? Or did they just think that we could wait Hitler out? Or in the long run they’d be in a better position to wage war against him? Did people think that this was a sustainable arrangement?

Sarah Paine 01:12:10

I don't think Hitler was about sustaining any arrangement. He was about taking over Lebensraum in Slavic lands, Russia. But flip it the other way, which is what the Soviet Union is thinking: “It'll be great when the Germans go after the British. What's not to like about that one?” It's the pounce and absorb. They will weaken each other that will open opportunities for Russia. That would be the plan. Then he's looking at it and I don't know the details of what his timeline is, whether he thinks it's just a delaying act. I have not gone into the archives to read what Stalin was up to in great detail, but I think it's basically that one. He’s thinking that the British and the Germans could go at it.

Also the Russians had tried to work out some deal with the British prior to all of this. With the British, the communists had been trying to destabilize them ever since the Bolshevik revolution, and they didn't get just how lethal the fascists were in Germany. They didn’t get that they weren't just run of the mill authoritarians, they had this whole genocide component that went with them.

The British are thinking the real threat is the communists in Russia. They're right, long term they are the real threat. But in the more medium term it's the Germans. It takes blitzkrieg for the British to go, uh oh. Then they're willing to team up with Russians. Russia's willing to team up with them. When German invasion transforms what had been primary enemies, he takes that role being primary enemies for both Britain and Russia. Then you're gonna glue another alliance.

That's the trick in our own day. You do not wanna do things that make the United States the primary enemy of those two, of Xi Jinping and Putin. Currently we're not. It's Taiwan or India or Vietnam. Whatever Xi Jinping's gonna steal currently, that's his primary problem. It's definitely Ukraine for Putin. You want to keep that all divided up to the extent that you're able to do that. A lot of things aren't feasible. These are high stakes things.

This is where just not thinking about your foreign policy… It takes a lot of experts and a lot of people to game out what is your best course of action. What is your least bad option? That's usually what foreign policy is. It's not what the great option is, there are none of those. It's what the least bad one is. It takes a lot of thought and gutting the State Department will not get you there. It's profoundly dangerous.

Dwarkesh Patel 01:15:03

You described Hitler's strategy as pursuing a continental strategy and that was his undoing. But it seems like there's many other wars where a more reasonable person can pursue a continental strategy and do just fine. Bismarck does multiple continental wars—Franco-Prussian War, Austro-Prussian War—and achieves his objectives. Hitler is just freaking crazy. He continuously does things. He knows going to war with Poland in 1939 that France and Britain will join. He's like, "I will fight both of them for Poland for some reason."

Sarah Paine 01:15:36

Resources, big resources.

Dwarkesh Patel 01:15:38

Somehow that works. Somehow, France falls in a few weeks. Instead of stopping there, he decides, “Now I'm going to go to war with the whole of the Soviet Union.” Then in what way is it a continental strategy to declare war on America, which is an ocean away? It's not even an adjacent continental power. It seems like just him being crazy explains more than him pursuing a continental strategy.

Sarah Paine 01:15:59

Let's talk about Bismarck. This is the difference between limited and unlimited objectives. These concepts, which I've tried to transmit in these lectures, are useful and you apply them to other things. An unlimited objective means the state in question is not going to exist at the end of the war. If it's really unlimited, and Hitler does the most unlimited variant, he's going to kill all the people. With Slavs, he's going to literally butcher them all. That's one kind of war. A limited war means you want maybe a hunk of territory, something less. The opposing government's going to live.

Bismarck runs three wars. One is the Danish war, then there's the Austro-Prussian war, and then there's a Franco-Prussian war. In the Danish war, it's about a couple of provinces in the far south of Denmark that Bismarck covets because he wants to unify for the first time all these Germanic states. He trounces Denmark in that war. Does he go for regime change in Copenhagen? No way. He just says, "Cede this place that you shouldn't have had anyway because it's got a bunch of Germans in it."

This is another concept, value of the object. How much is victory worth to you versus the other guy? For Denmark, just making Bismarck go away for this little state, the value of the object's lower to them. They aren't thinking in terms of creating a greater Germany, which Bismarck is. You have two things going on. You're giving the Danes, given how much the Prussians trounced them, a generous peace. You're saying, "Just get rid of this Germanic part that you don't really care about." He gets away with that one.

Then in the war against Austria, he just slams them. He's got this railway system that allows them to deploy where the Austrians can't. There are all these little Germanic states watching all this going on and they're starting to confederate sort of with Prussia because they're scared of all of this. Instead of doing regime change in Vienna, it's just saying, "Hey, let's cede a few more of these things to Germany." It was a big empire where you have pieces, you own pieces, you lose pieces. You lose pieces, you own pieces. For the Austrians, the value of the object again is much lower, given how they were slammed in the war. It seems to be a generous peace.

Then when he gets into the Franco-Prussian War, all these little Germanic states are joining with Prussia and that's going to create a unified Germany. Then he takes a bite too far when he takes Alsace-Lorraine, which the French are totally upset about. They never give up on that one. But all of these objectives are limited objectives. He's not overthrowing the Danish government, the Austrian government, or the French government.

The value for Prussians of unifying the Germanic states is far greater than the places that are losing these itty bitty places. Also, they don't get the big picture. They don't see it until it's done. By doing all of this, Prussia has transformed itself from the weakest of the five great powers of Europe to second only to Britain. That's ex post facto. It's a real reason why you don't want royalty running stuff. You want people who get their jobs based on merit rather than who their dad was. The world that Bismarck's dealing with is kings who are just clueless what they're up to.

Dwarkesh Patel 01:19:33

Last year, I interviewed Daniel Yergin, who wrote this big history of oil. He has like 300 pages about World War II in his book and specifically claims that Barbarossa was motivated by Hitler's desire to get the oil fields.

Sarah Paine 01:19:49

Yes. Yeah, he needed it because he had this big petroleum deficit zone. When you say “stupid to invade Russia”, probably it’s that once you've done this number, you have to go into Russia because you've got to take those oil fields. Then you can go, "Buddy, don't go for Moscow. Forget those people. Just go straight for the oil fields. Take that only." That would be more of the blunder.

Dwarkesh Patel 01:20:13

Is there any world in which Hitler would have surrendered by 1943 when it was clear that the war was going to go the other way?

Sarah Paine 01:20:22

Dictators don't. They just double down. This is the beauty of elections. I get it. It's a mess with a multiparty system and with all the crazy politics that come and go with it. But elections are a moment to reassess when the party in power sufficiently screws things up, where people finally go, “Ooh, this is a mess.” The party in power isn't reassessing. The election is reassessing for them.

Human beings, we've all met many. Most human beings don't like to change their mind. It's a terrible mistake. That's why I like the argument/counterargument structure, to try to take some of the sting out of changing our minds. Doubling down on bad decisions is a mistake, but we human beings do it. If you have a dictator, you're guaranteed. Putin is not going to back down in Ukraine. He's going to be going after it from now until doomsday. Then if he wins there, he's going to then go after the Baltic States. It's not going to end.

The enemy gets a vote, and you may go, "These are idiot decisions." But they're very dangerous ones. You have to do things to counter them.

Dwarkesh Patel 01:21:35

In World War I, you do have a lot of powers which have at least some amount of franchise. In that war, it would have been much better to lose the war in 1914 or 1915 than to continue waging it until 1917. But even Britain doesn't back down. France doesn't back down.

Sarah Paine 01:21:52

The problem with that war…This is another thing that is really helpful to do. To the best of your abilities, write down primary adversary, secondary adversary, tertiary adversary. Take all the powers in World War I on both sides and do their primary enemy, secondary, third. You find out it is a mess. Nothing aligns. So that, if you try to get out of that war, somebody's not going to like it because that's their primary objective. There's a lot of bad overlaps in who's going to get whatever parts of the Balkans at the end of this war. You really have to sit down and look at it. If I were doing a lecture on that, I have a whole other set of slides that would show you this. They're fighting parallel wars.

It's really not a good idea to have royalty running the show. World War I is the war where militaries on all sides are running the show and civilians are taking the backseat, saying, "I don't really understand military operations. We're going to let you boys do it." Disaster. That war sets communism and fascism up, it puts them on steroids. You can argue we've been dealing with that mismanagement in one way or another ever since. For most of my life, I thought we had put fascism back in its box and we were down to dealing with communism.

Now it seems that this very authoritarian vision which seems to rhyme with fascism… You can get into a big argument, what precisely is fascism? You try to define it. But it seems like we've come full circle in the collective West. It's a toxic way of dealing with all this.

World War I is an absolutely mismanaged war. Then the people who live through that set up these institutions that have held the peace until the present day, when irresponsible leaders across the globe are not cautious enough. Here we are. There's a price to pay if you have a bunch of reckless drivers on the road trying to play chicken with each other. How this will all end, I don't know the answer.

That's my reason for doing lectures on strategy. I'm going to give you concepts, some data to think about. You're going to have to form your own opinions and your own conclusions. But try to make it evidence-based and think about things. I recommend reading things more deeply. Don't just take what any one person has to say. Come to your own terms. That's the reason for doing all this.

We're living in really portentous times where people may make decisions where there's no going back. If we blow our alliance system, we're going to be dealing with China alone. That won't go well. Particularly if we think we're going to corner them in some way, corner a great power like China and think that'll go well for us. It won't.

Dwarkesh Patel 01:24:56

The reason I brought up earlier the primacy of industrial output and population size in determining who wins a war is that at the end, you sort of made a comparison that Russia and China today are in a similar position as the Axis was preceding World War II, at least in terms of the disadvantages that they had.

Sarah Paine 01:25:14

No, maybe I wasn't clear. I was just saying that we saw how it all worked in World War II, of people blockading or not blockading or doing peripheral operations or not. If you occupy Russia's position or China's position, it's not about who's going to win or who's going to lose. Both of them are much more vulnerable to blockade, period.

Dwarkesh Patel 01:25:35

Then it depends on do you think that was more central than the fact that during World War II, the Allies just had three to five times the combined output of the Axis?

Today, if you looked at the manufacturing output of China versus the United States, the shipbuilding capacity of China versus the United States, even if you looked at the international trade of China… China has more international trade with the world than the United States does. You can ask, “Who is in a more similar position to the United States during World War II?” Is it China or America?

Sarah Paine 01:26:10

More concepts. I did limited and unlimited objectives and I did “value of the object”, how much winning is worth to you. Another one is, what is “win”? If you're talking about the United States invading China, that’s insane. Who would ever do that? If you're thinking of whatever conflict you're imagining, if you’re thinking about if “win” is China…Tell me what this war is, who's going for what, and what a win is going to look like, and then we can talk about what feasible is.

Dwarkesh Patel 01:26:41

The most likely thing would be a war over Taiwan. The value of the object is clearly higher to China there.

Sarah Paine 01:26:46

If you look at Taiwan, maybe China can take it. It may have to leave it as glowing embers. That lovely chip foundry that everyone cares about, I imagine would be the first casualty of that war. That will be blown and be gone. There's no way you'll capture that intact. Everybody else is going to be terrified of this. It's not blockade nowadays. It's rather sanctions, saying, “If you're not going to play nice, we're not going to trade with you. You're going to have a massive timeout from the global order.” China is very trade-reliant.

Dwarkesh Patel 01:27:29

It's precisely because it has such a dominant position in trade that it would be difficult to coordinate a coalition against them.

Sarah Paine 01:27:34

Oh, why?

Dwarkesh Patel 01:27:35

The entire world benefits more from trading with them than the United States. There's more volume of trade with China than the United States globally. Then it would be difficult to coordinate a coalition to say, "Look, you’ve got to stop trading with the person you're trading most with."

Sarah Paine 01:27:46

The story is about sanctions. They don't have to be leak proof. If you can knock off one or two percentage points of Chinese growth just because some people are sanctioning like war materiel, that would be a very likely thing. After Taiwan, maybe you're trading with China, but you're not doing certain categories of war materiel with them.

The cumulative effects of doing that to a country over generations is the difference between North and South Korea. It adds right up. You could do the math better than I can. If you knock off 2% growth per annum, I think the doubling time of an economy… I don't know, if you knock a couple of percentage points off, it goes from 25 years to 75 years. You will be able to do the math better than I can rapidly. But it's very consequential and China's neighbors will be scared to death of this.

Also, you're right that trade patterns don't go overnight. They go gradually over like a 10 year period in which China will find it's already happening, as supply chains are moving out of China. It’s slow.

Dwarkesh Patel 01:28:55

The larger thing I'm trying to get at is this. In the immediate term, it seems like the thing that mattered in World War II was the fact that the United States was outproducing all the Axis combined in planes and munitions and tanks. Who is in a comparable position today? You'd say it's China.

Over the long run, perhaps we can cobble together a coalition which over many generations reduces China's growth or ability to compete. But I'm not even confident that the coalition would center around the United States rather than around China.

Sarah Paine 01:29:33

Not today. No one's centering around us right now because we're not playing nicely with allies right now. Rather, I think you will find that there are many countries that are eager to follow the Chinese model for development. They have cheaper labor now, places like Vietnam. India has also been doing all sorts of things. My impression with India is that one of India's problems is that it needs better infrastructure. There are investments going on in that.

There are many other people in the world who are interested in making money. They've watched how China did it. You can emulate that. If China is foolish enough to do Taiwan, these will be the long term ramifications. It's a mistake. As annoying as other countries are, it's better to just argue with them at international institutions. China has a big presence at these international institutions and should be able to get others to go along with however it wants to change rules. Do it that way. If they're going to do it the Putin way, you can look at it, it's negative sum and they will get a timeout from the global order.

Ever since the Industrial Revolution introduced compounded economic growth, this continental solution is a mistake. It is much better to forget about invading territory, allow the free traffic of people, trade, goods, ideas, and then we'll all grow together.

Dwarkesh Patel 01:31:02

On the general framework of how continental powers should act and how maritime powers should act. I wonder if it's worth even designating the optimal strategy based on your geography. To the extent that you think positive sum trade is better than invading and doing bloody wars, that's true regardless of whether we have a maritime or continental situation.

Sarah Paine 01:31:22

Yeah, I've just discussed geography because it's something I think most people don't think about. Particularly what's hilarious about people who buy navies—I've sadly had to read these naval theory books—they never look at the geography. They occasionally do, but it's sort of, "I'm going to get a big navy." Well, it depends where you live.

I've just finished and edited a book about fleets and being. If you have a fleet, what does that foreclose for whoever's around you? What does it open for you? It's not a great read but I learn a lot from doing it. It becomes really clear.

You can see that map there. You see Italy. You have in World War I, with that particular position, the Austro-Hungarian Empire has quite a good little submarine fleet. Captain von Trapp in The Sound of Music, he was a war hero, a submarine captain. Just by having that submarine fleet in there, it forced Britain to convoy the whole Mediterranean. Doesn't mean Austria-Hungary wins the war, but a very simple investment prevents anyone from invading over its oceans, which is important. It also makes the British do very expensive convoy duty because Captain von Trapp is quite good at sinking naval ships of the British.

That's the difference. Your geographic position limits what you can do. But now we have a whole rules-based order, that if you join the party you have a massive alliance system. That's why I look at China and it doesn't have any allies. It has itself and it likes to have bilateral relations. But actually these international institutions, for those who actually are serious about them, it's a de facto alliance. NATO is most certainly an alliance.

Dwarkesh Patel 01:33:26

If you are the biggest trading partner with say 100 countries, do you have a lot of alliances? If not, then that seems more significant than just calling yourself an ally. You just have an incredibly strong vested interest in the other country. The reason America has a lot of alliances is because we have a lot of trade.

Sarah Paine 01:33:44

Oh yeah, yeah, yeah it's important. It's just that the Chinese also come with this little piece of baggage: “We're big, you're small, so we get to tell you what to do.”

Dwarkesh Patel 01:33:53

Is that true though? What is it that they actually tell?

Sarah Paine 01:33:56

Oh, they've literally said it over and over again. They'll tell their neighbors, like the Vietnamese, to just basically shut up and color because we're a big country and you're little. The Vietnamese don't like it.

Dwarkesh Patel 01:34:05

One thing I often hear in these discussions is that it's the United States, which actually has stronger preferences in, for example, how your political system works. It may be reasonable to have, but China is usually more willing to just trade with you if you're willing to trade with them.

Sarah Paine 01:34:18

Oh yeah, yeah, yeah we've been really intrusive in other people's domestic politics, without a doubt. That is the American busybody gene that's gotten us into a lot of trouble. No one's perfect in this world, far from it. We all make horrendous errors and then do things we shouldn't do.

Dwarkesh Patel 01:34:40

I think that's an excellent place to close. Sarah, thanks so much.

Sarah Paine 01:34:43

Thank you.

Jacob Kimmel, the Co-founder & President of NewLimit, thinks he can find the transcription factors to reverse aging. We do a deep dive on why this might be plausible and why evolution hasn’t already optimized for longevity. We also talk about why drug discovery has been getting exponentially harder, and what a new platform for biological understanding to speed up progress would look like. As a bonus, we get into the nitty gritty of gene delivery and Jacob’s controversial takes on CAR-T cells.

For full disclosure, I am an angel investor in NewLimit. This did not impact my decision to interview Jacob, nor the questions I asked him.

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Hudson River Trading uses deep learning to tackle one of the world's most complex systems: global capital allocation. They have a massive in-house GPU cluster, and they’re constantly adding new racks of B200s to ensure their researchers are never constrained by compute. Explore opportunities at hudsonrivertrading.com/dwarkesh

• Google’s Gemini CLI turns ideas into working applications FAST, no coding required. It built a complete podcast post-production tool in 10 minutes, including fully functional backend logic, and the entire build used less than 10% of Gemini’s session context. Check it out on Github now!

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – Three reasons evolution didn’t optimize for longevity

(00:12:07) – Why didn't humans evolve their own antibiotics?

(00:25:26) – De-aging cells via epigenetic reprogramming

(00:44:43) – Viral vectors and other delivery mechanisms

(01:06:22) – Synthetic transcription factors

(01:09:31) – Can virtual cells break Eroom’s Law?

(01:31:32) – Economic models for pharma

Dwarkesh Patel 00:00:00

Today I have the pleasure of interviewing Jacob Kimmel, who is the president and co-founder of NewLimit, where they're trying to epigenetically reprogram cells to their younger states.

Jacob, thanks so much for coming on the podcast.

Jacob Kimmel 00:00:09

Thanks so much for having me. Looking forward to the conversation.

Dwarkesh Patel 00:00:11

All right, first question. What's the first principles argument for why evolution just discards us so easily? I know evolution cares about our kids. But if we have longer, healthier lifespans, we can have more kids, right? We can care for them longer, we can care for our grandkids. Is there some pleiotropic effect that an anti-aging medicine would have which actually selects against you staying young for longer?

Jacob Kimmel 00:00:37

I think there are a couple different ways one can tackle this. One is you have to think about what's the selective pressure that would make one live longer and encode for higher health over longer durations. Do you have that selective pressure present? There's another which is, are there any anti-selective pressures that are actually pushing against that? There's a third piece of this, which is something like the constraints of your optimizer. If we think about the genome as a set of parameters and the optimizer is natural selection, then you've got some constraints on how that actually works. You can only do so many mutations at a time. You have to spend your steps that update your genome in certain ways.

Tackling those from a few different directions, what would the positive possible selection be? As you highlighted, it might be something like, “If I'm able to extend the lifespan of an individual, they can have more children, they can care for those children more effectively. That genome should propagate more readily into the population.” One of the challenges then—if you're trying to think back in a thought experiment style of evolutionary simulation here—would be: What were the conditions under which a person would actually live long enough for that phenotype to be selected for, and how often would that occur?

This brings us back to some very hypothetical questions. Things like, what was the baseline hazard rate during the majority of human and primate evolution? The hazard rate is simply, “What is the likelihood you're going to die on any given day?” That integrates everything. That's diseases from aging, that's getting eaten by a tiger, that's falling off a cliff, that's scraping your foot on a rock and getting an infection and dying from that.

From the best evidence we have, the baseline hazard rate was very, very high. Even absent aging, you're unlikely to actually reach those outer limits of possible health where aging is one of the main limitations. The number of individuals in the population that are going to make it later in that lifespan, where using some of your evolutionary updates to try and push your lifespan upward, is relatively limited. The amount of gradient signal flowing back to the genome then is not as high as one might intuitively think.

Dwarkesh Patel 00:02:29

On that, often people who are trying to forecast AI will discuss how hard evolution tried to optimize for intelligence, and what were the things which optimizing for intelligence would have prevented evolution from selecting for at the same time? So even if intelligence were a relatively easy thing to build in this universe, it would have taken evolution so long to get at human-level intelligence. And, potentially, if intelligence were really easy, then it might imply that we're going to get superintelligence and Jupiter-level intelligence, etc. The sky's the limit.

One argument is birth canal sizes, etc., or the fact that we had to spend most of our resources on the immune system. But what you just hinted at is an independent argument that if you have this high hazard rate, that would imply you can't be a kid for too long. Kids die all the time and you have to become an adult so that you can have your kids.

Jacob Kimmel 00:03:27

You’ve got to contribute resources back to the group. You can't just be a freeloader. You need to get calories, go out in the jungle, get some berries.

Dwarkesh Patel 00:03:33

If you're just hanging out learning stuff for 50 years, you're just going to die before you get to have kids yourself.

Obviously, humans have bigger brains than other primates. We also have longer adolescences, which help us make use potentially of the extra capacity our brain gives us. But if you made the adolescence too long, then you would just die before you get to have kids. If that's going to happen anyways, what's the point of making the brain bigger? AKA, maybe intelligence is easier than we think, and there's a bunch of contingent reasons evolution didn't churn as hard on this variable as it could have.

Jacob Kimmel 00:04:05

I entirely agree with that particular thesis. In biology in general, when you're trying to engineer a given property, be it being healthier longer or be it making something more intelligent… This is true even at the micro-level of trying to engineer a system to manufacture a protein at high efficiency. You always have to start by asking yourself, “Did evolution spend a lot of time optimizing this? If yes, my job is going to be insanely hard. If not, potentially there are some low-hanging fruit.” This is a good argument for why, potentially, intelligence wasn't strongly selected for.

The lifespan argument plays back into intelligence to a degree. You start to ask, “If I have intelligence that's able to compound over time and for instance, in some hypothetical universe, my fluid intelligence lasts much longer into my lifespan…” If the number of people who are reaching something like 65 is very small in a population, you're not necessarily going to select for alleles that lead to fluid intelligence preservation late into life.

This is part of my own pet hypothesis around some of the interesting phenomenology in when discoveries are made throughout lifespans. There are some famous results. For instance—and I'm going to get the exact age a little bit wrong—but in mathematics, most great discoveries happen roughly before 30. Why should that be true? That doesn't make sense.

You can put down a bunch of societal reasons for it. Maybe you become staid in your ways. Your teachers have caused you to restrict your thinking by that point. But really, that's true across centuries? Is that true across many different unique cultures around the world? That's true in both cultures from the East and cultures from the West? That seems unlikely to me.

A much simpler explanation is that for whatever reason, our fluid intelligence is roughly maximized at the time where the population size during human evolution was maximal. If you had to pick an age at which fluid intelligence was selected most strongly for, it's probably around 25 or 30. That's probably about the age of the adults in the large populations that were being selected for during most of evolution.

There's a lot of reason here to think that there's interplay between many features of modern humans and how long we were living, and how that dictates some of the features that occur that rise and fall throughout our lives.

Dwarkesh Patel 00:06:01

In one way, this is a very interesting RL problem. It's a long-horizon RL problem, a 20-year horizon length, and then there's a scalar value of how many kids you have, I guess that survive, etc. If you've heard from your friends about how hard RL is on these models for just very intermediate goals that last an hour or a couple of hours, it's surprising that any signal propagates across a 20-year horizon.

On the point about fluid intelligence peaking, it’s not only the case that in many fields achievement peaks before 30. In many cases, if you look at the greatest scientists ever, they had many of their greatest achievements in a single year.

Jacob Kimmel 00:06:42

Yeah, the annus mirabilis.

Dwarkesh Patel 00:06:44

Yeah, exactly.

Jacob Kimmel 00:06:44

Yeah, exactly.

Dwarkesh Patel 00:06:45

Newton, what is it? Optics, gravity, calculus at 21.

Jacob Kimmel 00:06:49

Do you know the Alexander von Humboldt story?

Dwarkesh Patel 00:06:51

No.

Jacob Kimmel 00:06:52

Alexander von Humboldt is one of the most famous scientists in history who is kind of forgotten now. He had this one expedition to South America where he climbed Mount Chimborazo at a time when very few Europeans had done that. He was able to observe various ecological layers that were repeated across latitudes and across altitudes. It caused him to formulate an understanding of how selection was operating on plants at different layers in the ecosystem. That one expedition was the basis of his entire career.

When you see something named Humboldt, just to give you a sense of how famous this guy is, it's usually Alexander von Humboldt. It's not like this is some massive, prosperous German family name that just happens to be really common. It's this one guy. So really it was like this singular year in which he conceived a lot of our modern understanding of botany and selective pressure.

Dwarkesh Patel 00:07:35

Interesting. So that's one out of three components of the evolutionary story.

Jacob Kimmel 00:07:39

The next piece of the evolutionary story is, “Is there anything selecting against longevity?” Let's just pretend everything I said was wrong. Can I still make an argument that maybe evolution hasn't maximally optimized for our longevity?

One argument that comes up, and I'll caveat and say I don't know how strong some of the mathematical models that people put together here are. You can find people using the same idea to argue for and against. But there's this notion of what's called kin selection.

If you take a selfish gene view of the world—that really this is the genome optimizing for the genome's propagation, it's not trying to optimize for any one individual—then actually optimizing for longevity is a pretty tricky problem because you have this nasty regularization term.

If you're able to make a member of the population live longer, but you don't also counteract the decrease in their fitness over time—meaning you maybe extend maximum lifespan but you haven't totally eliminated aging—then the number of net calories contributed to the genome as a function of that person's marginal year and their own calorie consumption is less than if you were to allow that individual to die and actually have two 20 year olds, for instance, that follow behind them.

So there is a notion by which a population being laden demographically with many aged individuals, even if they did have fecundity persisting out some period later in life, is actually net negative for the genome's proliferation and that really a genome should optimize for turnover and population size at max fitness.

Dwarkesh Patel 00:08:56

I love this idea of aging as a length regularizer. People might be familiar with the idea that when companies are training models, they'll have a regularizer for, “You can do chain of thought, but don't make the chain of thought too long.” You're saying how many calories you consume over the course of your life, is one such regularizer? That's interesting.

The third point was...

Jacob Kimmel 00:09:20

The third piece is basically optimization constraints. So this is where another ML analogy is helpful. Well, actually a two-layer neural network is technically a universal approximator, but we can never actually fit them in such a way. Why does that occur? People will wave their hands, but it basically comes down to the fact that we don't really know how to optimize them, even if you can prove out in a formal sense that they are universal approximators.

I think we have similar optimization challenges with our genome as the parameters and evolution as the optimization algorithm. One of those is that your mutation rate basically bounds the step size you can take. So if you imagine that at each generation, you get some number of inputs, you can select for some number of alleles. The max number of variations in the genome is set by your mutation rate. If you dial your mutation rate up too high, you probably get a bunch of cancers, so you're selected against. If you have it too low, you can't really adapt to anything. You end up with this happy medium, but that limits your total step size. Then the number of variants you can screen in parallel is basically limited by your population size. So for most of evolution, there are lots of forces constraining population size as well.

One of the dominant sources of selection on the genome is really prevention of infectious disease. It seems like when you study the history of early modern man, infectious disease is actually what shaped a lot of our population demographics. There's a lot of pressure pushing for those step sizes, those updates to the genome, really to be optimizing for protection against infectious disease rather than other things.

Even if you imagine that maybe the arguments on the first and the second of these possible positive selection being absent for longevity and potentially some negative selection existing, you could construct a reasonable argument for why humans don't live forever and why the genome hasn't optimized for that, simply based on these optimization constraints. You have to imagine not only that the positive selection is there and the negative selection is absent, but that when you think about the weighted loss term of all the things the genome is optimizing for, that the weight on longevity is high enough to matter. Even if you imagine it's there, if you simply imagine that the lambdas are dialed toward infectious disease resilience more effectively, then you can construct an argument for yourself.

And so I think really when you start to ask “why don't we live forever, why didn't evolution solve this?” you actually have to think about an incredibly contingent scenario where both the positive selection is there, the negative selection is absent, and you have a lot of our evolutionary pressure going toward longevity to solve this incredibly hard problem in order to construct the counterfactual in which longevity is selected for and does arise in modern man and in which we are optimal.

So I think that puts human aging and longevity and health really in this category of problems in which evolution has not optimized for it. Ergo, it should be, relatively speaking, relative to a problem evolution had worked on, easy to try and intervene and provide health. In many ways, the existence of modern medicines, which are incredibly simplistic—we are targeting a single gene in the genome and turning it off everywhere at the same time—the fact that these provide massive benefit to individuals is another sort of positive emission or piece of evidence.

Dwarkesh Patel 00:12:07

Antibiotics are an even more clear case of that because here's something that evolution actually cares a lot about. It feels like antibiotics should be…

Jacob Kimmel 00:12:15

Why didn't humans evolve their own antibiotics?

Dwarkesh Patel 00:12:16

Yeah.

Jacob Kimmel 00:12:17

It's an excellent question that I haven't heard posed before. Where do antibiotics come from? To your point, we could synthesize them. They're just metabolites, largely of other bacteria or other fungi. You think about the story of penicillin. What happens? Alexander Fleming finds some fungi growing on a dish. The fungi secrete this penicillin antibiotic compound. So there's no bacteria growing near the fungi. He says he has this light bulb moment of, “Oh my gosh, they're probably making something that kills bacteria.”

There's no prima facie reason that you couldn't imagine encoding an antibiotic cassette into a mammalian genome. Part of the challenge that you run into is that you're always in evolutionary competition. There's this notion of what's called the Red Queen hypothesis. It's an allusion to the story in Lewis Carroll's Through the Looking Glass, where the Red Queen is running really fast just to stay in place.

When you look at pathogen-host interactions or competition between bacteria and fungi that are all trying to compete for the same niche, what you find is they're evolving very rapidly in competition with one another. It's an arms race. Every time a bacteria evolves a new evasion mechanism, the fungus that occupies the niche will evolve some new antibiotic.

Part of why there is this competitiveness between the two is they both have very large population sizes in terms of number of genomes per unit resource they're consuming. There are trillions of bacteria in a drop of water that you might pick up. There's trillions of copies of the genome. Massive analog parallel computation. And at the same time, they can tolerate really high mutation rates because they're prokaryotic. They don't have multiple cells. If one cell manages to mutate too much and it isn't viable, or it grows too fast, it doesn't really compromise the population and the whole genome. Whereas for metazoans like you and I, if even one of our cells has too many mutations, it might turn into a cancer and eventually kill off the organism.

What I'm getting at, and this is a long-winded way of getting there, is that bacteria and other types of microorganisms are very well adapted to building these complex metabolic cascades that are necessary to make something like antibiotics. It's necessary to maintain that same mutation rate and population size in order to maintain the competition. Even if our human genome stumbled into making an antibiotic, most pathogens probably would have mutated around it pretty quickly.

Dwarkesh Patel 00:14:18

That should imply that through evolutionary history there are millions of “naive antibiotics” which could have acted as antibiotics, but now, basically, all the bacteria have evolved around it.

Do we see evidence of these historical antibiotics that some fungi came up with and the bacteria revolved around and there's evidence for remnants in their DNA?

Jacob Kimmel 00:14:41

I'm going a bit beyond my own knowledge here, but my strong hypothesis would be yes. I can't point to direct evidence today. There are some examples of this. For instance, bacteria that fight off viruses that infect them, bacteriophages, have things like CRISPR systems. You can actually go and look at the spacers, the individual guide sequences that tell the CRISPR system, “Which genome do you go? Where do you cut?”

And you find some of these guides that are very ancient. It seems like this bacterial genome might not have encountered that particular pathogen for quite a while. So you can actually get an evolutionary history of what the warfare was like, what the various conflicts were throughout this genomic history just by looking at those sequences.

In mammals where I do know a bit better, we do have examples of this where there is this co-evolution of pathogen and host. Imagine you have some antipathogen gene A fighting off some virus X. Well you then actually update. Now you have virus X’ and antipathogen gene A’. Now virus X’ goes away, but actually virus X still exists and we've lost our ability to fight it.

Those examples really do happen. There's a prominent one in the human genome. We have a gene called TRIM5alpha. It actually binds an endogenous retrovirus that is no longer present, but was at one point actually resurrected by a bunch of researchers. It was demonstrated that this is the case. We have this endogenous gene which basically fits around the capsid of the virus like a baseball in a glove and prevents it from infecting.

It turns out if you look at the evolutionary history of that gene and you trace it back through monkeys, you can actually find that a previous iteration inhibited SIV, which is the cousin of HIV in humans. Old World monkeys actually can't get SIV, whereas New World monkeys can and humans can, obviously.

So it seems like what happened—and you can actually make a few mutations in TRIM5alpha and find that this is true—is that TRIM5alpha once protected against an HIV-like pathogen in the primate genomes. And then there was this challenge from this massive endogenous retrovirus. It was so bad that the genome lost the ability to fight off these HIV-like viruses in order to restrict this endogenous retrovirus. You can see it because that retrovirus integrates into our genome. There are latent copies, like the half bodies of this virus all throughout our DNA code. Then this particular retrovirus went extinct. Reasons unknown, no one knows why. But we didn't re-update that piece of our host defense machinery to fight off HIV again.

So we're in a situation where you can go in and take human cells and make just a couple edits in that TRIM5alpha gene. It's currently protecting against a virus which no longer exists. You can edit it back to actually restrict HIV dramatically. So there are plenty of examples. You could imagine the same thing for antibiotics where like, “hey, this particular defense mechanism went away because the pathogen evolved its own defense to it.” Well, the pathogen might have lost that defense long ago. If you could extract that historical antibiotic, that historical antifungal, potentially it actually has efficacy.

Dwarkesh Patel 00:17:24

Isn't the mutation rate per base pair per generation like one in a billion or something?

Jacob Kimmel 00:17:27

It's quite low.

Dwarkesh Patel 00:17:29

You're saying that in our genomes we can find some extended sequence which encodes how to bind specifically to the kind of virus that SIV is. The amount of evolutionary signal you would need in order to have a multiple base pair sequence… So each nucleotide consecutively would have to mutate in order to finally get the sequence that binds to SIV. That seems almost implausible. I guess evolution works, so we can come up with new genes, but how would that even work?

Jacob Kimmel 00:18:00

A great explanation for understanding a lot of evolution and how you're able to actually adapt to new environments, new pathogens, is that gene duplication is possible. This explains a whole lot. If you look at most genes in the genome, they actually arise at least at some point in evolution from a duplication event.

That means you've got gene A, it's performing some job, and then some new environmental concern comes along. Maybe it's a lack of a particular source of nutrient, maybe it's a pathogen challenging you. Maybe gene A, if it were to dedicate all of its energies, so to speak and you were to mutate it to solve this new problem, could be adapted with a minimal number of mutations. But then you lose its original function.

So we have this nice feature of the genome which is that it can just copy and paste. So occasionally what will happen in evolution is you get a copy paste event. Now I've got two copies of gene A and I can preserve my original function in the original copy. Then this new copy can actually mutate pretty freely because it doesn't have a strong selective pressure on it. So most mutations might be null. I've got two copies of the gene, I can have lots of mutations in it accumulate. Nothing bad really happens because I've got my backup copy, my original. So you can end up with drift.

Dwarkesh Patel 00:19:04

You're saying that even though the per base pair mutation rate might be one in a billion, if you've got 100 copies of a gene, then the mutation rate on a gene, or on a low Hamming distance sequence to the one you're aiming for, might actually be quite high, and you can actually get the target sequence.

Jacob Kimmel 00:19:21

It's not that the base rate goes up. It's not like DNA polymerase is more erroneous or that you're just doubling it. That is true, but I don't think it's the main mechanism.

One of the main mechanisms that just makes it difficult for evolution to solve a problem is if a mutation breaks a gene. Somewhere along the path of edits, imagine there are three edits that take a host defense gene from restricting SIV to restricting this new nasty PT endogenous retrovirus. Well, if one edit just breaks the gene, two edits just breaks the gene, three edits fixes it, it's really hard for evolution to find a path whereby you're actually able to make those first two edits because they're net negative for fitness. So you need some really weird contingent circumstances.

Through duplication, you can create a scenario where those first two edits are totally tolerated. They have no effect on fitness. You've got your backup copy, it's doing its job. Even though the mutation rate is low, some of these edits actually aren't that large. I'm going to forget the number of edits, for instance in TRIM5alpha, for this particular phenomenon, but it's in the tens. It's not that you need massive kilobase scale rearrangements. It's actually a fairly small number of edits. Basically you can just align the sequence of this gene in New World versus Old World monkeys and then for humans and you find there's a very high degree of conservation.

Dwarkesh Patel 00:20:36

Conceptually, is there some phylogenetic tree of gene families where you've got the transposons and you've got the gene itself, but then you've got the descendant genes which are low Hamming distance? Is there some conceptual way in which they're categorized?

Jacob Kimmel 00:20:51

You can arrange genes in the human genome by homology to one another. What you find is even in our current genome, even without having the full historical record, there are many, many genes which are likely resulting from duplication events.

One trivial way that you can check this for yourself is just go look at the names of genes. Very often you'll see something where it's like gene one, gene two, gene three or type one, type two, type three. If you then go look at the sequences, sometimes those names arise from the fact that they were discovered in a common pathway and they have nothing to do with each other.

A lot of the time it's because the sequences are actually quite darn similar. Really what probably happened is they evolved through a duplication event and then maybe did some swapping with some other genes. And you ended up with these quite similar, quite homologous genes that now have specialized functions.

So when evolution has a new problem to solve, it doesn't have to start from scratch. It starts from what was the last copy of the parameters for encoding a gene that is getting close to solving this. Okay, let's do a copy paste on that and then iterate and fine-tune on those parameters as opposed to having to start with “ab initio, some random stretch of sequence somewhere in the genome has to become a gene.”

Dwarkesh Patel 00:21:50

This is fascinating. Back to aging. You’ll have to cancel your evening plans. I've got so many questions for you.

Jacob Kimmel 00:21:58

Keep going man.

Dwarkesh Patel 00:22:00

So the second reason you gave was that there's selective pressure against people who get old but still keep living, but they're slightly less fit.

Jacob Kimmel 00:22:17

They’re suboptimal from a calorie input perspective, the number of calories they can gather for the population is lower.

Dwarkesh Patel 00:22:21

That's how people love thinking about their grandpas.

Jacob Kimmel 00:22:25

Suboptimal calorie provider right there.

Dwarkesh Patel 00:22:28

A concern you might have about the effects of longevity treatments on your own body is that you will fix some part of the aging process, but not the whole thing. It seems like you're saying that you actually think this is the default way in which an anti-aging procedure would work, because that's the reason evolution didn't optimize for it. We're only fixing half of the aging process and not the whole thing. Whereas sometimes I hear longevity proponents be like, “No, we'll get the whole thing. There's going to be a source that explains all of aging and we'll get it.”

Whereas, your evolutionary argument for why evolution didn't optimize against aging relies on the fact that aging actually is not monocausal and evolution didn't bother to just fix one cause of aging.

Jacob Kimmel 00:23:17

That's correct. I don't think that there is a single monocausal explanation for aging. I think there are layers of molecular regulation that explain a lot. For instance, I have dedicated my career now to working on epigenetics and trying to change which genes cells use because I think that explains a lot of it. But it's not that there is some upstream “bad gene X” and all we have to do is turn that off and suddenly aging is solved.

The most likely outcome is that when we eventually develop medicines that prolong health in each of us, it's not going to fix everything all at once. There's not going to be a singular magic pill. Rather you're going to have medicines that add multiple healthy years to your life, years you can't otherwise get back. But it's not going to fix everything at the same time. You are still going to experience, for the first medicine, some amount of decline over time. This gives you an example, if you think about evolution as a medicine maker in this sort of anthropomorphic context, of why it might not have been selected for immediately.

Dwarkesh Patel 00:25:26

So evolution didn't select for aging. What are you doing? What's your approach at NewLimit that you think is likely to find the true cause of aging?

Jacob Kimmel 00:25:38

We're working on something called epigenetic reprogramming, which very broadly is using genes called transcription factors. I like to think about these as the orchestra conductors of the genome. They don't perform many functions directly themselves, but they bind specific pieces of DNA and then they tell which genes to turn on, which genes to turn off.

They eventually put chemical marks on top of DNA, on some proteins that DNA surrounds. This is one of the answers, this particular layer of regulation called the epigenome. It's the answer to this fundamental biological question of how do all my cells have the same genome but ultimately do very different things? Your eyeball and your kidney have the same code, and yet they're performing different functions. That may sound a little bit simplistic, but ultimately, it's kind of a profound realization.

That epigenetic code is really what's important for cells to define their functions. That's what's telling them which genes to evoke from your genome. What has now become relatively apparent is that the epigenome can degrade with age. It changes. The particular marks that tell your cells which genes to use can shift as you get older. This means that cells aren't able to use the right genetic programs at the right times to respond to their environment. You're then more susceptible to disease, you have less resilience to many insults that you might experience.

Our hope is that by remodeling the epigenome back towards the state it was in when you were young right after development, that you'll be able to actually address myriad different diseases whose one of strong contributing factors is that cells are less functional than when you were at an earlier point in your life. We're going after this by trying to find combinations of these transcription factors that are able to actually remodel the epigenome so that they can bind to just the right places in the DNA and then shift the chemical marks back toward that state when you are a young individual.

Dwarkesh Patel 00:27:13

If you're just making these broad changes to a cell state through these transcription factors which have many effects, are there other aspects of a cell state that are likely to get modified at the same time in a way that would be deleterious. Or would it be a straightforward effect on cell state?

Jacob Kimmel 00:27:33

How I wish it were straightforward. No, it's very likely. Each of these transcription factors binds hundreds to thousands of places in the genome. One way of thinking about it is if you imagine the genome as the base components of cell function, then these transcription factors are kind of like the basis set in linear algebra. It's different combinations and different weights of each of the genes. Most of them are targeting pretty broad programs.

There are no guarantees that aging actually involves moving perfectly along any of the vectors in this particular basis set. And so it's probably going to be a little tricky to figure out a combination that actually takes you backward. There's, again, no guarantees from evolution that it's just a simple reset. It's actually a critical part of the process that we run through as we try to discover these medicinal combinations of transcription factors we can turn on, ensuring that they not only are making an aged cell revert to a younger state...

We measure that a couple different ways. One is simply measuring which genes those cells are using. They use different genes as they get older. You can measure that just by sequencing all of the mRNAs, which are really the expressed form of the genes being utilized in the genome at a given time. You see that aged cells use different genes. Can I revert them back to a younger state? Colloquially, we call this a "looks like" assay. Can I make an old cell look like a young one based on the genes it's using?

More importantly, we go down and drill to the functional level. We measure, “Can I actually make an aged cell perform its functions, its object roles within the body, the same way a young cell would?” These are the really critical things you care about for treating diseases. Can I make a hepatocyte, a liver cell in Greek, function better in your liver so it's able to process metabolites like the foods you eat, how it's able to process toxins like alcohol and caffeine? Can I make a T cell respond to pathogens and other antigens that are presented within your body?

These are the ways in which we measure age. We need to ensure that not only does the combination of TFs that we find actually have positive effects along those axes. But we then want to also measure any potential detrimental effects that emerge. There are canonical examples where you can seemingly reverse the age of a cell, for instance, at the level of a transcriptome, but simultaneously, you might be changing that cell's type or identity.

Shinya Yamanaka was a scientist who won the Nobel in 2012 for some work he did in about 2007, where he discovered that you could just take four transcription factors and actually, just by turning on these four genes, turn an adult cell all the way back into a young embryonic stem cell. It's a pretty amazing existence proof that shows that you can reprogram a cell's type and a cell's age simultaneously, just by turning on four genes. Out of the 20,000 genes in the genome, the tens of millions of biomolecular interactions, just four genes is enough. That's a shocking fact.

We actually have known for many years now that you can reprogram the age of a cell. The challenge is that simultaneously, you're doing a bunch of other stuff, as you alluded to. You're changing its type, and that might be pathological. If you did that in the body, it would probably cause a type of tumor called a teratoma.

So we measure not only at the level of the genes a cell is using. Do you still look like the right type of cell? Are you still hepatocyte? Are you still a T cell? If not, that's probably pathological. You can also use that same information to check for a number of other pathologies that might develop. Did I make this T cell hyperinflammatory in a way that would be bad? Did I make this liver cell potentially neoplastic, proliferate too much even when the organism's healthy and undamaged?

You can check for each of those at the level of gene expression programs and likewise, functionally. Before you put these molecules in a human, you actually just functionally check in an animal. You make an itemized list of the possible risks you might run into. Here are the ways it might be toxic, here are the ways it might cause cancer. Are we able to measure deterministically and empirically that that doesn't actually occur?

Dwarkesh Patel 00:30:52

This is a dumb question, but it will help me understand why an AI model is necessary to do any of this work. You mentioned the Yamanaka factors. From my understanding, the way he identified these four transcription factors was that he found the 24 transcription factors that have high expression in embryonic cells, and then he just turned them all on in a somatic cell. Basically, he systematically removed from this set until he found the minimal set that still induces a cell to become a stem cell.

That doesn't require any fancy AI models. Why can't we do the same things for the transcription factors that are expressed more in younger cells as opposed to older cells, and then keep eliminating from them until we find the ones that are necessary to just make a cell young?

Jacob Kimmel 00:31:42

I wish it were so easy. You're entirely right. Shinya Yamanaka was able to do this with a relatively small team, with relatively few resources, and achieve this remarkable feat. It's entirely worth asking. Why can't a similar procedure work for arbitrary problems in reprogramming cell state? Whether it be trying to make an aged cell act like a young one, a disease cell act like a healthy one, why can't you just take 24 transcription factors and randomly sort through them?

There were two features of Shinya's problem that I think make it amenable to that sort of interrogation that aren't present for many other types of problems. This is why he's such a remarkable scientist. Most of science is problem selection. You don't actually get better at pipetting or running experiments after a certain age, but you do get better at picking what to do. He's amazing at this.

The first feature is that measuring your success criterion is trivial in the particular case he was investigating. He's starting with somatic cells that, in this case, were a type of fibroblast, which literally is defined as cells that stick to glass and grow in a dish when you grind up a tissue. It sounds fancy, but it's a very simplistic thing. He's starting with fibroblasts, you can look at them under a microscope, and you can see they’re fibroblasts just based on how they look.

Then the cells he's reprogramming toward are embryonic stem cells. These are tiny cells, they're mostly nucleus. They grow really fast. They look different, they detach from a dish, they grow up into a 3D structure. They express some genes that will just never be turned on in a fibroblast by definition.

How he ran the experiment was he just set up a simple reporter system. He took a gene that should never be on in a fibroblast, should only be on in the embryo, and he put a little reporter behind it so that these cells would actually turn blue when you dumped a chemical on them. Then he ran this experiment in many, many dishes with millions upon millions of cells.

The second really key feature of the problem is this notion that those cells he's converting into amplify. They divide and grow really quickly. In order for you to find a successful combination, you don't actually need it to be efficient almost at all. The original efficiency Yamanaka published, the number of cells in the dish that convert from somatic to an induced pluripotent state, back into a stem cell, is something like a basis point or a tenth of a basis point, so 0.01%, 0.001%. If these cells were not growing and they were not proliferating like mad, you probably would never be able to detect that you had actually found anything successful. It's only because success is easy to measure once you have it and—even being successful in very rare cases, one in a million—amplifies and you can detect it, that this was amenable to his particular approach.

In practice, what he would do is dump these factors or this group of 24 minus some number, eventually whittling it down to four. He would dump these onto a group of cells and over the course of about 30 days, just a few cells in that dish, like a countable number on your fingers, would actually reprogram. But they would proliferate like mad. They form these big colonies. It's a single cell that just proliferates and forms a bunch of copies of itself. They form these colonies. You can see with your eyeballs by holding the dish up to the light and looking for opaque little dots on the bottom. You don't need any fancy instruments. Then you could stain them with this particular stain and they would turn blue based on the genetic reporter he had.

We look at those key features of the problem and we pick any other problem we're interested in. I'm interested in aging, so that's the one I'm going to pick for explanation. How difficult is it to measure the likelihood of success or whether you've achieved success for cell age? It turns out age is much more complicated in terms of discriminating function than actually just comparing two types of cells. An old liver cell and a young liver cell, prima facie, actually look pretty darn similar. It's actually quite nuanced the ways in which they're distinct. There isn't a simple, trivial system where you just label your one favorite gene or you can just…

Dwarkesh Patel 00:35:09

Give the young cells cancer. They'll grow.

Jacob Kimmel 00:35:13

Just make the old ones cancer, and then they'll grow. Dwarkesh, you've solved it for me. There's no trivial way that you can tell whether or not you've succeeded. You actually need a pretty complex molecular measurement. For us, a real key enabling technology—I don't think our approach would really have been possible until it emerged—was something called single-cell genomics. You now take a cell, rip it open, sequence all the mRNAs it's using.

At the level of individual cells, you can actually measure every gene that they're using at a given time and get this really complete picture of a cell's state, everything it's doing, lots of mutual information to other features. From that profile, you can train something like a model that discriminates young and aged cells with really high performance. It turns out there's no one gene that actually has that same characteristic. Unlike in Yamanaka's case, where a single gene on or off is an amazing binary classifier, you don't have that same feature of easy detection of success in aging. The second feature is, as you highlighted, we can't just turn these into cancer cells. Success doesn't amplify.

In some ways, the bar for a medicine is higher than what Yamanaka achieved in his laboratory discovery. You can't just have 0.001% success and then wait for the cells to grow a whole bunch in order to treat a patient's disease or make their liver younger, make their immune system younger, make their endothelium younger. You need to actually have it be fairly efficient across many cells at a time. Because of this, we don't have the same luxury Yamanaka did of taking a relatively small number of factors and finding a success case within there that was pretty low efficiency. We actually need to search a much broader portion of TF space in order to be successful.

And when you start playing that game, and you think “How many TFs are there?” Somewhere between 1000 and 2000, it depends on exactly where you draw the line. Developmental biologists love to argue about this over beer, but let's call it 2000 for now. You want to choose some combination. Let's say you guess somewhere between one and six factors might be required. The number of possible combinations is about 10^16. If you do any math on the back of a napkin, in order to just screen through all of those, you would need to do many orders of magnitude more single-cell sequencing than the entire world has done to date cumulatively across all experiments. It's just not tractable to do exhaustively.

That's where actually having models that can predict the effect of these interventions comes in. If I can do a sparse sampling, I can test a large number of these combinations. I can start to learn the relationship of what a given transcription factor is going to do to an aged cell. Is it going to make it look younger? Is it going to preserve the same type? I can learn that across combinations. I can start to learn their interaction terms. Now I can use those models to actually predict in silico for all the combinations I haven't seen, which are most likely to give me the state I want. You can actually treat that as a generative problem and start sampling and asking which of these combinations is most likely to take my cell to some target destination in state space. In our case, I want to take an old cell to a young state, but you could imagine some arbitrary mappings as well.

As you get to these more complex problems, you don't have the same features that Shinya benefited from, which were the ability to measure success really easily—you can see it with your bare eyes, you don't even need a microscope—and two, amplification, as you get into these more challenging problems. You're going to need to be able to search a larger fraction of the space to hit that higher bar.

Dwarkesh Patel 00:38:11

So we can think of these transcription factors as these basis directions, and you can get a little bit of this thing, a little bit of that thing and some combination.

And evolution has designed these transcription factors to…Is that your claim? They have relatively modular, self-contained effects that work in predictable ways with other transcription factors and so we can use that same handle to our own ends?

Jacob Kimmel 00:38:37

That would be very much my contention. One piece of evidence for this is that's the way development works. It's a crazy thing to think about, but you and I were both just a single cell. Then we were a bag of undifferentiated cells that were all exactly alike. Somehow we became humans with hundreds of different cell types all doing very different things.

When you look at how development specifies those unique fates of cells, it is through groups of these transcription factors that each identify a unique type. In many cases, the groups of transcription factors, the sets that specify very different fates, are actually pretty similar to one another. Evolution has optimized to just swap one TF in or swap one TF out of a combination and get pretty different effects. You have this sort of local change in sequence or gene set space leading to a pretty large global change in output.

Likewise, many of these TFs are duplicated in the genome. Because mutations are going to be random and they're inherently small changes at the level of sequence at a given time, evolution needs a substrate where, in order to function effectively, these small changes can give you relatively large changes in phenotype. Otherwise it would just take a very long time across evolutionary history for enough mutations to accumulate in some duplicated copy of the gene for you to evolve a new TF that does something interesting.

I think we're actually in most cases in biology—due to that evolution constraint, small edits need to lead to meaningful phenotypic changes—in a relatively favorable regime for generic, gradient-like optimizers. It would be a little bit overstating to say evolution is using the gradient, but there is a system. If you've heard of evolution strategies, where basically the way you optimize parameters is you can't take a gradient on your loss. So you make a bunch of copies of your parameters, you randomly modify them, and then you compute a gradient on your parameters against your loss, and so you can take a gradient in that space. That's how I imagine evolution is working. So you need lots of those little edits to actually lead you to have meaningful step sizes in terms of the ultimate output that you have.

Dwarkesh Patel 00:40:32

Interesting. You're just like designing a little LoRA that goes on top.

Jacob Kimmel 00:40:37

In a way. Maybe this is getting too giga-brained about it, but why does the genome even have transcription factors? What's the point? Why not just have it so every time you want a new cell type, you engineer some new cassette of genes or some new, totally de novo set of promoters or something like this?

One possible explanation for their existence, rather than just an appreciation for their presence, is that having transcription factors allows a very small number of base pair edits at the substrate of the genome to lead to very large phenotypic differences. If I break a transcription factor, I can delete a whole cell type in the body. If I retarget a transcription factor to different genes, I can dramatically change when cells respond and have hundreds of their downstream effector genes change their behavior in response to the environment.

It puts you in this regime where transcription factors are a really nice substrate to manipulate as targets for medicines. In some ways they might be evolution's levers upon the broader architecture of the genome. By pulling on those same levers that evolution has gifted us, there are probably many useful things we can engender upon biology.

Dwarkesh Patel 00:41:42

You're sort of hinting that if we analogize it to some code base, we're going to find a couple of lines that are commented out that's like, "de-aging," and then "un-hyphen" or "un-parenthesize."

Jacob Kimmel 00:41:53

I don't know about that, but I can give you a real cringe analogy that sometimes I deploy. It requires a very special audience. I think you'll probably be one who fits into it.

Dwarkesh Patel 00:42:01

You're flattering our listeners. “Only cringe listeners will appreciate it, but your audience will love this.”

Jacob Kimmel 00:42:09

I don't know about your audience, but you will. You can think about it like this. If you think about how attention works—queries, keys, values—TFs are like the queries. The genome sequences they bind to are like the keys. Genes are like the values. It turns out that that structure then allows you to very efficiently, in terms of editing space, change just one of those embedding vectors, in this case one of those sequences, and get dramatically different performances or total outputs.

So I do think it's interesting how these structures recur throughout biology, in the same way that the attention mechanism seems to exist in some neural structures. It's interesting that you can very easily see how that same sort of querying and information storage might exist in the genome.

Dwarkesh Patel 00:42:50

Interesting. A previous guest and a mutual friend, Trenton Bricken, had a paper in grad school about how the brain implements attention.

Jacob Kimmel 00:42:58

Eddie Chang has found positional encodings probably exist in humans using neuropixels, if you haven't read these papers. He implants these neuropixel probes into individuals and then he's able to talk to them, look at them as they read sentences. What he finds is that there seem to be certain representations which function as a positional encoding across sentences. They fire at a certain frequency and it just increases as the sentence goes on and then resets. It seems exactly like what we do when we train large language models.

Dwarkesh Patel 00:43:24

It's so funny the way we're going to learn how the brain works is just trying to first-principles engineer intelligence in AI. Then it just happens to be the case that each one of these things has a neural correlate.

Dwarkesh Patel 00:44:43

If you're right that transcription factors are the modality evolution has used to have complex phenotypic effects, optimize for different things...

Two-part question. One, why haven't pathogens, which have a strong interest in having complex phenotypic effects on your body, also utilized the transcription factors as the way to fuck you over and steal your resources?

Two, we've been trying to design drugs for centuries. Why aren't all the big drugs, the top-selling drugs, ones that just modulate transcription factors?

Jacob Kimmel 00:45:23

Why don't we have a million of these pills? I'll try and take those in stride. They're pretty different answers.

The first answer is that there are pathogens that utilize transcription factors as part of their life cycle. A famous example of this is HIV. HIV encodes a protein called Tat, and Tat actually activates NF-κB. HIV, to back up a little bit, is a retrovirus. It starts out as RNA, turns itself into DNA, shoves itself into the genome of your CD4+ T cells

It needs this ornate machinery to actually control when it makes more HIV and when it goes latent so it can hide and your immune system can't clear it out. This is why HIV is so pernicious. You can kill every single cell in the body that's actively making HIV with a really good drug. But then a few of them that have lingered and hunkered down just turn back on. People call this the latent reservoir.

Dwarkesh Patel 00:46:08

Similar to Hep B, right?

Jacob Kimmel 00:46:09

Hep B, Hep C, can both do this sort of latent behavior. HIV is probably the most pernicious of these. One way it does it is that this gene called Tat actually interacts with NF-κB. NF-κB is a master transcription factor within immune cells. Typically if I'm going to horribly reduce what it does, and some immunologists can crucify me later, it increases the inflammatory response of most cells. They become more likely to attack given pathogens around them on the margin. It'll turn on NF-κB activity and then use that to drive its own transcription and its own life cycle.

I can't remember quite all the details now exactly of how it works. But part of this circuitry is what allows it to—in some subset of cells where some of that upstream transcription factor machinery in the host might be deactivated—it goes latent. As long as the population of cells it's infecting always has a few that are turning off the transcription factors upstream that drive its own transcription, then HIV is able to persist in this latent reservoir within human cells. It's just one example offhand.

There are a number of other pathogens. Unfortunately, I don't have quite as much molecular detail in some of these. But they will interface with other parts of the cell that eventually result in transcription factor translocation to the nucleus and then transcription factors being active. This actually segues a little bit to your second question on why there aren’t more medicines targeting TFs.

In a way many of our medicines, ultimately downstream, are leading to changes in TF activity, but we haven't been able to directly target them due to their physical location within cells. So we go several layers upstream. If you think about how a cell works in sensing its environment, it has many receptors on the surface. It has the ability to sense mechanical tension and things like this. Ultimately, most of what these signaling pathways lead to is to tell the cell, "Use some different genes than you're using right now." That's often what's occurring. That ultimately leads to transcription factors being some of the final effectors in these signaling cascades.

A lot of the drugs we have that, for instance, inhibit a particular cytokine that might bind a receptor, or they block that receptor directly, or maybe they hit a certain signaling pathway… Ultimately, the way that they're exerting their effect is then downstream of that signaling pathway, some transcription factor is either being turned on or not turned on. You're using different genes in the cell.

We're kind of taking these crazy bank shots because we can't hit the TFs directly. That sort of begs the question, “Why can't you just go after the TF directly?” Traditionally, we use what are called small molecule drugs, where they're defined just by their size. The reason they have to be small is they need to be small enough to wiggle through the membrane of a cell and get inside.

Then you run into a challenge. If you want to actually stick a small molecule between two proteins that have a pretty big interface—meaning they've got big swaths on the side of them that all sort of line up and form a synapse with one another—then you would need a big molecule in order to inhibit that. It turns out that TF's binding DNA is a pretty darn big surface. Small molecules aren't great at disrupting that and certainly even worse at activating it. Small molecules can get all the way into the nucleus, but they can't do much once they're there. They're just too small.

The other classic modalities we have are recombinant proteins. We make a protein like a hormone in a big vat. We grow it in some Chinese hamster ovary cells, we extract it, we inject it into you. This is how, for instance, human insulin works that we make today. Or you make antibodies produced by the immune system. These run around and find proteins that have a particular sequence, they bind to it, and often they just stop it from working by glomming a big thing onto the side. Those are too big to get through the cell membrane. Then they can't actually get to a TF or do anything directly. So we take these bank shots.

What changes that today, and why I think it's pretty exciting, is we now have new nucleic acid and genetic medicines where you can, for instance, deliver RNAs to a cell that can get through using tricks like lipid nanoparticles. You wrap them in a fat bubble. It looks kind of like a cell membrane. It can fuse with a cell, and put the mRNAs in the cytosol. You can make a copy of a transcription factor there, and then it translocates to the nucleus the same way a natural one would and exerts its effect.

Likewise, there are other ways to do this using things like viral vectors, but we've only very recently actually gotten the tools we need to start addressing transcription factors as first-class targets rather than treating them as maybe some ancillary third-order thing that's going to happen.

Dwarkesh Patel 00:50:07

Interesting. So the drugs we have can't target them, but your claim is that a lot of drugs actually do work by binding to the things we actually can target and those have some effect on transcription factors.

This brings us to questions about delivery, which is the next thing I want to ask you. You mentioned lipid nanoparticles. This is what the COVID vaccines were made of. The ultimate question if we're going to work on de-aging… Even if you identify what is the right transcription factor to de-age a cell, and even if they're shared across cell types, or you figure out the right one for every single cell type, how do you get it to every single cell in the body?

Jacob Kimmel 00:50:49

How do you deliver stuff? How do you get them in there?

There are many ways one could imagine solving it. I'll narrow the scope of the problem. Delivering nucleic acid is a pretty good first-order primitive. Ultimately, the genome is nucleic acids, the RNAs that come out of it are nucleic acids. If you can get nucleic acid into a cell, you can drug pretty much anything in the genome effectively. You can reduce this problem to asking, “How do I get nucleic acids wherever I want them to any cell type very specifically?”

Today, there are two main modalities that people use, both of which have some downsides. The first one that we've touched on already is lipid nanoparticles. These are basically fat bubbles. By default, they get taken up by tissues which take up fat, like the liver. They can be used like trojan horses. They can release some arbitrary nucleic acid—usually RNA, maybe encoding your favorite genes, in our case, transcription factors—into the cell types of interest. You can play with the fats, and you can also tie stuff onto the outside of the fat. You can attach part of an antibody, for example, to make it go to different cell types in the body. The field is making a lot of progress on being able to target various different cell types with lipid nanoparticles. Even if nothing else worked for the next several decades, companies like ours would have more than enough problems to solve with the cells that we can actually target.

Another prominent way people go after this is using viral vectors. The basic idea being viruses had a lot of evolutionary history and very large population sizes. They've evolved to get into our cells. Maybe we can learn something from them, even better than Trojan horses. One type of virus people use a lot is called an AAV. Those AAVs carry DNA genomes. You can get genes, whole genes, into cells. They've got some packaging sizes. You can think of it like a very small delivery truck, so you can't put everything you want into it. They can go to certain cell types as well. On top of just where you actually get the nucleic acid to begin with, you can engineer the sequences a bit, and that basically allows you to add a NOT gate on it. You can make it turn off the nucleic acid in certain cell types, but you're never going to use the sequence engineering to get nucleic acid into cells where it didn't get delivered in the first place. You can start broad with your delivery vector and then use sequence to narrow down to make it more specific, but not the other way around.

Both of those methods are super promising. If nothing else emerged for decades, we'd still have tons and tons of problems as a therapeutic development community to solve, even using just those.

I have one very controversial opinion which people can roast me for later.

Dwarkesh Patel 00:53:04

You have just one? You're trying to solve aging and you have only one?

Jacob Kimmel 00:53:08

I have many controversial opinions. One of them is that both of these probably in the limit will not be the way that we're delivering medicines in the year 2100. If you think about viral vectors, no matter what, they're always going to be some amount of immunogenic. You're always going to have your immune system trying to fight them off. You can play tricks, you can try and cloak them, etc., but they're always going to have some toxicity risk. They also don't go everywhere. It's not that we have examples of a single viral species that infects every cell type in the body and we just need to engineer it to make it safe. We would have to also engineer the virus to go to new cell types. There's some limitations there.

LNPs likewise have some problems. They can go to tons of cell types. That's largely what we're working on. We're super excited about it. But there are some physical constraints. They just have a certain size. They have to get from your bloodstream out of your bloodstream toward a given target cell, and they have to not fuse into any of the other cells along the way. There's a whole gamut they have to run.

Ultimately, we're probably going to have to solve delivery the way that our own genome solved delivery. We have the same problem that arose during evolution. How do I patrol the body, find arbitrary signals in the environment, and then deliver some important cargo there when some set of events happens? How do I find a specific place and only near those cell types release my cargo? The problem was solved by the immune system. We have cell types in our body, T cells and B cells, which are effectively engineered by evolution to run around, invaginate whatever tissues they need to. They can climb almost anywhere in the body. There's nowhere they can't get access to, almost. Once they sense a particular set of signals—and they've got a very ornate circuitry to do this, they run basically an AND gate logic—they can release a specified payload.

Right now, the way our genome sets them up, the payload they release is largely either enzymes that will kill some cell that they're targeting or kill some pathogen, or some signal flares that call in other parts of the immune system to do the same thing. So that's super cool. But you can think about it as a modular system that evolution's already gifted us. We've got some signal and environmental recognition systems so we can find particular areas of the body that we want to find. Then we have some sort of payload delivery system. I can deliver some arbitrary set of things.

I imagine if we were to Rip Van Winkle ourselves into 2100 and wake up, the way we will be delivering these nucleic acid payloads is actually by engineering cells to do it, to perform this very ornate function. Those cells might actually live with you. You probably will get engrafted with them, and they might persist with you for many years. They deliver the medicine only when the environment within your body actually dictates that you need it. You actually won't be seeing a physician every time this medicine is active. Rather, you'll have a more ornate, responsive circuit.

The other exciting thing about cells is that they're big and they have big genomes. You actually have a large palette to encode complex infrastructure and complex circuitry. You don't need to limit yourself to the very small RNAs you can get in that might encode a gene or two, or in our case, a few transcription factors. You don't have to limit yourself to this tiny AAV genome that's only a few kilobases. You've got billions of base pairs to play with in terms of encoding all your logic. So I think that's ultimately how delivery will get solved. We've got many, many stepping stones along the way. But if I could clone myself and work on an even riskier endeavor, that's probably what I would do.

Dwarkesh Patel 00:56:12

In a way, we treat cancer this way with CAR-T therapy, right? We take the T cells out and then we tell them to go find a cancer with this receptor and kill it. Is the reason that works that the cancer cells we're trying to target are also free floating in the blood? Is that what it targets?

Basically, could this deliver to literally every single cell in the body?

Jacob Kimmel 00:56:34

Not literally every single cell. I'll asterisk it there. For example, T cells don't go into your brain. They can, but it's generally a pathology when they get in there. It's not literally every cell, but almost every cell in your body is surveilled by the immune system. There are very, very few what we call immune-privileged compartments in your body. It's things like the joints of your knees and your shoulders, your eyeball, and your brain. There might be a couple of others. The ear probably falls into that category.

A funny way of thinking about this is that all the gene-therapy people using viruses, they want to deliver to the immune-privileged compartments because their drugs are immunogenic, and they're limited to a very, very small set of diseases. In a way, it's like the shadow of all the diseases you can't address with viruses is what you can address with cells. Given the complementarity between them, you can probably cover the entire body.

They can't literally go everywhere. But your analogy to the CAR-T work is very apt as well. You can think about that two-component system. I've got some detection mechanism for the environment I want to sense to perform some function, and then I have some sort of payload that I deliver. CAR-Ts engineer the first of those and leave the second exactly the same as the immune system does. They engineer—go recognize this other antigen that you wouldn't usually target, some protein on the surface of a cell, for instance—and then deliver the payload you would usually deliver if it was infected by a virus or if you saw that it was foreign in some way. Whereas cancer cells usually don't actually look that foreign. Most of their genes are the same genes that are in your normal genome, and that's why it's hard for the immune system to surveil it.

Dwarkesh Patel 00:57:57

Interesting. Interesting. It's funny that whenever we're trying to cure infectious diseases, we just have to deal with, "Fuck, viruses have been evolving for billions of years with our oldest common ancestor, and they know exactly what they're doing, and it's so hard." Then whenever we're trying to do something else, we're like, "Fuck, the immune system has been evolving for billions of years, and it knows what it's doing, and how do we get past it?"

Jacob Kimmel 00:58:20

The Red Queen race is quite sophisticated. If you want to just throw a new tool into biology, you somehow have to get around one side of that equation.

Dwarkesh Patel 00:58:27

Given the fact that it's somewhere between impossible and very far away but it's necessary for full curing of aging, does that mean that in the short run, in the next few decades, we'll have some parts of our body which will have these amazing therapies, and then other parts which will just be stuck the way they are?

You mentioned hepatocytes are some of the cells that you're able to actually study in or deliver to. These are our liver cells. So you're saying, “Look, I can get drunk as much as I want and it's not going to have an impact on my long-run liver health because then you'll just inject me with this therapy.” But for the rest of my body, it's going to age as normal? What is the implication of the fact that the delivery seems to be lagging much behind your understanding of aging?

Jacob Kimmel 00:59:19

Just to give the delivery folks credit, they're currently ahead. There are currently no reprogramming medicines for aging, and there are medicines that deliver nucleic acids. They're still winning the race against us right now, but to your point, I hope the lines cross. I hope we outcompete them.

Even if you were able to only target some subsets of cells, it's not that you would see this strange, Frankensteinian benefit in health in some aspects and lack of benefit entirely in others. What we've found across the history of medicine is that the body's an incredibly interconnected, complex system. If you're able to rescue function even in one cell type in one tissue, you often have knock-on benefits in many places that you didn't initially anticipate.

One way we can get examples of this is through transplant experiments. Both in bone marrow and in liver, for example, we have fairly common transplant procedures that occur in humans. We can compare old humans who get livers from young people or old people and ask a pretty controlled question. What occurs as a function of just having a young liver? Is it that, for example, you can eat a lot of fatty food and drink a lot and be fine? Or is it that you see broader benefits?

The latter seems to be true. They have reduced risk of several other diseases and overall better survival as a function of having a younger liver than they do for an older one. Suggesting that because these tissues are so interconnected, many of these organs like the liver, like your adipose tissue, are endocrine organs. They're also sending out signals to many other places in your body, helping coordinate your health across multiple tissue systems. Even just one tissue can benefit other tissue systems in your body at the same time.

HSCs are another example. These are mostly examples taken from a wonderful book by Frederick Appelbaum, who trained with Don Thomas, the physician who invented human bone marrow transplants. There are many circumstances where patients got a bone marrow transplant and actually cured another disease they had as a result, maybe unanticipated, where it's even just the replacement of this one special cell type, HSCs, that has knock-on effects throughout the body. There were symptoms of these diseases that presented in myriad ways throughout their system, but ultimately, its root cause was even just a single cell.

There are counterexamples as well where you can go into animals and break even just one gene in one specific subset of T cells. You can break a gene in there that encodes for a transcription factor in their mitochondria called TFAM, and you dramatically shorten the lifespan of mice. One gene in one special type of T cells can give you that type of pathology. So it implies the inverse may also exist.

Dwarkesh Patel 01:01:50

Is this related to why Ozempic has so many downstream positive effects that seem even not totally related to its effects just on making you leaner?

Jacob Kimmel 01:02:01

I think it's one example. It is a hormone, and your endocrine system coordinates a lot of the complex interplay between your tissues. I don't think the story is fully written yet on exactly why GLP-1 and GIP-1, broadly incretin mimetic medicines like Ozempic, have so many knock-on benefits, but they're a great example of this phenomenon.

If someone told you, "I'm going to find a single molecule, and I'm going to drug it, and it's not only going to have benefits for weight loss but also for cardiovascular disease, also possibly for addictive behavior, and maybe even preventing neurodegeneration," you would have told them they were crazy. Yet, just by acting on the small number of cells in your body which are receiving this signal, the interplay and the communication between those cells and the rest of your body seems to have many of these knock-on benefits.

It's just one existence proof that very small numbers of cells in your body can have health benefits everywhere. Even if cellular delivery does not emerge by 2100, as I imagine it will, I still think that you're going to have the ability to add decades of healthy life to individuals by reprogramming the age of individual cell types and individual tissues.

Dwarkesh Patel 01:03:04

How big will the payload have to be?

Jacob Kimmel 01:03:06

How many transcription factors? I think it's just a countable number. Some of those that we've found today that have efficacy are somewhere between one and five. That's a small enough number that you can encapsulate it in current mRNA medicines. Already in the clinic today, there are medicines that deliver many different genes as RNA. There are medicines where, for instance, it's a vaccine as a combination of flu and COVID proteins, and they're delivering 20 different unique transcripts all at the same time. When you think about that already as a medicine that's being injected into people in trials, the idea of delivering just a few transcription factors is seemingly quotidian. Thankfully, I don't think we'll be limited by the size of the payloads that one can deliver.

One other really cool thing about transcription factors is that the endogenous biology is very favorable for drug development. The expression level of transcription factors in your genome relative to other genes is incredibly low. If you just look at the rank-ordered list of what are the most frequently expressed genes in the genome by the count of how many mRNAs are in the cell, transcription factors are near the bottom.

That means you don't actually need to get that many copies of a transcription factor into a cell in order to have benefits. What we've seen so far, and what I imagine will continue to play out, is that even fairly low doses of these medicines, which are well within the realm of what folks have been taking for more than a decade. They are able to induce really strong efficacy. We're hopeful that not only will the actual size of the payload in terms of number of base pairs not be limiting, but the dose shouldn't be limiting either.

Dwarkesh Patel 01:04:36

Would it have to be a chronic treatment, or could it just be a one-time dose?

Jacob Kimmel 01:04:41

In principle, it could be one time. I think that would be an overstatement for today. I can talk you through the evidence from the first principles back to the reality of what's the hardest thing we have in hand.

Epigenetic reprogramming is basically how the cell types in our bodies right now are able to adopt the identities that they have. The existence proof that those epigenetic reprogramming events can last decades is that my tongue doesn't spontaneously turn into a kidney. These epigenetic marks can persist for decades throughout a human life, or hundreds of years if you want to take the example of a bowhead whale which uses the same mechanism.

We also know that with very targeted edits, other groups have done this, folks like Luke Gilbert now at the Arc Institute, who I think of as one of the great unsung scientists of our time, have been able to make a targeted edit in a single locus and then show that you can actually make cells divide 400-plus times over multiple years in an incubator in the lab. Imagine a hothouse where you're just trying as hard as you can to break this mark down, and it can actually persist for many years. Other companies have actually now dosed some editors similar to the ones that Luke developed in his lab in monkeys and shown they last at least a couple of years.

In principle, the upper bound here is really long. You could potentially have one dose and it lasts a very long time, potentially decades, as long as it took you to age the first time maybe. We don't have data like that today. I don't want to overstate. We do have data that these positive effects can last several weeks after a dose. You could imagine, even without many leaps of faith up toward this upper bound limit of what's possible just from the data we have in hand now, that you could get doses every month, every few months and actually have really dramatic benefits that persist over time, rather than needing to get an IV every day, which might not be tractable.

Dwarkesh Patel 01:06:22

We've got 1600 transcription factors in the human genome. Is it worth looking at non-human TFs and seeing what effects they might have, or are they unlikely to be the right search space?

Jacob Kimmel 01:06:34

I think it's less likely. I think you have a prior that evolution has given you a reasonable basis set for navigating the states that human cells might want to occupy. In our case, we know that the state we're trying to access is encoded by some combination of these TFs. It does arise in development obviously. We're trying to make an old cell look young, not look like some Frankenstein cell that's never been seen before.

That said, we don't have any guarantees that the way aging progresses is by following the same basis set of these transcription factor programs in the genome that are encoded during development. I don't think it's unreasonable to ask, “Would your eventual ideal reprogramming medicine necessarily be a composition of the natural TFs, or would it include something like TFs from other organisms, as you posit, or even entirely synthetic transcription factors as well?”

Things like Super-SOX. Super-SOX is a particular publication from Sergiy Velychko where they mutated the SOX2 gene and they made more efficient iPSC reprogramming. They could take somatic cells and turn them into pluripotent stem cells more effectively than you could with just the canonical Yamanaka factors, which are Oct-4, Sox2, KLF4, and Myc.

iPSC reprogramming never happens in nature, so there's no reason to necessarily believe that the natural TFs are optimal.

So even really simple optimizations, like just mutagenizing one of the four Yamanaka factors we already know about or swapping some domains between a few TFs, seem to improve things dramatically. I think that's a pretty good signal that actually there's a lot of gradient to climb here and that potentially for us, the end-state products we're developing in 2100 are more like synthetic genes that have never existed, rather than just compositions of the natural set.

Dwarkesh Patel 01:08:10

What about the effects of aging? Your skin starts to sag because of the effects of gravity over the course of decades? Is that a cellular process? How would some cellular therapy deal with that?

Jacob Kimmel 01:08:23

The best evidence is that it's probably not cellular. The reason your skin sags is there's a protein in your skin called elastin, which does exactly what you'd think it would based on the name. It kind of keeps your skin elastic-y, like a waistband, and holds it to your face. You have these big polymerized fibers of elastin in your face. As far as we understand it, you only polymerize it and form a long fiber during development. Then the rest of your life, you make the individual units of the polymer, but for reasons no one as far as I can tell understands, they fail to polymerize. You can't make new long cords to hold your skin up to your face.

So the eventual solution for something like that is likely that you need to program cells to states that are extra-physiological. There might not be a cell in your body. It's not just like a young skin cell from a 20-year-old is better at making these fibers. As far as we can tell, they don't. But you could probably program a cell to be able to reinvigorate that polymerization process, to run along the fiber and repair it in places where it's damaged. Obviously these things get made during development, so it's totally physically feasible for this to occur. Maybe there's even a developmental state which would be sufficient to achieve this. I don't think anyone knows. But that would be the kind of state that one might have to engineer de novo, even if our genome doesn't necessarily encode for it explicitly.

Dwarkesh Patel 01:09:31

Interesting. Okay, what is Eroom's Law?

Jacob Kimmel 01:09:35

Eroom's Law is a funny portmanteau created by a friend of mine, Jack Scannell. He inverted the notion of Moore's Law, which is the doubling of compute density on silicon chips every few years. Moore's Law has graciously given us massive increases in compute performance over several decades. Eroom's Law is the inverse of that. In biopharma, what we're actually seeing is that there's a very consistent decrease in the number of new molecular entities, new medicines that we're able to invent, per billion dollars invested.

This trend actually starts way back in the 1950s and persists through many different technological transitions along the way. It seems to be an incredibly consistent feature of trying to make new medicines.

Dwarkesh Patel 01:10:13

In a weird way, Eroom's Law is actually very similar to the scaling laws you have in ML, where you have this very consistent logarithmic relationship. You throw in more inputs and you get consistently diminishing outputs. The difference, of course, is that this trend in ML has been used to raise exponentially more investment and to drive more hype towards AI. Whereas in biotech, modulo NewLimit's new round, it has driven down valuations, driven down excitement and energy.

With AI at least you can internalize the extra cost and the extra benefits because there's this general purpose model you're training. This year you spend $100 million training a model, next year $1 billion, the year after that, $10 billion. But it's one general purpose model, unlike, “We made money on this drug and now we're going to use that money to invest in 10 different drugs in 10 different bespoke ways.”

I was gearing up to ask you, what would a general purpose platform—where even if you had diminishing returns, at least you can have this less bespoke way of designing drugs—look like for biotech?

Jacob Kimmel 01:11:15

I'm going to slightly dodge your question first to maybe analyze something really interesting that you highlighted. You have these two phenomena: ML scaling and then scaling in terms of the cost for new drug discovery. Why is it that the patterns of investment have been so different? There are probably two key features that might explain this difference.

One is that the returns to the scaled output in the case of ML actually are expected to increase super exponentially. If you actually reach AGI, it's going to be a much larger value than just even a few logs back on the performance curve that people are following. Whereas in the life sciences thus far, each of those products we're generating further and further out on the Eroom's Law curve as time moves forward, haven't necessarily scaled in their potential revenue and their potential returns quite so much. You're seeing these increased costs not counterbalanced by increased ROI.

The other piece of it that you highlighted is that it’s unlike building a general model where potentially by making larger investments, you can be able to solve a broader addressable market, moving from solving very narrow tasks to eventually replacing large fractions of white collar intelligence. In biotech, when you're traditionally able to develop a medicine in a given indication—”I was able to treat disease X”—it doesn't necessarily engender you to be able to then treat “disease Y” more readily.

Typically where these biotech firms in general have been able to develop unique expertise is on making molecules to target particular genes, so “I'm really good at making a molecule that intervenes on gene X or gene Y.” It turns out that the ability to make those molecules more rapidly isn't actually reducing the largest risk in the process. This means that the ability to go from one or two outputs one year to then four the next is much more limited.

This brings us then to the question of what the general model would be in biology. I think it reduces down to how do you actually imbue those two properties that create the ML scaling law curve of hope and bring those over to biology so that you can take the Eroom's law curve and potentially give it the same sort of potential beneficial spin.

There are a few different versions of this you could imagine. But I'll address the first point. How do you get to a place where you're actually able to generate more revenue per medicine so that potentially the outputs you're generating are more valuable, even if each output might cost a bit more? Traditionally, when we've developed medicines, we go after fairly narrow indications, meaning diseases that fairly small numbers of people get. That's actually increased, in terms of the narrow scope of what medicines are addressing, as we've gone forward in time.

This is sort of an ironic situation where we've gone from addressing pretty broad categories of disease, like infectious disease, to narrower and narrower genetically-defined diseases that have small patient populations. Because these only affect a few people—if you think about the value function of a medicine as how many years of healthy life it gives how many people—if the “how many people” is pretty small, it just really bounds the amount of value you're able to generate.

You need to then be able to find medicines that treat most people. All of us will one day get sick and die. So arguably, the TAM for any really successful medicine could be everybody on planet Earth. We need to find a way to be able to route toward medicines that address these very large populations.

The second piece then is, how do we actually build models that enable us to take the success in one medicine we've developed and lead that to an increased probability of success on the next medicine? Traditionally, we haven't been able to do that. Maybe you're better at making an antibody for gene Y because you made one for gene X five years ago. But it turns out making an antibody isn't really the hard part of drug discovery.

Figuring out what to make an antibody to target is the hard thing about drug discovery. What gene do I intervene upon in order to actually treat a disease in a given patient? Most of the time, we just don't know. That's why even if a given drug firm becomes very good at making antibodies to gene X and they have a successful approval, when they then go to treat disease Y they don't necessarily know what gene to go after. Most of the risk is not in how to make an antibody to treat my particular target, it's in figuring out what to target in the first place.

Dwarkesh Patel 01:14:59

I'm not sure how to understand this claim that we know how to engage with the right hook, we just don't know what that hook is supposed to do in the body. I don't know if that's the way you describe it. Another claim that I've seen is that with small molecules we have this Goldilocks problem. They have to be small enough to percolate through the body and through cell walls, etc., but big enough to interfere with protein-protein interactions that transcription factors might have.

There it seems like getting the hook is the big problem.

Jacob Kimmel 01:15:32

In this particular case, if we bound ourselves to, "We must use small molecules as our modality," then there are lots of targets which are very difficult to drug.

There are many other modalities by which you can drug some of these genes. I would say–I don't have formal way of explaining this–if you were to write out a list of well-known targets that many, many folks would agree are the correct genes to go after and to try and inhibit or activate in order to treat a given set of diseases—and the only reason we don't have medicines is that we can't figure out a trick in order to be able to drug them—it's a fairly small list. It would probably fit on a single page. Whereas the number of possible indications that one could go after, and the number of possible genes that one could intervene upon especially when you consider their combinations, is astronomical.

The experiment you could run here is if you lock 10 really smart drug developers in a room. You tell them to write down some incredibly high-conviction target disease pairs where they're sure if they modulate this biology, these patients are going to benefit. All they need is some molecular hook, as you put it, in order to do this. It's a relatively short list. What you're not going to get is anything approximating the panoply of human pathologies that develop.

You can actually look for this. There are some existence proofs you can look for out in the universe. If the only problem was that we didn't have the ability to drug something using current therapeutics that we can put in humans, we should still be able to treat it in the best animal models of that disease because we can use things like transgenic systems. You can go in and you can engineer the genome of that animal. This gives you all sorts of superpowers that you don't have in patients, but allow you to, for instance, turn on arbitrarily complex groups of genes in arbitrarily specific or broad groups of cells in the organism, at any time you want, at any dose you want in the animal.

For the majority of pathologies, we just don't have many of those examples.

Dwarkesh Patel 01:17:13

What is the answer to what is the general purpose thing where every marginal discovery increases the odds you make the next discovery?

Jacob Kimmel 01:17:23

There are multiple ways one might approach this problem. The most common today, this is often what people are describing when they talk about a virtual cell. This is a very nebulous idea, sometimes numinous, if you'll let me describe it in that way as well. Concretely, what most people are trying to do is measure some number of molecules, or perceived emissions like the morphology of a cell, and then perturb it many times, turn some genes on, turn some genes off, and measure how that molecular morphological state changes.

The notion is that there's a lot of mutual information in biology. If I measure something, most commonly all the genes the cell is using at a given moment, which you can get by RNA sequencing, I get a decent enough picture of most of the other complexity going on. I can take a bunch of healthy cells and a bunch of cells that are in a diseased or aged state. I'm able then to compare those profiles and say, “Okay, my diseased cells use these genes, my healthy cells use these. Are there any interventions that I'm able to experimentally find in the lab that shift one toward the other?”

The hope would be that, because you're never going to be able to scan combinatorially all the possible groups of genes to make it concrete. There's something like 20,000 genes in the genome. You can then choose however many genes in your combination you want. It's not crazy to think of hundreds at a time. That's what transcription factors control. That's how development works. So the number of possible combinations is truly astronomical. You just can't test it all.

The hope would be that by doing some sparse sampling of those pairs—what your inputs are, here's what the cell looked like beforehand, here's the particular genes I perturbed—you have some measurement then of the state that the cell resulted in. Here's which genes went up, here's which went down. Once I've trained a model to predict from the perturbations to the output on the cell state, you can start to ask what would happen for some arbitrary combinations of genes.

Now in silico I can search all possible things that one might do and potentially discover targets that take my diseased cells back to something like healthy cells. So that's another version of what an all-encompassing model would look like where you actually have compounding returns in drug discovery.

Dwarkesh Patel 01:19:21

You basically described one of the models you guys are working on at NewLimit. You're training this model based on this data where you're taking the entire transcriptome and just labeling it based on how old that cell actually is.

If you've got all this data you're collecting on how different perturbations are having different phenotypic effects on a cell, why only record whether that effect correlates with more or less aging? Why can't you also label it with all the other effects that we might eventually care about and eventually get the full virtual cell? That's a more general purpose model, not just the one that predicts whether a cell looks old or not.

Jacob Kimmel 01:20:08

Absolutely, we actually do both today. We can train these models where the inputs are a notion of what that cell looked like at the starting place, here's what a generic old cell looked like, and then representations of the transcription factors themselves. We derive those from protein foundation models. They're language models trained on protein sequences. It turns out that gives you a really good base level understanding of biology. The model's starting from a pretty smart place.

Then you can predict a number of different targets from some learned embedding, the same way you could have multiple heads on a language model. One of those for us is actually just predicting every gene the cell is expressing. Can I just recapitulate the entire state and guess what effect these transcription factors will have on every given gene? You can think about that as an objective rather than a value judgment on the cell. I'm not asking whether or not I want this particular transcriptome. I'm just asking what it will look like.

Then we also have something more like value judgments. I believe that that transcriptome looks like a younger cell. I'm going to select on that and train ahead to predict it where I can denoise across genes and then select for younger cells. But you could do that for arbitrary numbers of additional heads. What are some other states you might want? Do I want to polarize T cells to a less inflammatory state in somebody with an autoimmune disease? Do I want to make liver cells more functional in a patient who's suffering from certain types of metabolic syndrome, maybe even orthogonal to the way that they age? Do I want to go in and change the way a neuron is functioning to a different state to treat a particular type of neurodegenerative disease? These are all questions you can ask. They're not the ones we're going after, but that is the more general, broader vision.

Dwarkesh Patel 01:21:36

This is so similar to, in LLMs, you first have imitation learning with pre-training that builds a general-purpose representation of the world. Then you do RL about a particular objective in math or coding or whatever that you care about. You are describing an extremely similar procedure where first you just learn to predict perturbations in genes to broad effects on the cell. That's the pre-training, just learning how cells work. Then there's another afterward layer of these value judgments of, “How would we have to perturb it to have effect X?” That actually seems very similar to “How do we get the base model to answer this math problem or answer this coding problem?”

I don't know if people usually put it this way, but it actually just seems extremely similar. That makes me more optimistic on this. LLMs work and RL works.

Jacob Kimmel 01:22:34

Yeah, they do. I think the conceptual analogy is very apt. We don't actually use RL at the moment, so I don't want to overstate the level of sophistication we've got. But I think the general problem reduces down in a similar way.

You can think about your earlier question of what does the general model look like that enables you to actually have compounding returns in drug discovery. You might have something like this base model, which as you said, just predicts this object function of, “How are these perturbations hitting these targets going to change which genes are turned on and off in this cell?” Then there's an entirely other task, which is, well, which genes do you want to turn on and off? What state do I want the cell to adopt?

Our lens on that is that across many different diseases people have, age is one of the strongest predictors of how they're going to progress, whether that disease arises. In many, many circumstances you have evidence in humans where you can say, “Ah, if I could make the cell younger, maybe that's not a perfect fix, but that's going to dramatically benefit not only patients who have a diagnosed disease, but it might actually help most of us stay healthier longer, even subclinically before anyone would formally say that we're sick.”

Now that's another more general function. The same way that in LLMs, you might have to create these particular RLHF environments, you need to have places where you can state a value function of the particular task that you're trying to optimize for. In drug discovery, you would then need to know, “Well, what are the cell states I want to engineer for?”

That's kind of the next generation of what a target might be. Beyond just which genes do I want to move up and down, and which gene perturbations do I put in, you then need to know what cell state am I engineering for? What do I want this T cell to do?

Dwarkesh Patel 01:24:01

You’ll have a bunch of labelers in Nigeria clicking different pictures of cells. Like, “Oh, this one looks young. This one looks old.”

Jacob Kimmel 01:24:06

“This one looks really great. I love that one.” Potentially. Potentially. It's more like developmental biologists locked in a room, as my friend Cole Trapnell would say.

Dwarkesh Patel 01:24:15

It seems like what you're describing seems quite similar to Perturb-seq. I don't know when it was done, what year was it?

Jacob Kimmel 01:24:22

There were three papers almost simultaneously in 2016.

Dwarkesh Patel 01:24:25

Okay, so almost a decade. We're still waiting, I guess, for the big breakthroughs it's supposed to cause. This is the same procedure, so why is this going to have an effect?

Jacob Kimmel 01:24:37

Why has this taken so long? Good question. The original procedure was created by a bunch of brilliant folks. There was a group in Ido Amit's lab at the Weizmann Institute, Aviv Regev's lab at the Broad, where Atray Dixit, a friend of mine, helped work on this, and then Jonathan Weissman's lab at UCSF, where Britt Adamson did a lot of the early work. They all constructed this idea where you can go in and you label a perturbation that you're delivering to a cell.

This is typically a transgenic perturbation, meaning you're integrating some new gene into the genome of a cell. That turns another gene on or off. They used CRISPR, but there's lots of ways to do it and the concept's pretty general. Then you attach on that new transgene, that new gene you put into the genome of the cell, some barcode that you can read out by DNA sequencing. Now when you rip the cells open, you're able to not only measure every gene they're using, but you also sequence these barcodes, and you know which genes you turned on and which are off. You can then start to ask questions like, “Well, I've turned on genes A, B, and C, what did it do to the rest of the cell?”

That's the general premise of the technology. It's useful to just set that up because it explains why this didn't all happen earlier. The actual readout, ripping the cells open and sequencing them, used to be pretty bad and it used to be really expensive. It's gotten much better over time. The metric people often think about here is like cost per cell to sequence. It used to be measured in dollars and now it's measured in cents, and down to the fractions of cents, because that cost curve has improved dramatically. The cost of sequencing has likewise come down. So even beyond the actual reagents necessary to rip the cell open and turn its mRNAs into DNAs that are ready for the sequencer, now the sequencer is cheaper.

The other piece is, actually getting these genes in and then figuring out which ones are there, it started out pretty bad. When we started with this technology, it was a beautiful proof of concept, but I don't think anyone would tell you it was 100% ready for prime time. When you sequenced a cell, only about 50% of the time could you even tell which perturbation you put in. Sometimes you just wouldn't detect the barcode and you'd have to throw the cell away. Or you detect the wrong barcode and now you've mislabeled your data point.

This might sound like a trivial sort of technical piece, but imagine you're running this experiment the old-fashioned way. You test different groups of genes in different test tubes on a bench. Now imagine you hired someone who every other tube labels it wrong. When you then collect data from your experiment, you basically have no idea what happened, because you've just randomized all your data labels. You wouldn't do much science. You wouldn't get very far that way.

A lot of those technologies have improved to the point where you had a number of processes which are pretty inefficient and you multiplied a lot of these things together and ended up with a very small outcome of successful cells you could actually sequence. They've all improved to the degree where now you can actually operate at scale.

Groups like ours have had to do a bunch of work in order to actually enable combinatorial perturbations, turning on more than just one gene at a time, which it turns out is much, much harder for the same reason we were just alluding to. Imagine you're having trouble figuring out which one gene you put in this cell and turned on or off. Now imagine you have to do that five times correctly in a row. Well, if you start out with the original performance where you could detect roughly 50% of them, then the fraction of cells that would be correctly labeled is like 1/2^n, where n is the number of genes you're trying to detect. Very quickly more of your data is mislabeled than is labeled.

There's lots of technical reasons like this that have gotten worked out over time. Only now are we really able to scale up where we're able to run experiments that are in the millions of cells in just a single day at, for instance, a small company like NewLimit. There was a point even just six or seven years ago where the companies that made these reagents were publishing the very first million-cell data set just as a proof of concept and only they could do it as the constructors of the technology. Now two scientists in our labs can generate that in an afternoon.

Dwarkesh Patel 01:28:02

If it actually is the case that this is very similar to the way LLM dynamics work, then once this technology is mature and you get the GPT-3 equivalent of the virtual cell, what you would expect to happen is you get many different companies, at least a couple, that are doing these cheap, Perturb-seq-like experiments and building their own virtual cells. Then they're leasing this out to other people who then have their own ideas about, "We want to see if we can come up with the labels for this particular thing we care about and test for that."

What it seems like is happening right now is, at least at NewLimit, you are like, "We know the end use case we're going after." It would be as if Cursor in 2018 was like, "We're going to build our own LLM from scratch so that we can enable our application," rather than some foundation model company being like, "We don't care what you use it for, we're going to build this."

Does that make sense? It seems like you're combining two different layers of the stack. Because nobody else is doing the other layer, you're just doing both of them. I don't know to what extent this analogy maps on.

Jacob Kimmel 01:29:18

To play with the analogy a bit, imagine that you think about NewLimit as an LLM company. If I'm going to put us in the shoes of Cursor, which oh I so wish, imagine we're trying to, in 2018, create Cursor Tab, but we're not trying to create a full LLM. I don't know enough about the underlying mechanics to know if that would have been feasible, but it's a much more feasible problem than trying to create the most recent Cursor agent or compete with modern Claude Code.

That's roughly the equivalent. The problem we're breaking off is a subset of the more general virtual cell problem. We're trying to predict, “What do groups of transcription factors do to the age of very specific types of cells?” We only work on a few cell types at NewLimit because those are some of the only cell types today with which we believe we can get really effective delivery of medicines. We think they're just more important because we can act on them today. If we solve the problem of what TFs to use, we can make a medicine pretty quickly.

In a way, we're carving out a region of this massive parameter space and saying, "If we can learn the distribution of effects even just in this small region, it's going to be really effective for us, and we can make really amazing products, unlike the world has ever seen." Over time, we can expand to the corpus of predicting every possible gene perturbation in every possible cell type.

I think that's maybe the way the analogy maps on, but it is true that we are vertically integrating here. We're generating our own data in a way that's proprietary. We think we have a much, much larger data set for this particular regime than the rest of the world combined. That enables us to build what we think are the best models.

In many cases, what we found is that unlike with LLMs, where a lot of the data that was necessary to build these was a common good—it was produced as a function of the internet and shared across everyone, it's pretty common across all the domains everyone wants to use it for—this biological data is still in its infancy. Imagine we're in the early 1980s and we are just now thinking about trying to create some of the first web pages. That's the era we're in.

We're going after and generating some of our own data in this very niche circumstance, building the very high-quality corpus, the Wikipedia that you might train your overly analogized- LLM on, and then building the first products based on that and then expanding from there. We think that's necessary because of where we are today. There isn't this Internet-like equivalent of data that everyone can go out and reap rewards from.

Dwarkesh Patel 01:31:32

Interesting. This is more a question about the broader pharma industry rather than just NewLimit. In the future, how are people going to make money? With the GLPs, we've got peptides from China that are just a gray market that people can easily consume. Presumably, with these future AI models, even if you have a patent on a molecule, finding an isomorphic molecule or an isomorphic treatment is relatively easy.

If you do come up with these crazy treatments and if pharma in general is able to come up with these crazy treatments, will they be able to make money?

Jacob Kimmel 01:32:07

The gray market piece, I'll put aside and say that’s IP enforcement at a geostrategic level that I'm not qualified to speak to. It comes down to IP enforcement effectively. For that gray market piece, another reason that the traditional pharmaceutical industry will still continue to reap the majority of rewards here is that most of the payment in the United States, which provides most of the revenue for drug discovery in the world, goes through a payment system that is not just direct-to-consumer. It goes through payers.

If you have the opportunity to either order a sketchy vial off of some website from some company in Shenzhen, or you can go through your doctor and get a prescription with a relatively low co-pay for Tirzepatide, the real thing, most patients will go for Tirzepatide. You and I probably live in a milieu of people who are much more comfortable with ordering the vials from Shenzhen than most people might be. I don't consider that to be a tremendous concern writ large.

The broader point is, if you have medicines with very long-term durability, how do you reimburse them? If the benefits are very long term and accrue in the out-years… A challenge we have in the US system is that the average person churns insurers every three to four years. That number fluctuates around, but that's the right order of magnitude. That means that if you had a medicine which dramatically reduced the cost of all other healthcare incidents, but it happened exactly five years after you got dosed with it, no insurer is technically economically incentivized to cover that.

I think there are a couple of models here that can make sense. One is something called pay-for-performance where, rather than reimbursing all of the cost of the drug upfront, you reimburse it over time. Say you get a medicine that just makes you generically healthier and you can measure the reduced rates of heart attack and reduced rates of obesity and various other things, and you get this one dose and it lasts for 10 years. Each year you would pay something like a tenth of the cost of the medicine contingent on the idea that it was actually still working for you and you had some way of measuring that.

That's a big challenge in this industry. How would you demonstrate that any one of these medicines is still working for the patient? In the few examples we have today, these are things like gene therapies where you can just measure the expression of the gene and it’s like, “Okay, the drug is still there.” It gets more complicated when you have some of these longer-term net benefits.

The idea would be that then each insurer is incentivized to just pay for the time of coverage that you're on their plan. We already have a framework for this post-Affordable Care Act in the US where pre-existing conditions no longer really exist. Patients are able to freely move between payers, and you could sort of treat the presence of one of these therapeutics lowering this patient's overall healthcare costs the same way we treat a pre-existing condition. This is something that the system is still overall figuring out. What I'm saying here is one hypothesis about what the future might look like, but there are alternative clever approaches people might think about for reimbursement.

I think over time we're going to move more toward a direct-to-consumer model for many of these medicines which preserve and promote health rather than just fixing disease. You're seeing what are really some of the most innovative examples of this right now from Lilly around the incretin mimetics, where they actually launched LillyDirect. For the first time, rather than going to a pharmacy, which interacts with a PBM, which interacts with your primary care physician… You can get a prescription from your doctor, go straight to Lilly, the source of the good stuff, and you're able to order high-quality drugs from them, and not involve some intermediary compounder in the middle that might not even make your molecules properly.

As these medicines develop that have actual consumer demand—because you feel it in your daily life and you're actually seeing a benefit from it, it's not just something that your physician is trying to get you to take—that model will start to dominate. That means that this payment over time for some of these long-term benefits might be able to be abstracted away from our current payer system where it churns every few years. A payment-over-time plan, the same way we finance other large purchases in life, seems very feasible.

Dwarkesh Patel 01:36:01

The reason I'm interested in this is that healthcare is already 20% of GDP. It's grown notable percentages in the last few years. This is a fraction that is quickly growing. The overwhelming majority of this is going towards administering treatments that have already been invented. That’s good, but nowhere near as good as spending this enormous sum of resources towards coming up with new treatments that in the future will improve the lives of people that will have these ailments.

One question is just how do we make it so that more… If we're going to spend 20% of GDP on healthcare, it should at least go towards coming up with new treatments rather than just paying nurses and doctors to keep administering stuff that kind of works now.

Two, if the cost of drugs, at least from the perspective of the payer, ends up being, you need a doctor to give you some scan before he can write you a prescription and then they need to administer it and they need to make sure that you're doing okay, etc… Even if for you to manufacture this therapy it might cost tens of dollars per patient, for the healthcare system overall, it might be tens of thousands of dollars per patient. I'm curious if you agree with those orders of magnitude.

Jacob Kimmel 01:37:29

I think that's correct. I think the stat is something like drugs are roughly 7% of healthcare spend. I could be a little bit wrong on that, but the OOM is right.

Dwarkesh Patel 01:37:36

Basically, even if we invent de-aging technology, or especially if we invent de-aging technology, how should we think about the way it will net out in the fraction of GDP that we have to spend on healthcare? Will that increase because everybody's lining up at the doctor's office to get a prescription and you gotta go into the clinic every week? Or will that decrease because the other downstream ailments from aging aren't coming about?

Jacob Kimmel 01:37:59

I think the latter is much more likely to be the case. Here are some quick heuristics. There are many reasons that healthcare costs so much in the US. One of them is something like Baumol's cost disease, which is very unrelated to pharmaceutical discoveries but is something that we will have to solve in the system. Part of it's the disintermediation of the actual customer and the actual provider. These are things that biotech probably isn't going to be able to solve as an industry alone. That's probably a larger economic problem.

But when you think about how this will affect the total amount of healthcare that will need to be delivered. If you have more of these medicines for everyone, medicines that keep you healthier longer rather than medicines that only fix a problem once you're already very sick, I think you actually avoid a lot of the types of administration costs. It’s not just administration like admins at hospitals, but the cost of administering existing medicines and therapies to you. That’s going down.

One stat on why I think that's true. Something like a third of all Medicare costs are spent in the final year of life, which is shocking when you realize that the average person on Medicare is probably a decade-plus covered by it. There's an incredible concentration of the actual expenses once someone is already terribly sick. Helping prevent you from ever having to access the intensive healthcare system, something like an inpatient hospital visit, if you can prevent even just a couple of those visits over a long period of someone's life with a medicine like an incretin mimetic, like a reprogramming medicine that keeps your liver, your immune system younger, on net that actually starts to drive healthcare spend down because you're shifting some of that burden from the administration system to the pharmaceutical system.

The pharmaceutical system is the only piece of healthcare where technology has made us more efficient. As drugs go generic, the cost of administering a given unit of healthcare is going down.

The grand social contract is that they eventually go generic. That's the way our current IP system works. So if you were to get the question of, “When would you like to be born as a patient?” you always want to be born as close to today as possible. Because for a given unit in terms of pharmaceuticals, for a given dollar unit of expense, you can access more pharmaceutical technology today than has ever been possible in history, even as healthcare costs everywhere else in the system have shot up. Pharmaceuticals are the one place where, because of the mechanism of things going generic and the fact that our old medicines continue to work and persist over time, you're able to get more benefit per dollar.

Dwarkesh Patel 01:40:21

Okay, final question. Pharma is spending billions of dollars per new drug it comes up with. Surely they have noticed that the lack of some general platform or some general model has made it more and more expensive and difficult to come up with new drugs. You say Perturb-seq has existed since 2016. As far as you can tell, you have the most amount of that kind of data which would feed into a general-purpose model.

What is the traditional pharma industry on the other coast up to? If I went to the head of R&D at Eli Lilly or Pfizer or something, do they think that this is like they have some different idea of the platform that needs to be built or they're like, “No, we're all in on the bespoke game, bespoke for each drug?”

Jacob Kimmel 01:41:06

I'll just correct one thing to make sure I'm not overstating. We have way more data for the limited subproblem we're tackling, which is overexpressing TFs in combinations. We have way more data than anyone, full stop, there. But even more specifically, I feel very, very confident we have more data than anyone looking at trying to reprogram a cell's age. That's where we're way larger than the rest of the world.

When we think about just general single-cell perturbation data of various flavors, there are other groups which have very large data sets as well. We're still differentiated because we do everything in human cells with the right number of chromosomes, whereas it's very common to do things in cancer cell lines which have 200 chromosomes. Is that human? I don't know. Depends on how you actually quantify these things.

So then, if you’re going to go ask the leaders of some of the traditional pharmaceutical firms, "Are you trying to build a general model?" I think some of them have in-house AI innovation teams that are working on this. There are really smart people there. But as a general trend, you can think about some of the modern pharmas a bit like venture capital firms. They've over time externalized a lot of their R&D. They often have divisions of external innovation, which you can think of as the corp dev version of venture capital. They work with the biotech ecosystem to have a number of smaller, nimble firms explore really pioneer ideas, the types of things we're working on, and then eventually partner with them once they have assets that are later downstream.

The industry has sort of bifurcated where smaller biotechs like ours take on most of the early discovery. I'm going to get it a little bit wrong from memory, but it's something like 70% of molecules approved in a given year come from originally small biotechs rather than large pharmas, even though you look at the actual dollars of R&D spend on the balance sheet and it's largely in big pharma.

Dwarkesh Patel 01:42:47

Another level of disintermediation.

Jacob Kimmel 01:42:52

Part of the reason for that difference in cost is they're running most of the trials. Most people partner with pharma to run trials where a lot of the costs are incurred. It's not just that all large pharmas are horribly inefficient or anything like that. Some of them would tell you, “These ideas are really exciting. We have an external innovation department, if we don't have one internally, or we're collaborating with a startup that's doing something similar.”

You can think of the market structure like you have a bunch of biotechs, which are the startups in your ecosystem, and then they're working with something like an oligopsony of pharmas. It's a limited number of buyers for this particular type of product, which is a therapeutic asset that is ready for a phase one, phase two trial. There's a very liquid market for the phase one, phase two assets, and that's the point at which these partnerships can come to fruition. That's what a lot of those leaders would say.

By contrast, for instance, Roche bought Genentech back in 2013. R&D is currently run by Aviv Regev, one of the scientists I admire most in the world, who's like a thousand times smarter than me. She's one of the people who invented this technology and has a big group doing this sort of work there. So it's not like every pharma takes that view, but that's a general trend.

Dwarkesh Patel 01:43:56

Full disclosure, I am a small angel investor in NewLimit now, but that did not influence the decision to have Jacob on. This is super fascinating. Thanks so much for coming on the podcast.

Jacob Kimmel 01:44:08

Awesome. Thanks, Dwarkesh.

How will we feed the 100s of GWs of extra energy demand that AI will create over the next decade?

On this episode, Casey Handmer (Caltech PhD, former NASA JPL, founder & CEO of Terraform Industries) walks me through how we can pull it off, and why he thinks a major part of this energy singularity will be powered by solar. His views are contrarian, but he came armed to defend them.

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Lighthouse helps frontier technology companies like Cursor and Physical Intelligence navigate the U.S. immigration system and hire top talent from around the world. Lighthouse handles everything for you, maximizing the probability of visa approval while minimizing the work you have to do. Learn more at lighthousehq.com/employers

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – Why doesn’t China win by default?

(00:08:28) – Why hyperscalers choose natural gas over solar

(00:18:01) – Solar's astonishing learning rates

(00:27:02) – How to build 50,000 acre solar-powered data centers

(00:40:24) – Environmental regulations blocking clean energy

(00:44:04) – Batteries replacing the grid

(00:49:14) – GDP is broken, AGI's true value must be measured in total energy use

(00:58:45) – Silicon wafers in space with one mind each

Dwarkesh Patel 00:00:00

Today I'm interviewing Casey Handmer. Casey has worked on a bunch of cool things. Caltech PhD on some gravitational wave black hole gimmick stuff, then Hyperloop, then the Jet Propulsion Laboratory at NASA. Now he is founder and CEO of Terraform Industries. Casey, welcome.

Casey Handmer 00:00:16

Thank you. It's great to be here finally.

Dwarkesh Patel 00:00:19

Big picture question I'm interested in. To the extent that AI just ends up being this big industrial race - who can build the most solar panels? Who can build the most batteries? Who can build the most GPUs and transmission lines and transformers? This is not what the US is known for, at least in recent decades. This is exactly what China is known for. They have 20x the amount of yearly solar manufacturing the US has. Obviously we have export controls right now, but over time SMIC will catch up to TSMC's leading edge.

What is the story of how the United States wins this? Why does China just not win by default?

Casey Handmer 00:01:00

Do you think that China is better at capital allocation than the United States? Do you think the Chinese business environment is better for business than in the United States?

Dwarkesh Patel 00:01:06

I feel you can make these first principles arguments about these other industries where they're killing it, but it doesn't seem to have hampered BYD or CATL.

Casey Handmer 00:01:11

People say they're so much better at building high-speed trains than the United States. I would never hold up a flag saying, “I'm really good at building high-speed trains.” That is just a sign that you're really bad at capital allocation. Why would you devote, in 2025, so much industrial effort and money…

Dwarkesh Patel 00:01:26

They're devoting a lot to solar overcapacity, which in your opinion is the key to future industrial growth.

Casey Handmer 00:01:33

I think they might be accidentally correct.

Dwarkesh Patel 00:01:35

They called the most important thing correct, right? Which should count for something.

Casey Handmer 00:01:37

Well, they're in a similar situation to Europe, but unlike the United States. The United States is the luckiest goddamn country on earth because it's surrounded on two sides by oceans and on the other two sides by friendly allies.

China's surrounded by 15 countries who are mostly hostile to it, with no good mountain ranges or rivers or anything to really separate them. They get almost all their oil from the Middle East. From countries that they don't control, don't have strong diplomatic relationships with, on fleets of oil tankers that they can't defend because their navy doesn't have the ability to operate effectively in the Indian Ocean.

Dwarkesh Patel 00:02:11

But you're working on this, right? If you get synthetic fuels working at Terraform. Doesn't that asymmetrically help China? Which might be fine.

Casey Handmer 00:02:17

It does. It absolutely asymmetrically helps China. We're not currently working with China, we don't plan to, but the physics is very obvious. Synthetic fuels have been around for 100 years. There are projects in China right now working on synthetic fuels. It would not surprise me if they were thinking pretty seriously about this.

Dwarkesh Patel 00:02:30

Just to spell out for the audience, China has all this electricity production. And the bottleneck is that only a third of final energy use in a modern economy comes from electricity. The rest, you need gas and whatever to transport things...

Casey Handmer 00:02:44

Or coal. They use a lot of coal in China.

Dwarkesh Patel 00:02:45

Right. What Casey is inventing is a technology to turn that electricity, which only can supply a third of end-uses right now, into synthetic fuels which can supply 100% of the electricity your civilization needs. China's energy advantage then becomes overwhelming.

Casey Handmer 00:03:04

This technology levels the playing field. It levels the playing field a lot. But at the end of the day, China still contains the poorest Chinese people anywhere on earth. Never underestimate the capacity for an autocratic dictatorship to shoot itself in the foot.

Dwarkesh Patel 00:03:17

I don't know. I agree that they've obviously made bad decisions, but even if you have the poorest Chinese people anywhere in the world, they can still be quite rich. Like Singapore is richer, whatever.

Also, there are parts of China which actually contain quite rich Chinese people. You have to compare not all of China against the US, but Shanghai and Guangdong against the United States. You can have a part of China that is as big as America and as wealthy as America and as innovative as America.

Casey Handmer 00:03:44

Like the Indian middle class is larger than the US middle class.

Dwarkesh Patel 00:03:47

But also it's nowhere near as wealthy. Whereas there are parts of China which are humongous, which are actually as wealthy as the United States, and in many cases as innovative, etc.

Casey Handmer 00:03:56

Yeah. I'm saying don't underestimate it, but at the same time, we want to find the truth here. The truth is we should not count the United States out of the battle and just give up. We're very much still in the race now, provided we don't take extra effort to shoot ourselves in the foot.

Dwarkesh Patel 00:04:09

Right now we are export controlling chips for the purpose of keeping our AI lead and we recognize this is a key input in our ability to compete in AI. So we are going to export control China's ability to have these chips.

Energy is also a key input in this AI race, and if China wanted to do the converse of what we're doing to them with these cheap imports, what they would do to us is to export control solar and batteries.

Casey Handmer 00:04:40

It would be asymmetrical, it would hurt them worse than us.

Dwarkesh Patel 00:04:43

If they did tariffs?

Casey Handmer 00:04:44

China obviously depends upon the US export market for its economic dynamism. It's going to hurt both parties to sever the link. But if you sever the link completely, China's ability to make advanced chips right now is basically not there, whereas the United States can make them. The United States’ ability to make solar arrays is embryonic, but it's actually not that far behind China's. It's maybe five years behind.

Dwarkesh Patel 00:05:09

If we decided we want to produce 100 gigawatts of solar capacity every single year…

Casey Handmer 00:05:13

We’re already on track to do that.

Dwarkesh Patel 00:05:15

Is it going to be as cheap as it is to do in China?

Casey Handmer 00:05:19

My views on this are somewhat different from the mainstream, which is great because this is a podcast. The mainstream view would say China has cheaper labor, which is no longer true, because they compare to Mexico. And it's got lower environmental regulations, which is true. And that it is more business friendly, which is absolutely crazy. There's no way you could justify that your company having to have an inspector from the CCP on its board who harasses you about Xi Jinping every day helps you do your business. And also the rule of law is not great. So you're constantly having to pay bribes to people in order to stay in business.

The idea that the United States cannot compete against that with mostly or fully automated solar panel manufacturing in the United States—which has cheaper natural gas by far, abundant oil, abundant human resources, great financial capacity, world leading automation, etc.—is crazy. We could literally copy paste solar manufacturing factories.

Dwarkesh Patel 00:06:05

How much additional solar power capacity do you think we could be putting on, that's manufactured in the US, by 2028?

Casey Handmer 00:06:13

This is a good question. When Russia invaded Ukraine, I thought, finally the Europeans will see sense and they'll pull the trigger on, “We need to localize production of solar panels from dirt to the finished module,” which is roughly a four-stage process. They didn't. They're still paying Russia a billion dollars a day for the privilege of being invaded.

But at the time I thought they could probably do that in about two years. I think the United States could probably do that in two years or less if you started today. It's currently 11 o'clock. So we're going to start cutting checks by noon.

You could ramp up pretty quickly. A lot of technology already exists here. It's not like it has to be invented from scratch. It's mostly a case of putting in a phone call to all the different manufacturers here, in Germany, and so on and saying, “We need you to 10x the size of your factory, starting today, blank check, go.”

Dwarkesh Patel 00:07:00

A lot of your predictions seem to be not predictions, but more like, “If we had World War II-levels of motivation, if we had Manhattan Project-level intensity around doing a specific thing, how fast could we do it? Like if Elon was running the government, how fast could it happen?”

Casey Handmer 00:07:18

He was for a brief period.

Dwarkesh Patel 00:07:20

Maybe then we should put it like, “If Elon ran the government like he ran SpaceX.” As opposed to the question of “What is actually practically likely to happen, given that we are not treating it with World War II-level intensity.”

Casey Handmer 00:07:30

If you look at xAI, which Elon is involved in obviously, what are they actually focused on right now? They're focused on the chips because they understand the key bottleneck is the chips, not the solar power. Even if Trump puts in a 200% tariff on Chinese solar and we're not able to bypass it via Vietnam or something, it's still a bargain. It doesn't matter. If you need solar to run your data center, it doesn't hurt in terms of the overall cost picture. It doesn't matter at all. What matters is having the chips at competitive capabilities per chip, and enough of them installed in your PCBs, in your data centers, hooked up to your liquid cooling, ready to go.

That's actually something that Elon and his companies are great at. It's figuring out this mass production, semi-automated mass production. They've got this facility in Texas which is making the Starlink receivers, completely automated.

At what point does, "Oh, we don't have a solar panel factory," become on the critical path? I very much doubt it's ever going to be on the critical path. There's dozens and dozens of manufacturers of solar panels worldwide that are all competing against each other.

Dwarkesh Patel 00:08:28

So you're a big solar bull.

Casey Handmer 00:08:30

Yeah.

Dwarkesh Patel 00:08:31

Right now the hyperscalers are making decisions about the data centers that they're building. They're going to be 1-2 gigawatts, 5 gigawatts in Meta's case. They’re making decisions about how they're going to be actually powered. The people with actual money on the line are choosing natural gas. It's not like they can't see the learning rate. They're building things which will be online in 2028 or 2030. Why are they wrong and you're right?

Casey Handmer 00:08:56

It's their job. They probably know more about it than I do. But in all seriousness, if you're like xAI right now trying to build the Colossus data center in downtown Memphis, you want to get it done super fast.

“What are all the different things we need? What are the factors of production to build this? We need a building. We don't have time to build a building so we'll buy a building. Okay. We'll adapt it. We need power, we need thermal cooling.” That stuff you can deliver on a truck so that's what they did. You need access to gas. They had access to gas there. They could tap into a local gas line.

If you can tap into a gas line, generally speaking, you can get enough power. The energy transmission capacity of your regular gas delivery pipelines is way, way higher than electricity overhead lines, and it's easy to upgrade.

So if you're in this situation right now, you say, “Are we constrained by our ability to go and rent gas turbines?” No, they're not, because there was enough available once, maybe twice. But at a certain point, you realize as you grow, you start to touch all these additional constraints. Some of those constraints include gas availability. So there's a lot of chat about doing this in Pennsylvania where there's quite a lot of stranded gas, and in parts of Texas. But at the same time, the United States is gearing up in its ability to export natural gas overseas, so the price will not be infinitely low forever.

You start to run into constraints around turbine manufacturing rate, around transformer production rate, around grid capacity, and also running into problems where the AIs and the humans who depend on legacy electricity production and delivery utilities are competing with each other. We just saw this recent forward auction in PJM result in very high, unsustainably high prices for consumers who depend on cheap electricity to heat and cool their houses and have general prosperity.

If you look far enough in the future, you can just turn up the dial arbitrarily high. You can say we're going to put in a gigawatt a year. Well, we can meet that constraint with gas turbines. We're not going to run out of natural gas at 1 gigawatt per year indefinitely. What if we're doing 5 gigawatts per year? What if we're doing 50 gigawatts per year? What if we're doing 100 gigawatts per year? You can just break the situation.

Not to reach prematurely for analogies, but Henry Kaiser set up the shipyard in Richmond, just down the road here in San Francisco near Berkeley. He was initially making ships for the British and by the end of the war he had four separate shipyards operating in parallel, to the point where he was bottlenecked on his supply of steel. Steel was rare enough in the war, because everyone was using it for different things, that Kaiser Industries went off and built not only a steel mill, but also a steel mine. They went and started digging rocks out of the ground to turn into ships.

That's the same sort of situation you have here where you have these massive industrial verticals. Here I'm quite bullish on xAI in particular because the Elon cinematic universe has just done so much industrial stuff compared to the Googles and Metas of this world. They can reach all the way through down into primary material supply if they need to.

Dwarkesh Patel 00:11:50

The reason that these current plans are being done based on natural gas is that this is the sort of…

Casey Handmer 00:11:56

PJM has all kinds of different sources of power. They have nuclear as well, they have gas, they have coal, all kinds of stuff. This price here is probably driven more by the delivery cost growth than by the generation cost growth, if that makes sense.

When you pay your utility bill, the cost is sometimes broken down into a delivery cost and a generation cost, sometimes importation costs and other things. The delivery cost is what it costs the utility to build and maintain all the power lines that connect all the houses to all the power plants in some gigantic area divided by your marginal usage, with all kinds of other complicated rules designed to make it fairer.

The problem that we see—and the reason that PG&E here in California, for example, is perpetually on the brink of bankruptcy—is that even though the cost of an additional solar panel or additional wind turbine or additional gas turbine or whatever is relatively cheap, getting that power to your house is really expensive.

Why? Because you've got generally unionized labor that has to build and maintain power lines in areas that already have built up infrastructure. You have multiple collisions, whether this is a power pole on your own street or building a new transmission line which requires you to eminent domain land. So you're in court for years and years and years spending public money litigating against other people who are also spending public money to litigate against you on behalf of other interest groups and so on and so forth. Then you've got wildfires. It's just the poster child for Baumol cost disease.

One of the reasons that we're going to see large-scale pruning of these grids is that we just can't afford under our current regulatory regime to maintain.

Dwarkesh Patel 00:13:34

When you say pruning, will everything just go off-grid?

Casey Handmer 00:13:37

It's fairly clear to me that for really large captive loads, like AI data centers or aluminum refineries or whatever, you're going to have to build your own power plant for them, which is how it used to work. If you had an aluminum plant back in the day, you would be building your own power plant for it as well.

Dwarkesh Patel 00:13:54

It seems inefficient to have redundant power plants at every single industrial site.

Casey Handmer 00:13:59

Let me paint a grand vision for you. It would seem inefficient, but if you are sensitive to the cost of power expressed in supply elasticity or something like that, you just have to do it. There's no two ways about it. Is it inefficient for the xAI Colossus data center to have its own captive power plant, which it does on the backs of a bunch of trucks in the parking lot? No, it's not inefficient. It's the cheapest way for them to get power.

Dwarkesh Patel 00:14:26

Okay, AI might be a special case. But big picture question. Across different kinds of ISOs, from Texas to Pennsylvania to whatever, people are building data centers which will not be online for many years. They're choosing natural gas. What's going on?

Casey Handmer 00:14:41

We haven't completely exhausted the supply of turbines relative to GPUs.

Dwarkesh Patel 00:14:47

Do you have some estimate of when we'll run out of them? Because we can also make more.

Casey Handmer 00:14:53

Everything before about 2030 is spoken for at this point. Yeah, you could make more turbines. The funny thing is that it's actually relatively expensive to spool up additional production of these turbines.

Here's one thing you have to grapple with sooner or later. Conventional power generation is a steam engine. You have some kind of chemical that you find inside the earth that is out of chemical equilibrium with the atmosphere and you burn it and it makes heat. It could be coal, could be gas, oil.

Dwarkesh Patel 00:15:20

You're giving me the true birds and bees here.

Casey Handmer 00:15:21

Yeah, exactly. And it makes heat and you boil water. The water goes through some kind of mechanical contrivance that creates motion. That motion twists a magnet and generates an electrical field which then pushes electrons down wires, which then push electrons through a series of gates that then approximate thinking.

It's kind of complicated. But the key step in this is converting heat into electricity in the most efficient way. The most common way is the same for a nuclear plant or a gas plant, combined cycle plant or a coal plant or whatever. It’s what's called a Brayton cycle. The jet engine on an aircraft is a Brayton cycle as well. Anytime you have a Brayton cycle with a bunch of Inconel spinning at high speed, it's just going to cost you a bunch of money.

Dwarkesh Patel 00:16:04

Because it's inherently inefficient or what?

Casey Handmer 00:16:06

It's just inherently expensive to build.

Dwarkesh Patel 00:16:08

Okay. What is the cost of… GE makes these 100 megawatt gas turbines, right?

Casey Handmer 00:16:15

I don't actually know what the retail price is. I would suspect that if their price is flexible, it would have gone up a lot. But if I recall correctly, $35 a megawatt hour is just the flag-four cost for..

Dwarkesh Patel 00:16:25

How much, sorry?

Casey Handmer 00:16:26

$35 a megawatt hour just for the Brayton cycle. We're not talking about the fuel, we're not talking about the heat exchanges, we're not talking about the cooling ponds or anything like that. Just the amortized cost of the high-speed, high-temperature spinning components is $35 a megawatt hour.

Dwarkesh Patel 00:16:43

Do you think the hyperscalers are being irrational, or do they have some reason?

Casey Handmer 00:16:47

To be clear, they don't care about the cost of power. This is very counterintuitive. For Grandma Kettle in Pennsylvania, she's very sensitive to electricity costs. We don't really want her to suffer in her retirement from unaffordable electricity costs and having to sit there shivering. That's not the image that we want.

At the same time, what is the economic value to you of using Claude or Grok or whatever you use on a monthly basis?

Dwarkesh Patel 00:17:11

A lot.

Casey Handmer 00:17:11

It's obviously much more than the subscription, but is it 10 times more than the subscription maybe?

Dwarkesh Patel 00:17:15

Yeah, easily.

Casey Handmer 00:17:16

Let's say the subscription is on the order of $10. The value is on the order of $100. No, it’s probably more like $100 and $1,000. How much does it cost xAI or Anthropic or whatever to serve your usage?

Dwarkesh Patel 00:17:27

The marginal variable cost of serving it, in electricity, is less than 10% of the actual cost of…

Casey Handmer 00:17:32

Well, their cost of serving it is maybe a buck per million tokens or something like that. The cost of electricity is about 10% of that. So 10 cents of electricity is generating $1,000 worth of economic value.

It's very obvious that Anthropic could be like, "Our electricity cost basis has increased by a factor of 100.” Now instead of paying 10 cents on your bill, on your $100 bill for power you're paying $10. So we're putting your subscription up to $110 for an electricity capacity charge. Then they could go out and buy turbines for prices that would make your eyes water.

Dwarkesh Patel 00:18:01

Okay, so then why are we going to get the solar future? In 2032, we're going to have hundreds of gigawatts of extra demand for data centers and at that point, most of it is coming from solar? Why is that?

Casey Handmer 00:18:15

There aren't enough turbines being manufactured.

Dwarkesh Patel 00:18:18

But also, I think in the early 2000s… We can probably overlay the graph of how many turbines were being manufactured. Right now, we're at historical—

Casey Handmer 00:18:27

They've ramped up basically to the early 2000s rate again.

Dwarkesh Patel 00:18:30

But I don’t know, you have to make more solar panels as well, right? There will be supply elasticities for both solar and natural gas. Is there some reason to think that it's worse for the supply chain involved in having a natural gas-powered data center than a solar one?

Casey Handmer 00:18:43

Yeah I do. The learning rate for natural gas is nowhere near as steep as solar. It just tells you that it's easy to make solar panels, much easier to make solar panels. There are very few manufactured products which are easier to make. The Wright's Law coefficient is 43%. So every time we double cumulative production, we get a 43% reduction in cost.

Dwarkesh Patel 00:19:00

What is the basis of that? Why are we finding 43% worth of things that can be made cheaper or more efficient every single year?

Casey Handmer 00:19:08

Roughly speaking, there's 10,000 manufacturing process engineers working on this full-time.

Dwarkesh Patel 00:19:14

That could be true of any process, but no other process sees the kinds of learning rates that solar is seeing.

Casey Handmer 00:19:17

That’s not strictly true. In order to sustain this over a long period of time, you obviously need to have demand elasticity that exceeds your learning rate. Otherwise, you would, after a couple of OOMs, saturate your market at the current price and you'd have no additional growth. But in this case, roughly every two years we're doubling production. Every 2-2.5 years, we’re doubling production and the price is coming down by a factor of ~40%. So it’s roughly 15-20% per year. Then just as a result of that price reduction, demand skyrockets by probably six times more than that additional marginal production capacity increase.

This is one point where I'll say the so-called pros are definitely wrong. Conventional wisdom is that solar demand is going to saturate this week. It's going to saturate. We've got a graph here somewhere that's like, "This year is it, it's never going to grow anymore." Instead, it's just blasting out the top of the graph. This conventional wisdom is wrong. Not only are solar adoption, production, and price decreases continuing, they're accelerating. And the rate at which they're accelerating is still accelerating.

Dwarkesh Patel 00:20:18

The rate at which it's accelerating is accelerating?

Casey Handmer 00:20:21

Yes.

Dwarkesh Patel 00:20:22

As measured in the total fraction of energy that’s coming from solar?

Casey Handmer 00:20:26

In the sense that its fitness for the markets that it is being produced for is increasing over time. So it's still extremely early. We’re still in the Apple II computer era of solar.

Dwarkesh Patel 00:20:39

Backing up, if the story is that the reason solar is getting cheaper is because there's a lot of demand for more solar, and that demand can sustain economies of scale or whatever is going on…

Casey Handmer 00:20:50

Yes. I'm going to go on a limb here and agree with Elon Musk on this.

Dwarkesh Patel 00:20:53

Then shouldn't that also be true of gas turbines and transformers and power stations and whatever else that's required for the non-solar future? We're expecting AI to drive up demand for power regardless of the source. To the extent the story for solar becoming cheaper over time is just that demand will go up and that will drive efficiencies, why isn't that true for…?

Casey Handmer 00:21:17

Let's say you're a bank, and you're trying to decide whether to lend GE a bunch of money to expand production of their gas turbines. You can write them the check today. They'll start scaling up their factories. They'll start to see the benefits of that in three or four or five years.

You don't know if the AI bubble will have burst by then. You don't know if China will have invaded Taiwan by then. You don't know if Siemens or Philips or someone will have outcompeted you. You don't know if GE's major looming structural problems will cause it to be unable to compete, as it has in the past. In order to make that money back, you also have to then operate that plant at that capacity for 20 years.

If I was looking at the same charts as they're looking at right now, I'd say, “What are the odds that in 25 years' time we can produce gas turbines at a price that is relevant in a world where solar is already at its current price and batteries are at the price where they're already?” You cannot win.

Dwarkesh Patel 00:22:10

I feel like there was actually a similar discussion a year back when AI people were like, "No, AI is real. This is going to happen." Then SK Hynix, Samsung, etc., were like, "We're not ramping up HBM production because HBM is used largely for AI workloads, and if this demand doesn't continue, then our additional manufacturing capacity for HBM will not have been worth it." Then there was another bottleneck with CoWoS.

Casey Handmer 00:22:34

What happened after that? Did they end up indeed ramping up their production?

Dwarkesh Patel 00:22:38

I think so.

Casey Handmer 00:22:39

Well, so when someone says, "We can't do it, we won't do it, no way, no how," what they're saying is, "Write me a check." And they did. Now Samsung's coming on board in the States to build AI6 with xAI, I think. So they all got there in the end.

Dwarkesh Patel 00:22:52

Maybe it’s worth going into the numbers. Right now, 43% of US data center power consumption is from natural gas. Basically, you think asymptotically that it will be 100% solar if you go to 2040?

Casey Handmer 00:23:05

Yeah. Obviously legacy production, coal and stuff, is going to retire over time. If a gas plant is still making money, people will keep operating it. But at a certain point… It is the case right now that operating a coal plant costs more than building a new solar plant. So it's just cheaper to turn off. Also, capacity is going to increase a lot so that helps to dilute the existing production.

Dwarkesh Patel 00:23:28

And the amount of use is going to increase a bunch. The amount of data center use of energy will just be exponentially higher. So the new stock matters a lot as compared to the existing stock.

Anyway, I want to know in 2027, what fraction is natural gas? In 2030, what fraction is natural gas versus solar?

Casey Handmer 00:23:49

For new load?

Dwarkesh Patel 00:23:50

Let's say new load. For new load, 2035, etc. If eventually you're right that we'll pave the earth in solar panels to sustain quadrillions of AI souls, what is the pace of that?

Casey Handmer 00:24:00

The question to ask is, what is the major constraint on that ramp up? Then everything else will just draft in behind. I suspect that the hardest thing to make will always be the silicon, like the GPUs. So the question is really, “How quickly does TSMC ramp up its production of GPUs?” That's a question for you, not for me.

Dwarkesh Patel 00:24:22

I'll use some numbers that AI 2027 used for their compute forecast. Even if you don't buy their singularity thing, I think they did a reasonably good job with crunching the numbers on their compute forecast. I think they said there's on the order of 10 million H100 equivalents in the world today. I think they said by 2028 there'd be 100 million, so basically 10x more H100 equivalents in the world.

Casey Handmer 00:24:44

About a kilowatt each, something like that. Okay, so that's like 100 gigawatts. That sounds roughly right. You're not the first person to give me a call and ask me about this. I'll put it that way. I'm not going to name names. Pretty much all the names you've heard of have given me a call and said, "We know that you're a minority voice on the paper that came out recently with Scale Microgrids talking about how you could do 90% solar, 10% gas."

I said, "You can go all the way 100% solar." I wrote a blog post about it. So they always call me up and say, "What about this?" They're all talking like 5 gigawatts in the next few years. That's just like 90+% solar for just those. So within a few years, we'll probably see that the majority of new data centers that are going in will be mostly solar.

Dwarkesh Patel 00:25:23

Within how long?

Casey Handmer 00:25:24

Let's say by 2027, the majority of new data centers going in at that point would be mostly solar.

Dwarkesh Patel 00:25:29

Going in as in…?

Casey Handmer 00:25:30

Groundbreaking at that point.

Dwarkesh Patel 00:25:32

But if you're groundbreaking in 2027, you're probably planning it now, right?

Casey Handmer 00:25:35

That's why they're calling me. My consulting fees are extremely affordable. But I don't have deep visibility—because I'm not in the same room with the Meta people—as to when we're going to hit the wall on transformers and when we're going to hit the wall on just how much municipal peak load we can shave off, which is the latest thing that's been doing the rounds.

It turns out there's a handful of places in the United States—and by handful, I mean literally a handful—where there might have used to be an aluminum smelter. There's a bunch of latent capacity in the grid. And there's also a bunch of generators on the grid that are notionally turned down. They operate at, say, 40-50% capacity factor, but they max out at about 80% capacity factor because you've got to bring them down for maintenance pretty often, especially if they're old.

So they're saying, "Well, you know, we could pay you just to operate this old coal plant or something at higher capacity. It'll go down this power line to this place where the smelter used to be. We'll set up there, and then we promise to curtail when you need the power."

That basically means they just have a massive captive battery plant as well. Which is fine, you just buy that and it arrives on a truck. The major advantage to doing that over the pure solar play is that the power is already there, so there's no risk there. And you don't need a massive amount of land. The problem with the solar approach is that there's no two ways about it. It's a farming operation. You need a huge amount of land.

The total amount of land that you're using, less than 1% is under batteries, under roads, under data center structures, et cetera, etc. It's mostly solar.

Dwarkesh Patel 00:27:02

Let's get into what this looks like. If you've got a 5-gigawatt plant you want to build, break down the numbers for me in how much land in terms of solar you need to farm this out.

I was talking to somebody in this space and they said, "Obviously, the cost of energy for these data centers is a small fraction of the total cost. Most of the cost is going towards chips. So then the issue is just, can you make the energy available?" They were saying that even though solar panels themselves you can acquire, the issue is getting that much contiguous land and getting the permitting to interconnect it. That’s apparently a big hassle.

Casey Handmer 00:27:46

It's kind of a nightmare.

Dwarkesh Patel 00:27:48

So they're like, "Well, at that point, is it actually easier than just getting on the grid or…?"

Anyway, if you need tens of thousands of acres of solar, where can you do that?

Casey Handmer 00:27:59

Basically in Texas. There's this very popular misconception that there's not enough land to do solar. This is garbage. If you've ever flown in an aircraft in the United States and you've ever looked out the window, you'd be like, "Oh, wow, look, there's a lot of land you could put solar on." Especially west of like 110°.

Dwarkesh Patel 00:28:16

Does it need to be flat, or no?

Casey Handmer 00:28:17

No. Doesn't matter. Do trees grow on mountain slopes? So it doesn't matter. For reference, Nevada is something like 80 million acres. Just Nevada, which is like 90% federal land, is 80 million acres.

I would never say that we should sacrifice Nevada to the AI and pave the entirety of Nevada from one wall to the other. But I just saw a bunch of things in my feed the last couple of days that Vegas is falling apart. The boomers are retiring. No one goes there anymore. People would go to see the 100 million acres of solar.

Dwarkesh Patel 00:28:56

Even if you did it in Nevada—

Casey Handmer 00:28:57

We can do it anywhere. You can do it anywhere you can find the land. People say that you can't do this in Europe because Europe doesn't have solar power. Europe has solar power. I've been to Europe in the summer. It's sunny for 20 hours of the day. It's a bit seasonal. But that's not a big deal.

Dwarkesh Patel 00:29:09

But I mean it is. Because energy is a small fraction of the cost, you care more about making sure the chips are running all the time, right?

Casey Handmer 00:29:15

In practice, what happens is… Let's say Europe hypothetically awakens from its slumber and decides it wants to participate in AI. I hope it does. They say, “Well, we're going to have to put 100 gigawatts of solar down at some point to build these data centers. It will most likely be in southern Europe. Spain is not particularly heavily populated. That's a great place to start. So we put in 100 gigawatts of solar data centers in Spain.”

Basically, if you're spending AI hyperscaling money on your GPUs, you want to have four nines of uptime in order to maximize your tokens per dollar spent on the entire project, not just on that. This is a very subtle point. I can go into vast detail on it later on maybe. Let's just say you need four nines of uptime. In order to achieve four nines of uptime in the middle of winter, you need to have a lot of solar overbuilt.

Is solar overbuilt a bad thing? No. Is the fact that we produce 40% more food than we need a bad thing? No. It's much better than producing 40% less than we need. It just means that effectively, you have a giant captive power plant attached to a data center that 99.9% of the time produces more power than it needs. 99% of the time it produces much more power than it needs.

That can now actually be the source of power for the local utility, which, instead of being like, “Naughty, naughty data center, you must disconnect when we tell you to”, they say, “Hey, data center, I noticed you've got a bunch of power you're not using 360 days of the year. Would you mind ever so much if we threw a power cable over the wall and we powered our entire town off your spare power at essentially zero marginal cost, plus whatever residential batteries that we need in addition to local power supply.”

Dwarkesh Patel 00:30:52

Brian Potter had a good analogy in his blog post about this. He's like, “My MacBook has a terabyte of storage and I use 100 gigabytes. I just got the terabyte version because it's cheap enough and I might need it at some point that it's worth it.” You're saying solar gets so cheap that it's the way we'll treat hard drive space. We get a bunch of excess.

Casey Handmer 00:31:13

Also the market will be made at the new marginal consumption and production. All the people who are working in the space right now are like, “Oh, I'm in the business of delivering power or storing power. I'm going to serve the AI market because that's where all the growth is occurring.” That's where all of US GDP growth is occurring right now.

Dwarkesh Patel 00:31:37

I guess you didn’t answer the question of , yes, theoretically we could do this, but is it going to be possible to get the permitting to have tens of thousands of acres of contiguous land?

Casey Handmer 00:31:48

It doesn't need to be contiguous. It helps if it's contiguous. It doesn't need to be convex. You can have a bit over here and a bit over there and you can wire them together relatively easily. In fact, in the limit, you have fields upon fields of solar arrays with…

Dwarkesh Patel 00:32:02

Tell me your dream, Casey.

Casey Handmer 00:32:03

Fields, just solar arrays as far as the eye can see. Then within the solar arrays, roughly in the middle of them, you have your batteries and your…

Dwarkesh Patel 00:32:11

I've played Factorio. I remember this optimal layout of batteries and solar.

Casey Handmer 00:32:16

You’ve got your batteries and you've got your data centers. So in terms of ground floor area, it's roughly, 10% racks, 10% access to the racks, maybe 50% batteries stacked up on top of each other, and there’s also cooling, something like that. That’s in terms of what sits in the centralized node.

That could be 100 megawatts or it could be 10 gigawatts, depending on how you want to scale this. But then all you need to connect that to the outside world is an optical fiber cable which you can string up on poles, you can run it underground. You could even use microwave links if you really wanted to. You could use Starlink if you really wanted to. I don't know if Starlink would be fast enough. I'm not sure if it's capacity is high enough. You could use laser links if you really needed to. That's it. It's this completely self-contained world of computation.

Dwarkesh Patel 00:33:04

Because it's off-grid.

Casey Handmer 00:33:06

Yeah it occurs off-grid, on private land somewhere in the backwoods of Texas where no one lives and no one will ever live because it's completely inhospitable to humans.

Dwarkesh Patel 00:33:13

In terms of the ratios, one trend that was impressed upon me is that the power density of racks is increasing a lot as the flops per GPU are increasing.

Casey Handmer 00:33:24

A megawatt per rack is what they're heading to now, which just seems bananas to me.

Dwarkesh Patel 00:33:28

I think it was even more than that.

Casey Handmer 00:33:33

Let's get concrete here for a second. Let's say you’ve got one rack and it's 1 megawatt. I'll leave the cooling to someone who specializes in air conditioners, but it's basically throwing air conditioners at the problem. Then you have batteries.

So in order to get four nines of uptime on this… In South Texas, you actually need less than this. But let's just say it’s 24 hours worth of battery storage. That means it'll get you through two bad nights in a row, basically. Actually, it turns out that you can significantly decrease power consumption with a very small reduction in overall compute. So if you've got like three really bad days in a row or something, you can dial back your power usage quite a lot without compromising your inference or training.

Okay, so you've got, say, a Tesla Megapack, something like four megawatt hours. So one megawatt rack, and then six Tesla Megapacks, each of which is roughly one truckload worth of stuff. So one truckload worth of rack, and then like six truckloads worth of batteries. Then in order to operate this at an average power of 1 megawatt, your solar arrays in Texas will be something like 25% utilization. So on average, if the sun came up every day and the day was the same length all the time, you would need 4 megawatts of solar arrays, which is about 4 acres of land. But in practice, because you're aiming for four nines instead of one nine, you need an overbuild of about 2.5x. So you've got about 10 acres of solar.

So 10 acres of solar, six truckloads of batteries, one truckload of data center, and some cooling stuff.

Dwarkesh Patel 00:34:58

For how big of a data center?

Casey Handmer 00:35:00

One megawatt. That's just for one megawatt. So 10 acres, one megawatt kind of situation at four nines.

If you want five gigawatts, then that's 5,000 times 10. So 50,000 acres. At a larger scale, you can probably cut all those numbers down by 10-20%, but it’s on that order. And 50,000 acres sounds like a lot.

Dwarkesh Patel 00:35:19

It does sound like a lot, is it not?

Casey Handmer 00:35:20

The amount of land put aside for Oak Ridge was 100,000 acres. The amount of land put aside for Hanford was about 100,000 acres.

Dwarkesh Patel 00:35:25

What's Hanford?

Casey Handmer 00:35:26

Hanford was where they made the plutonium in the Manhattan Project.

Dwarkesh Patel 00:35:30

But I don't know how big that was. Is it like, "Oh, this is so small," and then you're like, "Oh, but it's 100,000 acres," or…?

Casey Handmer 00:35:38

It's still largely unpopulated now because it's a National Laboratory. The reason they did that was they thought, “Oh, we're going to need four piles to produce plutonium.” These are not nuclear reactors that produce exothermal energy, so you can't actually make nuclear power with them, but you're making plutonium with them.

In the end, they only needed two. They wanted them spaced out because they thought they might just spontaneously explode, and a bunch of other facilities and plants and stuff as well.

Dwarkesh Patel 00:36:01

Austin Vernon had an interesting blog post where he said that if you have diesel generators or something which can take over 10% of the generation during winter, then you can have a 60% reduction in the amount of solar panels you need to install because you don't need to plan for that contingency.

Casey Handmer 00:36:26

Yeah, there's a balance here. This is not a very complicated optimization problem. For people who do optimization problems for fun, this is how you do it. You start off with a bunch of NREL data on what your solar abundance is in this particular part of the world, and then you just start throwing solar panels and batteries at it over the course of a one-year simulation until you hit the number of nines you want.

To an extent, you can trade the amount of panels and the amount of batteries you've got back and forth, and there's a very broad optimum. Or you can throw in a third thing like a diesel backup or a gas turbine.

Dwarkesh Patel 00:36:56

The issue here is—if Meta or Microsoft or whoever just wants to get something off the ground—this might be low opex to have this huge solar farm, but it's high capex, where you need to hire 30,000 people to go in the middle of a desert and install 50,000 acres' worth of solar panels. They're like, “Why would I not just buy 50 gas turbines instead?”

Casey Handmer 00:37:23

Why not just outbid Microsoft, or Meta outbids Google or something, for the last gas turbine that's available that year? Totally. The thing that Meta has realized is that Zuck is running out of time to spend his money to win. The capex is not crazy high, just to be clear. The capex is still dominated by just the GPUs. How much does five gigawatts' worth of GPUs cost?

Dwarkesh Patel 00:37:45

I don’t know if my numbers will be wrong but $250 billion or something?

Casey Handmer 00:37:48

$250 billion sounds about right. Is 50,000 acres going to cost $250 billion in Texas?

Dwarkesh Patel 00:37:56

That's so much money. Wait, I did the math in my head and like, that’s a lot of money.

Casey Handmer 00:38:00

We're talking maybe hundreds of millions of dollars, something like that. So it's like 0.1% of the cost is land.

How much does a megawatt of solar cost? If you go and ask the usual suspects, they'll tell you a million dollars. But this is one of the things that breaks my brain at Terraform, which is my day job. The modules themselves, without tariffs, would be 8 cents a watt, so that's $80,000…

Dwarkesh Patel 00:38:20

8 cents a watt? But they're like a dollar a watt, including installation and everything.

Casey Handmer 00:38:27

Including installation and everything. But the panels are the magic part. They're the thing that turns sunlight into pure electrical energy at 25% efficiency. Everything else should be less than that. If you want to work on that project, come and work with us at Terraform because we're very cost-sensitive.

Dwarkesh Patel 00:38:43

We'll give you an opportunity to shill, don't worry.

Casey Handmer 00:38:47

In all seriousness, the central takeaway is that the hyperscalers are not power cost sensitive. They are power availability sensitive. For all these things, you just run into this supply elasticity wall at the rates of increase that we're talking about. Solar is by far the best option for firehosing energy at a given problem because it rains down from the sky.

Dwarkesh Patel 00:40:24

Between the fact that maybe solar prices will go down and the fact that demand is going to go up. Do you think electricity prices are likely to rise?

Casey Handmer 00:40:33

Yes, but electricity prices at this point are a reflection of a regulatory irrationality. This is the same situation in Europe and Australia for that matter. Your prices will rise until you've had enough and you say, "No, we demand that you allow us to take advantage of power technology that's been invented in the last 50 years.”

In terms of things that are causing us to lose to China, tariffs are neither here nor there because as we've discussed, we're not sensitive to cost on power. But the environmental regulations that are actively preventing us from deploying renewable energy in the United States… This is the reason Texas is winning. Texas is out deploying California 10 to 1. The regulatory environment around solar is just insane. It's insane.

In the United States, part of the reason that solar has not been deployed at massive scale yet is that a bunch of laws went into action in the early 1970s that were intended to protect our environment. And that makes a lot of sense. And our environment's a great thing we should protect.

Dwarkesh Patel 00:41:31

I think people will be familiar with NEPA and whatever, but how is it especially impacting solar?

Casey Handmer 00:41:35

Let's say you've got a bunch of private land out in the middle of nowhere, and you want to build solar on it. You'll probably end up triggering NEPA, at which point you now have to do what is not in the law but considered necessary under current regulations. That’s your four-year environmental impact review, which generates so much paper that just the environmental impact of producing the report—because you have to cut down trees to make paper—is more than the environmental impact of just deploying the solar. This is bonkers. It is crazy town.

The thing that drives me particularly crazy in Southern California is that just because solar is kind of new, and off-grid solar is very new, unless you're very, very careful you end up getting regulated as though you're trying to build a chemical plant even though it's a solar array.

The impact of solar arrays on desert is arguably positive because it shades the ground and improves soil moisture retention. If you wanted to reverse desertification, you would basically just deploy solar panels on it and that would pay for the process. But you end up having to go through more stringent environmental review process than if you just wanted to grade the whole thing and cover it in concrete, or if you grade it and then park a bunch of old rusting cars that are dropping oil into the aquifer, which in many cases you don't need a permit for at all.

But to build solar, you have to go through this whole process. If there's one thing that anyone listening to this can do, it would be to have a categorical exemption for solar deployment. Or if I put money in an escrow account that says after 20 years we have to pull this out—we'll pull all the solar out of the desert and it goes back to being desert—I will do that in a heartbeat.

But if I have to hire another biologist for $10,000 to be like, "Well, on that 40-acre plot we found a tuft of grass which we believe might be one of the 20 species that this particular species of bee sometimes eats, and this species of bee is not technically endangered but it might be at some point in the future… Therefore, you can't deploy there." Even though it's zoned unrestricted industrial and it's sandwiched between a rocket test stand and a chemical plant, for example, in an industrial part of the desert… I'm going to become the Joker. It is insane.

We need to be a bit balanced about this. I don't want to drive species into extinction. But the meta problem here is if we don't move our industrial stack off fossil fuels in 10 or 20 years… First of all, we'll get poor the same way the UK did, because they ran out of coal, basically. The second thing is we'll get poor because we'll flood our coastal cities than Florida underneath climate change. We need solar synthetics for that part. We also need to do sulfur injection and a couple of other things.

Dwarkesh Patel 00:44:04

People will point out that transmission line growth has been stuck in a rut for decades. We have all these bottlenecks in terms of substations and transformers, etc. Why will this not hamper this abundant solar future?

Casey Handmer 00:44:22

That's a really great question. You and I had a conversation along these lines almost two years ago when we first met. It caused me to go and write a blog post. This is a good way of thinking about it.

Dwarkesh Patel 00:44:34

There's another blog post you wrote, which was also related to a conversation we had, which is “How to feed the AIs.”

Casey Handmer 00:44:39

That’s much more recent. That was after dinner, I think. To be fair, I usually am fairly clear in my blog posts if I'm shitposting or if I'm serious, but this one actually I'm dead serious on. It's actually the one where it's the most out of the money bet as well. Everyone else that I consider to be a respectable forecaster in this area disagrees with me on it. That to one side.

We know why the grid is expensive. It's a lot of wires strung up in hard to reach places that are hard to maintain, especially as the workforce ages, with regulations and all the rest and eminent domain and so on and so forth. So the grid's not going to get cheaper anytime soon or easier to build. If you look at the projections of how much grid the DOE would have us needing to build in the next 10 years versus how much's actually being built, it's not even in the same order of magnitude.

You say, “Are we totally screwed?” The answer is, “No, we're not totally screwed” because batteries actually do the same job that the grid does. This is kind of weird. Hear me out. The grid transports power from one place to another. It transports almost instantaneously at the speed of light, so it's actually performing a spatial arbitrage. The idea being that right outside the local nuclear power plant, power is really cheap because they make a lot of it. And in your house, power is really expensive because you don't have a power plant in your house. You pay the intermediary a small fee and they allow this trade to take place. That's basically how the grid works.

Until quite recently, the only way we had of meaningfully storing energy, storing electricity on the grid, was pumped hydro. That only works in a handful of places and with limited capacity. It doesn't work all that well either. The efficiency is not great.

Now we have batteries. Batteries store power at one time of day and they release it at another time of day. Batteries are performing a temporal arbitrage, an arbitrage over time. But they can be local or they can be more remote. I think we'll end up seeing batteries next to the solar arrays, and batteries in the middle of the grid at substations, and batteries on the sites of existing power plants that get turned off, and batteries in your house, and batteries everywhere in between.

One way of thinking of this is, what is your per capita allocation of batteries in kilograms per head? When you and I were much younger, the lithium ion battery was just in your cell phone. So let’s say it’s 10 grams per person or something. Nowadays half the people in this town drive Teslas, so your per capita allocation of lithium ion batteries is 100 kilograms or something like that. We're talking four or five OOMs of increase of total battery per person. That trend is only going to continue.

We've got batteries that are performing this temporal arbitrage. The sun comes up every day, right? So the power swings from midday—you're otherwise curtailing the solar array—to dusk when everyone's watching TV and cooking dinner or running the air conditioners to cool off in the evening. It’s very predictable. Whereas, "Oh, we had really bad weather, so we had to use the power line that runs to the extra power plants over by Hoover Dam or something." It doesn't get used nearly as much. Its peak utilization happens almost never, which means that the utilization of the batteries is on average, let's say 300 days a year. The utilization of your most expensive, highest voltage grid assets is much, much lower. That includes the substations and transformers and stuff that serve that.

So it's a really bad position to be in if you're a grid operator. You've got this aging existing thing that the batteries are cannibalizing. The batteries are being installed behind the meter. You don't have a say in whether they're being installed and how they're being used. All you know is that your utilization of your asset where you get to charge top dollar for it is just dropping year after year at the same time as your operating costs are increasing year after year.

So it's just very clear that the average distance the electron is going to travel between generation and consumption is going to decrease in the future pretty radically. It's already decreasing. It's going to continue to decrease.

Dwarkesh Patel 00:48:27

It's especially helpful for solar, but solar is the one that's most intermittent.

Casey Handmer 00:48:34

You can predict the amount of solar power you're going to get in three days pretty accurately because of weather prediction.

Dwarkesh Patel 00:48:38

But you can't change the amount of batteries you have.

Casey Handmer 00:48:42

Well, actually in the limit you can because you can put them on trucks and drive them around. There could be a capacity market for batteries where you drive them around to people who need them. In practice, it's going to be cheaper just to double the size of your battery because batteries are going to keep getting cheaper and cheaper.

But what it does mean is you can say, “Well, I know that I'm going to have three low days, so I will start curtailing now by 5% so I don't have to curtail by 50% in three days. Then overall for the whole year I'll only curtail five hours, so I'm still at four nines instead of having to curtail 24 hours because I can't predict the weather.”

Dwarkesh Patel 00:49:14

Okay, let's assume you're right. I think at some point, you will be right. Maybe we disagree about–sorry, I'm not qualified to disagree. Maybe you and some other person disagree about what year it happens. But it's hard to deny that in the asymptote, our civilization is headed towards lots of energy use for AI and a lot of that coming from solar. In that asymptote, I want to get to the crazy nerd sci-fi…. What does our civilization look like? What is happening?

Casey Handmer 00:49:43

Kardashev Level 1.

Dwarkesh Patel 00:49:45

Let's wait to get to turning the entire earth into an AI factory. But let’s say in the 2030s, where you've gotten multiple people who are building sites on the order of 5 gigawatts or 10 gigawatts.

The value of the hardware is dependent on its complement, which is the software. Right now, AI models are fine. The hardware they're running on, the economic value they can generate, is sort of bottlenecked by how good the software is. But if you actually had AGI, if you had human-level intelligence or maybe even better, running on an H100, that H100 is worth a lot. We're paying a lot for humans to do work. Right now, I don't think AI is that valuable. The models themselves aren't super, super valuable in terms of just pure economic value. OpenAI is generating on the order of $10-20 billion ARR.

Casey Handmer 00:50:38

That sucks. It's terrible. How can they sleep at night?

Dwarkesh Patel 00:50:40

But for context, McDonald's and Kohl's generate more yearly revenue than that.

But the promise of AGI is to automate human labor. Human labor generates on the order of $60 trillion of economic value. That's how much is paid out in wages to labor around the world. So that's what AGI can do. Even if you curtail it to just white-collar work, that's still tens of trillions of dollars of value.

So once we have models which are actually human-level, they will be worth at least that, pending the fact that you can build them or you can run them.

Casey Handmer 00:51:13

I don’t think we should constrain ourselves to being like, "Oh, well, maybe it'll be some fraction of current payroll," because that's very contingent on humans being humans.

Dwarkesh Patel 00:51:22

That's a lower bound, to be clear.

Casey Handmer 00:51:24

Oh yeah, lower bound for sure. But if you think about someone trying to estimate the upper bound for the market cap of Caterpillar based on, “Well, it takes this many men and wheelbarrows to dig a trench. So it couldn't be more than that.”

One way to think about the industrial revolutions is every time you figure out the industrial revolution, what you're doing is you're finding some way of bypassing a constraint or bypassing a bottleneck. The bottleneck prior to what we call the Industrial Revolution was metabolism. How much oats can a human or a horse physically digest and then convert into useful mechanical output for their peasant overlord or whatever? Nowadays we would giggle to think that the amount of food we produce is meaningful in the context of the economic power of a particular country. Because 99% of the energy that we consume routes around our guts, through the gas tanks of our cars and through our aircraft and in our grids and stuff like that.

Right now, the AI revolution is about routing around cognitive constraints, that in some ways writing, the printing press, computers, the Internet have already allowed us to do to some extent. A credit card is a good example of something that routes around a cognitive constraint of building a network of trust. It's a centralized trust.

Dwarkesh Patel 00:52:46

That's interesting. I want to credit James Bradbury and Gwern with making this interesting point when I was talking with them a couple of days ago. If you measure it by GDP, AI's outputs might be underwhelming. One of the complaints that economists have about the Internet is that it's hard to measure the consumer surplus that's created by the Internet because a lot of the goods and services that are made available, you pay zero for them. They don't show up in GDP.

Casey Handmer 00:53:14

Well, it's the same with oil.

Dwarkesh Patel 00:53:15

In the sense that energy’s like only 1% of GDP?

Casey Handmer 00:53:18

Well, oil is like $8 trillion a year or something, right?

Dwarkesh Patel 00:53:19

Yeah.

Casey Handmer 00:53:20

But if you said, “Well, one day we're going to consume 100 times more energy in the form of oil than in the form of food—and the per joule cost of food is whatever it is, the cost of a Big Mac—then oil should be like $800 trillion a year. Per unit energy, oil, like gasoline, is 100 times cheaper than the cheapest food that humans can digest. Does that mean that we've shot ourselves in the foot by using oil to run our economy because it's so cheap? No.

Dwarkesh Patel 00:53:50

Right. Also its fraction of GDP also doesn't correspond to how important it is. For example, oil is like 1% of GDP or something. But if you don't have oil, then you have these oil shocks, which cause double digit decreases in GDP. So the elasticity of demand often matters more than its raw fraction contribution to GDP.

Anyways, on the original point about AI, you're going to have this huge deflation. Gwern put it this way. He's like, “If you imagine Dario's data center of geniuses, how is that showing up in GDP? Well, it would be the inputs which are the chips, the energy, etc., and the outputs which are just the tokens. Neither of those is going to be that astronomical in comparison to the value that data center of geniuses is producing.”

In terms of GDP numbers, if that data center of geniuses automates or complements a bunch of human work, it might actually cause a nominal decrease in GDP while at the same time contributing massively to what we might think of as the valuable stuff human civilization can produce. In the long run, it might make more sense to think of the size of our economy, or the size of our civilization, as the raw energy use that we do rather than GDP. Again, GDP will see this huge deflation because the variable cost of running AI will just be pretty cheap as compared to paying humans wages.

Casey Handmer 00:55:23

At the point where you've got a mixed economy with an AI doing my job and also a human doing my job…

Dwarkesh Patel 00:55:31

I love how this is the new way we use the phrase "mixed economy".

Casey Handmer 00:55:35

Obviously, I still have some pricing power relative to humans, and the AI thus has pricing power. But if it were the case that a new kind of job emerges that AI is really well adapted to, because it's not competing against humans for most of those roles, it'd be competing against the other labs. You'd actually see the cost pushed down to a small multiple of whatever the marginal production cost of those tokens is. That would be my guess.

It might be a mistake to assume that if we're going to pay a top AI researcher $200,000 a year—Lol. Let's say for the sort of AI researcher that I could be, $200,000 a year—that if an AI comes along that's as good as me, even taking into account the fact that realistically speaking, I only get maybe 10 hours of really top cognitive work done a week, that it would also be worth $200,000.

Obviously, it'd be worth much more than that in the sense that you can copy-paste its output and much less than that in the sense of whatever the marginal additional cost of spooling up H100s is.

If some kind of role comes along that the AIs are really well specialized at and outcompete the humans quickly, then we'd also expect to see that both the cost of providing that service would drop drastically, at the same time as the overall value generated in the economy by that service would increase a lot.

Dwarkesh Patel 00:56:50

Exactly, if we think that the value of cognition is going to be unbounded, and the way to derive cognition—to the extent you think solar will eventually win—you can derive it from how much land it takes to power an H100 using solar panels. That is a very interesting derivation. At a minimum we're going to just fill up all the land. At some point you might have a declining marginal value of cognition or something.

Casey Handmer 00:57:20

We kind of discussed this earlier. If you have 10 acres of land feeding one megawatt of H100s or something, let’s say a megawatt is 1,000 humans. So one acre is a thousand humans’ worth of cognition. The implicit land value there is a lot higher than it is as undeveloped desert. It's also a lot higher than it is as the most productive farmland that humanity has ever had.

Dwarkesh Patel 00:57:47

At current hardware efficiencies. I don't know if it's worth spelling out. Basically, an H100 has the same amount of flops as a human brain, but also uses way more energy than a human brain. It uses 50x more energy.

Casey Handmer 00:57:58

Is that right?

Dwarkesh Patel 00:57:59

20 watts vs. 1000 watts? We know hardware can be at least as efficient as the human brain. The human brain can generate this many flops on 20 watts. If you do that calculation, that's 50x1000, so 50,000 AI souls off of one acre?

Casey Handmer 00:58:18

It could easily be much more than that because neurons are much slower than transistors, obviously. Probably 10 years ago, one of my friends reminded me, the way your phone saves power is it goes to sleep between you tapping out “hello”. H, it takes a nap, like 10,000 cycles. E… It's kind of nuts. I think Elon's talked about this in the context of self-driving cars as well. Anything humans do is glacially slow from the perspective of a computer.

Dwarkesh Patel 00:58:45

Let's go back to the original point. I was explaining why I think it's plausible that there could be more than hundreds of gigawatts of extra demand from AI in the 2030s. I want to understand what that looks like in the real world. At that point, it has become basically this industrial problem. Can you generate enough solar panels and solar modules and batteries, and not to mention the chips themselves?

Casey Handmer 00:59:09

That's the industrial point, and then there's a cultural point as well.

Dwarkesh Patel 00:59:11

Let's start with the industrial point. I want to know what the year 2035 looks like, if we've got AGI and we're just bottlenecked by the ability to deploy it.

Casey Handmer 00:59:21

What do you need in order to run? What is the minimum amount of matter that you need in order to perform these calculations? Right now we're talking about AI racking and grid and transmission and a bunch of ISOs and all the rest. You don't need any of that stuff. Obviously, xAI is on top of this because the first thing that Elon will always ask is, "delete anything you don't absolutely need."

What you actually need is a big slab of relatively cheap silicon to make the power, and then a small slab of relatively expensive silicon to do the thinking. If it's in space, that's all you need, because it's in the sun all the time, so you don't need a battery. If you're on the Earth, you need a battery as well, so you need some interconnects.

You don't need a transformer. You don't even need a DC-to-DC converter. You can actually make do with a buck converter or with relays or whatever to match the current output of your solar array with the charge state of your batteries and the power consumption of your GPU or something.

But a solar array about the size of this desk, for example, will generate about 500 watts in full sun. So you can actually imagine aliens who have different silicon technology stacks building their systems as an integrated solar array with a bit of computronium in the middle, for example, on the same wafer. But that's basically all you need.

Dwarkesh Patel 01:00:39

On the same wafer? Because it's all silicon?

Casey Handmer 01:00:41

It's all silicon all the way down. What's silicon made of? It's an element. It's chemically in the crust. There's no shortage of it.

Dwarkesh Patel 01:00:47

This is a great prompt for a sci-fi exercise, because especially in space, you don't need batteries. The future TSMC just manufactures integrated solar dies.

Casey Handmer 01:01:00

And they can fly around. They're solar sails, and they're relatively dense so they don't fly crazy fast, but they don't need to because they're immortal.

Dwarkesh Patel 01:01:05

Is this what the Dyson sphere will be made of, Casey? Is it just going to be computronium at the center of a solar cell?

Casey Handmer 01:01:10

They can fly closer to the sun to get more power, right up to the thermal limit, and they can fly further from the sun to go and explore or fly to other planets or something. They can adjust the orientation of the solar sail with LCD panels that could be integrated into the wafer itself. What's the post-human state? That's it.

Dwarkesh Patel 01:01:29

A solar sail with a silicon die in the middle for compute?

Casey Handmer 01:01:33

One human's worth of computation. One human brain can be simulated in roughly a square meter of silicon floating in space.

Dwarkesh Patel 01:01:39

How much, sorry?

Casey Handmer 01:01:40

1-square-meter of silicon, like the thickness of a sheet of paper, floating in space. That's the future human form. That's my final form.

Dwarkesh Patel 01:01:46

That's the attractor state.

Casey Handmer 01:01:47

That's assuming a little bit of software improvement, but I don't think that's…

Dwarkesh Patel 01:01:51

All that's assuming is software improvement. The Dyson sphere just needs a little bit of tweaking of the algorithm.

Casey Handmer 01:01:58

The area of the panel is the variable there. What do you need in order to make the silicon? Making solar arrays, making chips is a multi-stage process. Basically you start off with silicates, which are rocks ideally in a relatively pure form. You chemically reduce them. A couple of different processes can do that. Then you purify them into, ideally six nines of purity for solar arrays, maybe nine nines for really nice computers, and grow crystals, cut wafers, etc.

So then the constraint is, well, how quickly can you convert the crust into enough silicon to support silicon thought? What does the silicon ecosystem look like?

Dwarkesh Patel 01:02:38

Any thoughts?

Casey Handmer 01:02:39

Well, it's pretty quick. 1 kilowatt per square meter and then you use that just to rip oxygens off the underlying dirt, it doesn't take all that long too. You only need about 20 microns of silicon to make a solar PV array.

Dwarkesh Patel 01:02:52

You mean like actual dirt?

Casey Handmer 01:02:56

Yeah. Actual dirt has plenty of silicon in it. For example, setting up a brand new silicon refinery takes about 18 months. But that's just with the current technology, I actually think we may find new ways. One of the nice things about having infinite free solar power, approximately free solar power, is that you can revisit a bunch of legacy industrial processes that have been optimized for efficiency and say, “Well, what if we just use twice as much power and we just want to do them faster and cheaper?” Less capex, less lead time, more power.

Well, you can start solving problems. It turns out that if you want to chemically reduce silicon, you can do it electrolytically with less efficiency and under a hydrogen-rich atmosphere or something. One of the ways that silicon can be refined is by turning it into silane, which is a silicon tetrahydride. I'm not really a chemist, but I think that's right. So SiH4, which is a gas, it's actually like methane, but one down on the periodic table. Don't breathe it though.

Once it's a gas, you can filter it from all the contaminants which don't form gases or can be separated by density, much like how uranium is sometimes enriched, but much, much less difficult. You then heat it up to separate it back into pure silicon where you can then precipitate out a crystal.

Dwarkesh Patel 01:04:11

The reason I think this is interesting is because whenever people are talking about the AI singularity, often their expertise is not in energy or physics or whatever. They focus only on the cognitive elements of the singularity, which is like how much faster can we make AI smarter, etc.

I think this is really interesting. If we have unbounded cognition, which sets up both the ability to supply and to demand more energy, I'm very curious, what does the energy singularity look like? We're just trying to saturate as much energy that the earth receives and turn it into cognition.

Casey Handmer 01:04:49

I hadn't thought about that before, but there's this idea that evolution resulted in this continual ramification and complexification of the thermodynamic gradient.

You start with very simple RNA-based organisms. Now you get this industrial economy. But it may be the case—I don't have a strong reason to suspect one way or the other—that what we're seeing is the beginning stages of a collapse back towards the simplest possible thermodynamic-to-cognition stack. We have fusion in stars and the inky blackness of space and that provides our temperature gradient.

Then the most efficient way to convert that into usable cognition is silicon. Literally electrons being pushed across the Fermi gap in a solar array and then taking the return path through some set of gates, making decisions about things and then beaming lasers to their friends, saying, “Hey, I just made up a new meme.”

Dwarkesh Patel 01:05:37

That is an interesting concept. For 4 billion years we've been increasing the variance in complexity of creatures and then you might see this big collapse.

Dwarkesh Patel 01:05:50

Should I give you the opportunity to plug why people should work for Terraform?

Casey Handmer 01:05:55

Just to give you an introduction, Terraform is my day job. It's a company I founded almost four years ago. We are making synthetic natural gas from sunlight and air. We are also working on other core primary materials stuff. We also have a methanol process. Methanol and methane together are precursors to every hydrocarbon you could possibly want. Another chemical, ammonia, processed steel, desalination. We can also make cement and a few other things. Basically everything that the primary industry does, except for glass and paper.

We are hiring. Our jobs are available at terraformindustries.com. Yes, the website's meant to look like that because we're very cool. We are some very special people. I know a lot of smart people and I'm privileged to work with some of the smartest people I know. We are mostly mechanical engineers. I will never hire anyone who can't do math. I will never have the problem at Astronomer because we don't have a head of HR.

Dwarkesh Patel 01:06:45

Also the CEO is not having an affair.

Casey Handmer 01:06:48

Yeah, step one.

Dwarkesh Patel 01:06:51

I think that was a more crucial issue, Casey.

Casey Handmer 01:06:54

Heads of HR can get into trouble. I'm just saying, everyone does math. It's very important to me that Terraform is the place that ambitious hardware people go to become the best they can be. That is really important. It's not here to check in and get your paycheck and optimize some shiny widget. It's still a small team. It's still like a one project per person kind of situation. And I will level you up—maybe not quite like a Jensen “torture you into greatness” kind of situation, but at times it's going to feel that way.

You get to work with the best people that there are, at least on the West Coast of the United States, on this sort of thing. It's also a unique company. I thought years ago, by now we'll have competition. We don't. No one else is doing this except for a small startup in the UK. So you get in on the ground floor and it's going to be super cool technology. Eventually we get to go and build it all on Mars as well and help our robot overlords make more of themselves out of dirt. It's pretty cool.

Dwarkesh Patel 01:07:47

Nice.

Casey Handmer 01:07:48

Come work for us.

Dwarkesh Patel 01:07:49

Casey, thank you so much for coming on the podcast.

Casey Handmer 01:07:50

Thanks for having me.

Dwarkesh Patel 01:07:51

This was fun.

Casey Handmer 01:07:51

Yeah.

A deep dive with Lewis Bollard, who leads Open Philanthropy’s strategy for Farm Animal Welfare, on the surprising economics of the meat industry.

Why is factory farming so efficient? How can we make the lives of the 23+ billion animals living on factory farms more bearable? How far off are the moonshots (e.g., brainless chickens, cultivated meats, etc.) to end this mass suffering? And why does the meat industry have such a surprising amount of political influence?

For decades, innovation in the meat industry has actually made the conditions for animals worse. Can the next few decades of tech reverse this pattern?

Watch on YouTube; listen on Apple Podcasts or Spotify.

The welfare of animals on factory farms is so systemically neglected that just $1 can help avert 10 years of animal suffering.

I decided to give $250,000 as a donation match to farmkind.giving/dwarkesh after learning more about the outsized opportunities to help. FarmKind directs your contributions to the most effective charities in this area.

Please consider contributing, even if it’s a small amount. Together, we can double each other's impact and give a total of $500,000.

Bluntly, there are some listeners who are in a position to give much more. Given how neglected this topic is, one such person could singlehandedly change the game for 10s of billions of animals. If you’re considering donating $50k or more, please reach out directly to Lewis and his team by emailing [email protected].

(00:00:00) – The astonishing efficiency of factory farming

(00:07:18) – It was a mistake making this about diet

(00:09:54) – Tech that’s sparing 100s of millions of animals/year

(00:16:16) – Brainless chickens and higher welfare breeds

(00:28:21) – $1 can prevent 10 years of animal suffering

(00:37:26) – Situation in China and the developing world

(00:41:41) – How the meat lobby got a lock on Congress

(00:53:23) – Business structure of the meat industry

(00:57:42) – Corporate campaigns are underrated

Dwarkesh Patel 00:00:00

Today I'm chatting with Lewis Bollard, who is Farm Animal Welfare program director at Open Philanthropy. Open Philanthropy is the biggest charity in this animal welfare space. Lewis, thanks so much for coming on the podcast.

Lewis Bollard 00:00:10

Thanks for having me on.

Dwarkesh Patel 00:00:11

First question. At some point we'll have AGI. How do you think about the problem you're trying to solve? Are you trying to make conditions more tolerable for the next 10 years until AI solves this problem for us? Or is there some reason to think that the interventions we're making in terms of improvements, like in-ovo sexing or cage-free eggs, etc., will have an impact beyond this transformative moment?

Lewis Bollard 00:00:36

The end of factory farming is far from inevitable. Every year we're factory farming about 2% more animals globally. There are two possible trajectories we could go down. One is the trajectory that we have been on for the last century, which is that technology has made factory farming ever more efficient, resulted in ever more animals being abused and ever more intensive ways.

There is a trajectory where we reduce the number of animals on factory farms, where we reduce the suffering of each of those animals. So, even if we get AGI, I am really optimistic that that will accelerate forms of technological progress. It will bring us better alternative proteins and it will improve humane technology. But there are still huge cultural and political obstacles to alternatives. The cultural obstacles are that most people want real meat. Most people have the option already of plant-based meat that tastes about as good as real meat.

Dwarkesh Patel 00:01:30

Does it? I don't know.

Lewis Bollard 00:01:31

This is a debate. That's fair. This is a debate. But I don't think that's just the obstacle that people have. There are a lot of people who say, "I'm just not interested in the alternative. I want the real thing." Then there's also the political obstacle. Let's say that AGI solves cultivated meat for us. Cultivated meat is already illegal in seven US states. It might soon be illegal in the entire European Union. By the time we get AGI, will they even be able to sell it anywhere?

Again, there's a huge amount of good that technology can do in this space. I'm optimistic that AGI can accelerate that hugely. But at the same time, we should prepare for the significant possibility that AGI does not end factory farming. This is an incredibly efficient system that has persisted through all kinds of technological changes, and could persist through this technological change.

Dwarkesh Patel 00:02:16

What is it that makes it so efficient?

Lewis Bollard 00:02:18

The basic efficiency is that the animal, and the chicken in particular, has evolved over a very long time to be a being that can take in a relatively small amount of grain and convert it very efficiently into a form of protein that people like to eat. The feed conversion ratio for chickens, the amount of grain you put in to get meat out, is 2x. That grain is incredibly cheap.

The rest of the production process is incredibly cheap because they've removed everything that costs money, treating animals well and providing comfort and all that stuff. They've just gotten rid of it all. They've gotten down to the point where it's insanely cheap. You're trying to beat the price of grain times two, plus a few extra costs. That is actually a really hard target to meet. That's why factory farmed chicken is so insanely cheap today.

Dwarkesh Patel 00:03:08

Maybe an intuition pump here. We've been spending on the order of hundreds of billions of dollars a year in order to replicate human intelligence. And human intelligence has been developed—it depends on when you start counting intelligence as having started evolving—but on the order of hundreds to tens of millions of years ago, evolution has been trying to optimize for this intelligence thing. And we've had to spend all this effort in order to replicate it.

Converting calories into meat has been something that evolution has been optimizing for billions of years. Everything from the immune system to growth factors to delivering nutrition, etc, texture or whatever… This is what evolution is working on the entire time. So it makes sense why this is actually such a tough problem.

Are you ready to throw some cold water on your friends? How far away is cultivated meat actually?

Lewis Bollard 00:04:02

It completely depends on what we do from here. It also depends on what you mean by cultivated meat. There are companies right now that are selling cultivated meat, in very small volumes at very high price points, which is incredible. The challenge from here is how do you scale that and bring the price point down to compete with the incredibly low price point of factory farmed chicken.

How long it takes to get there, and indeed whether we get there, really depends on what happens from here. We are not on a path right now when it comes to the amount of venture capital funding available, when it comes to the current startups available… We're not on a path to reach cultivated meat that is cheaper than factory farmed chicken. We could get on that path…

Dwarkesh Patel 00:04:41

Ever?

Lewis Bollard 00:04:42

Depends. It’s contingent on AGI and contingent on what happens with AGI. I wouldn't rule it out, but I don't think it's the default path. I don't think it's the most likely outcome.

Dwarkesh Patel 00:04:49

Eventually, we'll have nanotech or whatever. At that point, raising chickens can't be the thing to do.

Lewis Bollard 00:04:53

You would think nanotech and bringing robotics in and all these things… But unless the cost of all those things goes down to close to zero, chickens are just going to be so insanely cheap. Maybe. It is totally possible that these AGI technologies introduce incredible new proteins that help solve this problem for us. But I don't think we should rely on it. First, because they might not be able to solve some of these problems to the point that it is as cheap as chicken. But second, because you still have these cultural and political barriers.

Dwarkesh Patel 00:05:20

The reason I think this is a very interesting example is because whenever people think about the use of technology to improve animal welfare, they're thinking about cultivated meat, lab meat, etc. They're thinking about these extremely far-off solutions. Then it makes sense why people who are especially concerned about this space, the first thought is not to just find ways to make the existing regime more tolerable, but to come up with some moonshot that changes the whole paradigm.

If you look at how much VC investment is going towards cultivated meat. Do you have some sense of how much goes into it a year? Versus how much VC investment goes into, “Okay, we've already got the farms. What is it that we need to do to come up with more things like putting the eggs through MRIs? Let's do these other small improvements in welfare”?

Lewis Bollard 00:06:13

There's a huge difference. The venture capital on humane technology is probably less than $10 million a year.

Dwarkesh Patel 00:06:19

$10 million?

Lewis Bollard 00:06:20

That would be my guess, whereas the venture capital on the alternative proteins has been in the billions over the last few years.

Dwarkesh Patel 00:06:26

Which has probably been motivated, at least partly, by the sense of, “We're going to make things more ethical.” People might not realize that in the near term, to actually make things more ethical, it might be just better to increase that $10 million pool.

Lewis Bollard 00:06:38

It's good to do both. Both of these are important. I can see why alternative proteins have a more promising allure to investors. First, it could be higher margins. But second, it feels more like the electric vehicle or the solar that just totally replaces the old practice. It's something totally new that replaces it. It has that potential for some portion of the market. But what I don't see happening anytime soon is the entire market switching over to these alternative proteins.

We need alternative proteins to meet the world's growing demand for protein so that we don't just have ever more factory farming. And we need humane technology to reduce the suffering within the factory farming that does exist.

Dwarkesh Patel 00:07:18

Whenever a discussion like this comes up, it's often phrased in the context of personal behavior. I think people will be assuming that what we're going to get up to is this push to make you vegetarian. I happen to have been vegetarian. I grew up a Hindu, so I've never eaten meat. Then I just stayed a vegetarian after I was no longer a Hindu. But then I started prepping to interview you and I'm like…

I don't know how valuable this is, especially if you look at some of these online charity evaluators and you're just like, “A dollar of your donation will offset this much meat-eating.” You're like, “What are we doing here?” But anyways, vegetarianism, overrated?

Lewis Bollard 00:08:02

I think we made a mistake as a movement making this about personal diet. It's great when folks want to make a personal diet decision, whether that is eating less meat or meat from more humane sources, but the focus should not be on the individual. This is not how large-scale social change occurs. We need government reform. We need corporate reform. People can be a part of that regardless of what they eat, regardless of what their diet is. We need people to be advocates and funders and supporters of this cause.

Dwarkesh Patel 00:08:30

How did we end up in this position? When people think about animal welfare, they think PETA, they think of protesters who are encouraging individuals to give up meat consumption? At the same time, these charities, which are so effective at corporate or policy changes, are just so neglected. How did this end up being the landscape of animal welfare activism and funding?

Lewis Bollard 00:08:59

It's a puzzle. It seems so obvious that you can have far larger-scale change at the level of governmental change and corporate change. Instead, we get fixated about whether someone is completely vegan or vegetarian. I think what happened is that when people started learning about this issue initially, it was just a few people and they felt totally powerless to achieve larger-scale change. They understandably focused on themselves.

Then it started to become a self-fulfilling prophecy. It started to become an end in itself, where it was about personal purity as much as about the impact you're having on the issue. It's much easier to measure your own personal purity than it is to measure your total impact on reforming factory farming. It just became this kind of inward focus. The good news is, that has changed tremendously in the last decade. The movement has gone from being one that was obsessed with personal purity, obsessed with dietary choices, to one that is much more obsessed with impact.

Dwarkesh Patel 00:09:54

This is why I really wanted to do this episode. I think people will be aware that there's a general problem here, but the actual politics, the actual economics, the actual state of the technology landscape here… there might be interventions which are stupendously effective, which we overlook just because people are not aware of what's actually happening in this space.

On that point, to use an analogy from global health and poverty. The Against Malaria Foundation estimates that it saved on the order of 180,000 lives, which is a lot. But then you compare it to China liberalizing, which brought a billion people out of poverty. That's just a many, many orders of magnitude bigger impact. In animal welfare, do you have some big take about what the “China-liberalizing” equivalent in this space is?

Lewis Bollard 00:10:42

There have been three large-scale drivers of progress so far. The first has been government policy. Advocates got the European Union to set basic animal welfare standards. That is billions of animals every year, billions every year. Then there’s corporate reforms. We see the same thing. There's this incredible scale across these corporate supply chains. McDonald's just implemented its pledge to go cage-free in the US. That alone is seven million hens every year out of cages, just in the McDonald's supply chain.

The third lever is technology. One example would be in-ovo sexing as a new technology that can get rid of the need to kill male chicks. In the egg industry, the unwanted chicks are killed at birth. In-ovo sexing has already spared about 200 million chicks from that fate. So there are these giant drivers. The good news is we're just getting started with them. There is the potential, I think, to help tens of billions of animals through these drivers.

Dwarkesh Patel 00:11:38

I want to go into in-ovo sexing. Just the fact that you can have a new technology and you can have basically Pareto improvements where things aren't getting more expensive… Maybe in the future they'll actually get cheaper because of this technology. At the same time, you're having improvements in animal welfare.

The problem, of course, with this industry has been that in the past, increases in efficiency have been coupled with increases in cruelty. I want to understand whenever the trend goes in the opposite direction, what causes that to be the case? What is the history of this technology? How does it work? Why did it take so long for it to come into common practice?

Lewis Bollard 00:12:14

The historical basis is a story of technology doing harm. Initially, the egg industry and the meat chicken industry separated because they realized they could grow meat chickens to be optimized for weight gain and laying hens to be optimized for laying eggs. That meant that the laying-egg industry had no need for male chicks. They couldn't lay eggs and they couldn't grow fast enough to be meat chickens. What they decided was to just kill them on the day they were born. The standard practice is that about 8 billion chicks globally every year are just thrown in a giant meat grinder or suffocated in bags the day they're born. Crazy.

This new technology is basically the application of existing technologies to scan the eggs in advance and work out whether they're going to be male or female. And then you can just get rid of the male eggs very early in the incubation phase. This technology went from, 10 years ago, just being a vague idea to today where it's already a third of the European egg industry. It just got introduced to the United States. We've got the first eggs coming out in the United States now. This is a technology that is growing rapidly. I'm really optimistic it can ultimately end this problem globally.

Dwarkesh Patel 00:13:27

How much was this driven by policy versus the tech being mature enough for it to be economical?

Lewis Bollard 00:13:33

It was both. First, there was some policy up front. Because advocates had drawn attention to this practice of killing male chicks. There was real impetus by governments and philanthropists to support kickstarting this technology. My estimate is it was about $10 million, a very little amount of public and philanthropic money that kickstarted this technology, got it to a point where startups could start to implement the technology.

Dwarkesh Patel 00:13:56

I'd be curious to understand exactly. MRIs have existed for a while, PCRs have existed for a while. Why did it take this long for this to be economical? What was the nature of that cost curve? I'm especially interested to understand this because it seems to imply… We didn't have to come up with some brand-new tech in order to enable this. Are there other things where somebody who is somewhat familiar with the technological landscape… People are always looking for startup ideas. Should they just spend a couple days at a big poultry farm or pig farm or something and see if things can't be improved?

Lewis Bollard 00:14:33

I think there's huge potential for technologists here. There is a lot of low-hanging fruit. This is primarily a commodity business that has only done things that reduce the price or increase production levels. It has not invested in animal welfare. And as a result, you find these things it's doing that just seem archaic.

Like the way that it is castrating piglets is with a blunt knife and with no pain relief. In that case, there was a new technology of immunocastration, an injection that achieved the same effects, and it was very easy to develop. I think there are a whole lot of other practices like that out there. There's a whole lot of these archaic practices being done where someone could come in, and with a little bit of smart work around this and an actual focus on animal welfare, bring in solutions that could potentially help billions of animals.

Dwarkesh Patel 00:15:20

One important dynamic to this industry that you've pointed out is that whenever we have to ruthlessly optimize for efficiency in one domain, it causes all kinds of other problems that we have to then make up for with even more cruelty.

Lewis Bollard 00:15:37

Think of what we've done to pigs. We took pigs inside from outdoors and we selected them to grow faster and to have this inadvertent greater aggression, the first thing they started doing was getting bored and biting each other's tails. That was a problem, so then we said, “We'll cut the tails off.”

Well, that didn't work. So then we had to start clipping the teeth and cutting part of their teeth off. That still wasn't enough when it came to the sows. So then we had to put them in crates to protect them from any other animal, and that wasn't enough. So then we gave them antibiotics and other drugs. At each step, there is a new solution that can't solve the fundamental underlying parts of the problem, and sometimes just makes it worse.

Dwarkesh Patel 00:16:16

Could we make chickens or pigs with no brains? Because it’s the suffering we care about. To the extent that their bodies are just these incredibly well-evolved bioreactors for converting grain into meat whereas optimization has led to more and more cruelty in the past, in this case, this is the ultimate optimization. They're not moving around at all. They are literally just a machine for producing more meat. The suffering is in some sense inefficient, if they're pecking at other animals, if they're catching on wires, etc. This is something that it would be better, even economically, to eliminate.

Lewis Bollard 00:17:02

You're right. The suffering is uneconomical at the level of an individual animal. The animals that we have selected for and the way we have treated them result in more of those animals dying, more of them having all kinds of welfare problems. The problem is that it is collectively more efficient.

If you can cram twice as many animals into a barn, it doesn't matter if 10% more of them die. That's been the underlying model of this industry. The reason welfare gets neglected is that it has a slight cost, but the efficiency gains are so much greater. I agree we should try and find things to reverse that. I am personally more optimistic about these kinds of incremental reforms. The average person listening to this is not thinking, "Oh yeah, I'm really pumped for the brainless chickens to come along and just persuade me."

Dwarkesh Patel 00:17:43

But they're not pumped about the cultivated meat either, right?

Lewis Bollard 00:17:44

No, sure. This is why you need a whole bunch of different approaches. There's no one solution that is going to satisfy everyone. What I would say on genetics is that what feels way more achievable to me in the near term is to get rid of the genetic physical problems that ail these animals. For instance, we've bred these chickens to be mutants that collapse under their own weight. We know that we can breed for far higher welfare birds that are still commercially viable. Indeed, there are companies, and there are places like Denmark, where the industry has already moved entirely toward these higher welfare birds. They have way better welfare outcomes that suffer way less.

Dwarkesh Patel 00:18:20

What's different about them?

Lewis Bollard 00:18:21

The first thing about them is that they are more balanced overall. Where the industry has just selected for rapid breast meat growth and for really efficient feed conversion, these birds have been bred to have robustness. They have broader legs so their legs don't collapse. They have better cardio systems, so they don't develop all these cardio problems. In general, they've just been bred for welfare outcomes. We're just like, “Let's just breed a bunch of birds and find the ones that die less.”

Dwarkesh Patel 00:18:48

Are they less economical?

Lewis Bollard 00:18:50

They're slightly less economical. This is why, because they haven't been ruthlessly selected for those two variables of breast meat yield and feed conversion. They cost a little bit more, and this is why you need advocacy to get people to adopt them. There has been huge advocacy in France, in Germany, in Denmark to get this. In fact just last month, the largest French chicken producer, the LDC Group, committed to moving its two main brands to these higher welfare genetics

Dwarkesh Patel 00:19:17

Why not think that they will just be eaten up in terms of their welfare impact? To the extent that the economics in the industry for a century have been to cram more things in, to figure out how to optimize along axes which just make the animal incredibly unhealthy and immiserated for longer and in more extreme ways? Okay, we'll come up with in-ovo sexing, but then there will be another thing which is the equivalent of gestation crates. Why think that even technologically the thing that is favored is the suffering-free optimizations?

Lewis Bollard 00:19:51

I think you’re right. This is the story of a lot of the industry's efforts to improve welfare. There was a study back in the 90s where they taught chickens how to select pain relief-laced feed. They found the broiler chickens were all selecting the pain relief-laced feed, suggesting they were all in chronic pain. The industry said, "Don't worry, we'll address it. We'll strengthen their legs." So they went away and they strengthened their legs for a bit.

Then they were like, "Wow, it's great. The chickens have stronger legs now. They can go and eat more stuff and we can put more weight on those legs." So then they made them bigger and essentially undid those gains. In recent years we've seen the mortality rate in the industry rising again and getting worse. So presumably they've just pushed so far again in that direction.

So that's a major risk. This is why you need the government or corporations involved. This is why you need the government setting out a baseline standard saying, “You can't go below this welfare floor.” For instance in Denmark, the government is strongly encouraging the move toward these higher welfare breeds and looking to ban low welfare outcome breeds entirely. You need to maintain those higher welfare outcomes.

This is what you need in corporate supply chains too. This is also what you see with the French retailers moving away from these low welfare breeds. You need them to maintain those standards. Because you're right, the industry left on its own will always find a race to the bottom.

Dwarkesh Patel 00:21:09

Potentially, we could find ways to make animals even bigger with the future forms of biological progress that some of my guests talk about. It's already the case that it's better to eat beef than chicken, because cows just have so much more meat per brain. What if we just got rid of the myostatin inhibitor genes or whatever, and then now there's even more meat per cow? Is that better because you have more meat per cow, or is it worse because it's potentially going to lead to the same dynamic of these overgrown, more suffering animals? Which way does that tilt?

Lewis Bollard 00:21:53

I think it probably tilts toward more suffering. This is what you see with the history of breeding these chickens to be the kind of mutants they are today. They've achieved a 4x gain in growth rates since the 1950s. That has led to a 2x drop in price, and that has led to a 3x increase in consumption. Because consumption has gone up so dramatically, and the suffering per bird has gone up so dramatically, that has outweighed the benefits of these birds being bigger.

Dwarkesh Patel 00:22:24

But the consumption might have gone up regardless. So actually then it's not clear. To the extent that we hold consumption constant, and maybe we shouldn't, they would have to be suffering 4x as much as a chicken in the 1950s for it not to be a net improvement. I don't know if you disagree with that.

Lewis Bollard 00:22:43

There is an in-between ground solution now. The higher welfare breeds that we are advocating for producers to adopt are not 1950s growth levels. They grow almost at 2025 growth levels scaled back slightly in a way that enables much larger welfare improvements. You don't have to go backwards to the level of these incredibly slow-growing animals.

Some people will want that. There'll be a market for like heritage chickens and people who are willing to pay for these extremely slow-growing animals. But the more realistic thing at scale is going to be these ones that still grow fast and still get big, but do so in a way that doesn't totally destroy their bodies and cause them to suffer so much.

Dwarkesh Patel 00:23:25

It's just striking me now that the way to think about what we're doing to these animals is not even… This would already be just incredibly immoral, finding creatures in the wild and then caging them up and then putting them through awful tortures. Rather, we are manufacturing creatures basically optimized for suffering. It's not even like, “We found this chicken and now we're going to put this in this little cage.” It's like, “We have designed this chicken to basically suffer as much as possible. We have literally genetically changed it as much as we can plausibly change it given the technology available to us today, in this Frankensteinian way, to suffer as much as possible.” That framing just makes it especially gruesome.

Lewis Bollard 00:24:18

I agree. This is the story of the chicken meat industry. They have just bred and bred so that these animals suffer more and more. I'll give you another example of that, which is the breeding birds. The birds that they have that are raised for meat are optimized to only survive until about seven or eight weeks of age. Even by that age, a lot of them are keeling over, getting lame, collapsing. But they're not at puberty yet. So they need to raise some of these birds past puberty to raise the next generation of birds.

For those birds to not totally collapse under their genetics, they have to starve them. What they do is they give the breeding birds about 30% of the feed that the birds would eat on their own. They're starving them 70%, because that is the only way to stop these birds from completely collapsing under the genetics that they've inflicted on them.

Dwarkesh Patel 00:25:05

In just reading about the accounts of, for example pigs in gestation crates, and the medical symptoms like swollen ankles, broken bones obviously from chewing the iron bars, all the bruises that cause ulcers, tumors, cancers, pus, etc… These are not rare medical emergencies, but the regular anticipated, expected outcomes across populations of pigs. Individual farms will house thousands of them and, of course, around the world a billion. I'm sure you've visited many of these places yourself or have had friends who've done so, right?

Lewis Bollard 00:25:51

I've visited factory farms. It is every bit as bad as it looks on the videos you can find online. It is every bit… The addition you see is, first, you hear the noise, the distress yelps from these animals. You smell it, it smells awful.

But the other thing I noticed, I visited one egg factory farm. It's impossible for farmers to provide individual care to each of these birds. This was a relatively well-run farm. Yet I still found a whole lot of hens stuck in the wire. Those hens are just going to slowly starve. Indeed many had. There were a lot of dead birds in with the live birds in other cages. That's just because of the scale. One farmer is trying to look after 200,000 hens. The only thing they can actually do is check the feed lines and check the water lines and remove some of the dead birds. In fact, that is the work of a factory farmer, largely removing dead animals.

It is just this dystopian thing where the industry presents this picture of, "Oh, we have individual care for our animals." The scale at which you were doing it has totally prohibited having any kind of individual care like that.

Dwarkesh Patel 00:26:57

This is an issue where scope sensitivity is just so insane, the magnitude. If this one battery cage farm was the only thing that existed in the world—if there was this one farm in India that had 100,000 chickens which were each just experiencing weeks upon weeks of pain through their life—that would already be a moral emergency. But it's so easy to forget that if there's 10 billion chickens that are alive at any point in the world. The whole problem is five orders of magnitude bigger than this one farm itself. So 100,000 times bigger than this one farm. It's just stupendous to comprehend the scale of the problem.

Lewis Bollard 00:27:44

It's crazy. You see this total confusion in the laws we pass, for instance, dog fighting. It’s a real evil, it's horrific, but we're talking about thousands of animals. Congress has passed multiple laws. Every state has made it a felony. It is being regulated, correctly, out of existence.

Meanwhile, the factory farming of pigs occurs on this far greater scale. We've even done the same thing with cockfighting, which is literally chickens. Again, it's literally thousands of chickens. We have rightly banned it. We've rightly made it felony animal cruelty. Yet when factory farmers do far worse to a far larger number of chickens, we call that commerce.

Dwarkesh Patel 00:28:21

Okay, so the positive spin on that can be that, because of how big the problem is and how neglected it is, the ability of any one person to have a big impact might genuinely shock them. Let's get into that. You are the biggest funder in this space, but cumulatively between you and the others, what is the amount of smart money that is being allocated to this problem?

Lewis Bollard 00:28:43

We think less than $300 million is being devoted to all work globally around every possible solution to factory farming across every country. Less than 200 million of that is what you would consider smart money going to evidence-based, effective interventions.

To put that into perspective, philanthropic climate advocacy alone is 50 times bigger than that. The work of cat and dog shelters and rescue groups in the US alone, 25 times bigger than that. There are individual conservation and poverty charities that are 5 to 10x bigger than that. So this is a tiny amount of money for the purpose of social reform. Yet it has achieved a huge amount, impacting hundreds of millions, billions of animals.

Dwarkesh Patel 00:29:27

What would happen if the amount of funding in this space doubled from the $200-300 million you mentioned that is being spent smartly?

I know you will say there's a bunch of things we could optimize around. There's so many neglected issues. But is there an immediate thing which is the thing that is directly at the margin, the next $100 million or the next $10 million would enable this?

Lewis Bollard 00:29:50

Additional funding would be transformative. We have a playbook that works on a number of these issues. One of the first things would be holding companies to account for animal welfare policies they've already made. We've got huge numbers of companies that made commitments to getting rid of battery cages and are now trying to back out of them or ignore them. With additional campaign funding, we could hold them to those, and as a result immediately improve the conditions of millions of animals.

For years, the industry used these battery cages that are these microwave-oven-sized cages. They cram as many hens in as they can and they leave them there for years. We know consumers don't think this is acceptable, but the industry doesn't disclose their use of them. It's not like when you pick up a pack of eggs, it has a big thing saying "from caged hens" or an image of where they came from.

Advocates went to the largest retailers, the largest fast-food chains, and said, "You need to move away from this because your consumers already expect this of you. This is what your consumers clearly want and they clearly don't accept this practice." They got pledges from almost all of the largest food companies, not just in the US but globally, to move away from these practices.

We're already seeing that this transition has already spared over 200 million hens a year from these battery cages. The US has gone from less than 10% cage-free to 47% cage-free. The European Union is now 62% cage-free. This is a huge transition.

Dwarkesh Patel 00:31:10

How did they do this?

Lewis Bollard 00:31:11

They captured this basic divide between what consumers expected was already happening and what was actually happening.

Dwarkesh Patel 00:31:19

I loved this specific example where there's a super tractable thing that is immediately available with the next millions of dollars in funding. Is there a particular charity which works on these campaigns in particular?

Lewis Bollard 00:31:32

I think that one great way to support them is to support a diversified portfolio of groups. There's a group, FarmKind, that allows people to donate to a variety of groups. Two of those groups that you can donate to through that platform, The Humane League and Sinergia Animal, are both working on exactly this.

Dwarkesh Patel 00:31:49

I think people might just not be aware of the ratio of dollars to suffering averted in this space. Can you give some sense of what we're talking about, dollar to suffering, here?

Lewis Bollard 00:32:01

The work to get hens out of cages has already spared over 200 million hens from cages. The work to improve the lives of broiler chickens has already benefited over a billion animals. That's just every year.

Dwarkesh Patel 00:32:17

Wait, sorry, it's 200 million a year?

Lewis Bollard 00:32:19

200 million a year.

Dwarkesh Patel 00:32:21

Sorry, I missed that. I thought it was cumulative across.

Lewis Bollard 00:32:23

No, cumulatively we're already well north of 500 million hens. We're into the billions of broiler chickens. If you assume these things weren't just around the corner, they weren't just going to happen anyway, if you think you probably sped up progress by years, decades, maybe it would never have happened… That cumulative impact over those years and decades is giant. We're talking billions, we're talking tens of billions.

The amount of money spent just on those corporate reforms was less than $100 million a year over a couple of years. We're talking about a ratio that is far less than 1:10 of a dollar per year of animal well being improved.

Dwarkesh Patel 00:33:02

So you're saying $1 can do more than 10 years of a better, more humane life? That is stupendous. A couple hours of pain is just awful and terrible. You're saying 10 years for a dollar? The reason why that’s so shocking…On its face it's shocking. But in other areas where you're trying to do global health or something, first, the problem is improving on its own. Second, with the Gates Foundation etc., there's tens of billions of dollars already being poured into the problem. It’s the same with climate change, etc. The idea that you would find an intervention where a single dollar can go this far is just genuinely crazy.

Lewis Bollard 00:33:52

It's very unique. The reason this philanthropic opportunity exists is because this area has been systematically neglected. Most people, when they think of philanthropy, do not think of farmed animals. Most people pile into popular areas like education and healthcare and climate. As a result, you end up with these outsized opportunities that no one has taken advantage of. If the space had billions of dollars in it, as other philanthropic areas do, you would not see opportunities like this.

Dwarkesh Patel 00:34:21

I won’t bury the lede any longer. I've always been interested in this issue. I lost track of it for a little while, to be honest. But I encountered you on Twitter and I started learning more about the issue. We chatted a few times in person, and that motivated me to have you on the podcast and also to donate myself.

As you mentioned, FarmKind Giving is this re-grantor. They don't keep any of the money themselves. They just regrant it to the most effective charities in this area. They're basically like an index fund across the most effective charities in animal welfare. It motivated me to donate to them. So I'm giving $250,000, and I'm doing this as a donation match.

This is to say that you the listener, if you contribute to this donation match, we can double each other's impacts. Between the two of us, we can allocate $500,000—if we saturate this, and I really want to saturate this—to the most effective charities in this area. Remember how neglected this area is. Lewis, as you were just mentioning, $1 that is donated in this area corresponds to 10 years of animal suffering that is averted, which is just stupendous to think about. There's no other cost area in the world which has such a crazy ratio. That has to do with how neglected this area is. The positive connotation of that neglectedness is just how big an impact any person listening to this podcast can have.

That's the donation match. The way you can contribute to it is to go to farmkind.giving/dwarkesh.

I also recognize that there's people in the audience who can do much more than this amount. Given how neglected this issue is… Remember, there's on the order of $100 million or $200 million that is being spent wisely on this topic. One such person listening could double the amount of money that is being spent effectively in this area.

That's crazy to think about. If you are one such person, just think about that. Even if you can't double the amount of money that's being spent in the area, you could cause a double-digit increase in the amount of funding that these effective causes are receiving. For those people in the position to contribute much more, or at least want to get their foot in the door and explore contributions of $50k or higher, Lewis, what's the best way they can reach you?

Lewis Bollard 00:36:46

In that case, we'd love to hear from you. People can message me on X, or they can reach out to my colleague Andres. That is Andres with one ‘a’. So it is [email protected]. He would love to work with you, and I'd love to work with you to help you spend that money as effectively as possible.

Dwarkesh Patel 00:37:08

Okay, but if you're like the rest of us and you need to start off on a more humble basis, again, your donation would already just have a huge impact given how neglected the space is. Again, the link is farmkind.giving/dwarkesh.

Dwarkesh Patel 00:37:26

Let's talk about other countries. You are not only the biggest funder in this cause area in the United States, but globally. Obviously, animal suffering in Sri Lanka or China is just as bad as animal suffering here. What is especially promising, especially given that more people in these countries will start eating meat? This problem is getting worse over time. It's getting worse because people are getting wealthier and eating more meat. What seems like the most useful intervention, or the useful thing to understand, about what to do about that?

Lewis Bollard 00:38:04

There are a couple of things. Countries, where their protein consumption is rapidly growing and there is not yet a deeply entrenched animal agricultural industry, have the ability to do things differently. In particular, they have the ability to support alternative protein work without that being politically toxic. For example, we see China investing very heavily in cultivated meat research. The majority of patents coming out globally on cultivated meat now are coming out from public universities in China. This is a case where the US is being overtaken because we have this entrenched industry that is ferociously lobbying.

I also think there's the potential to extend animal welfare policies globally. We're seeing multinationals like Unilever and Nestlé and even Burger King saying we shouldn't have cages in our supply chain globally. This creates the potential to spread best practices, just in the same way that factory farming spread from the United States globally.

Dwarkesh Patel 00:38:58

But factory farming spread because it was cheaper, not because there was some law passed that everybody else felt the need to copy.

Lewis Bollard 00:39:06

That's right. We essentially had economic efficiency spread factory farming. In some cases, that can spread higher welfare tech, for example, in-ovo sexing technology. Once that has been derisked enough, once it has been scaled up in Europe and the US, I'm optimistic it will become cheaper and then it will just be scaled out globally for economic reasons. But also, we can spread moral progress. We know that people in these countries also care about animal welfare.

I had a fascinating conversation. I went to a trade show and I talked with a company that manufactures gestation crates. I was like, what do you think about the future sales of these crates? They're like, "Well, we already have stopped selling them in Europe and the US." I was like, "Do you think you'll just be able to sell them in Asia forever?" They're like, "No way. As Asia gets richer and is on social media and sees the images and things, they're not going to be cool with this either. We know there is a limit to how long we're going to be able to sell these things for."

I think that gives me some optimism. As countries get richer, they generally get more concerned about this issue, and that then enables them to adopt animal welfare reforms as we've seen in the West.

Dwarkesh Patel 00:40:12

On net, is there a Kuznets curve here? Initially they get wealthy enough to afford the most economical forms of meat, which are battery cages, etc., and then they get even wealthier so that they can afford the potentially slightly more expensive versions of meat which are more humane. Or on net, is it just that you keep eating more meat through this whole process? Even if it gets slightly more ethical, the amount of meat consumption will rise 2x or 3x. So wealth always correlates with more suffering, basically.

Lewis Bollard 00:40:45

It's mixed. So far globally, wealth has heavily correlated with more suffering. The drive of people getting richer has led to them eating far more meat and far more of that coming from factory farming. We have overwhelmingly seen that trend across all countries.

In a few European countries, we are starting to see the dynamic where once countries have reached a certain degree of wealth, they are able to bring about reforms that actually reduce the total amount of suffering. I think it is quite likely that Germany has passed the top of that curve and is now on the other side of diminishing total animal suffering.

The critical thing to bear in mind is that this does not happen on its own. In Germany, this happened because there are very talented advocates who harnessed that public opinion and concern to drive corporate reforms to the retailers and to drive government policy reforms. We need to do that. I don't think you can just count that people are going to get to a certain degree of wealth and this is going to happen. I think it only happens if there is advocacy to mobilize that public opinion.

Dwarkesh Patel 00:41:41

A difficulty that these animal welfare policies have had is that even if you outlaw a practice domestically, to the extent that it's cheaper to produce meat that way, people will just import meat produced that way that is made elsewhere. States in the US who have tried to do this have had this problem. Countries in Europe that have tried to do this have had this problem. How do you solve the lowest common denominator problem in animal welfare standards?

Lewis Bollard 00:42:12

It's a huge problem. Advocates in the US passed ballot measures in Florida and Arizona to ban gestation crates. Then the pork industry just imported crated pork from other states into those states. So advocates then went to California and Massachusetts and passed ballot measures that extended the same standards to the sale of pork within the state. That you can't sell pork from crated pigs anywhere.

I think that is a critical move, and we're seeing the European Union now considering doing the same thing, imposing animal welfare standards equally on imports. That policy is critical to not just ensuring that you're not getting these laws undercut, but also to changing the political dynamic because domestic or local farmers are going to be very opposed to any law if they realize they're just going to get undercut by out-of-state competition, rightly so. This is a chance to also change that political dynamic so they can actually support the law, knowing that they are not at a relative disadvantage.

Dwarkesh Patel 00:43:05

Potentially reversed by an upcoming bill, right?

Lewis Bollard 00:43:10

This is right. The pork industry, unfortunately, has looked at these laws in California and Massachusetts and wants to do everything it can to undermine them. It knows this is the only way it can be effectively regulated, given that it has an absolute hold on the legislatures in Iowa and North Carolina, which are the main states for pork production. It knows that it needs to stop any other state from setting production standards or sales standards.

It first went to the Supreme Court. It first said this is unconstitutional, the states can't do this. The Supreme Court disagreed. We won at the Supreme Court. Now it has gone to Congress, and it's saying to Congress, "You need to wipe out these state laws. You need to stop them from doing this." The unfortunate thing is the Senate and the House are both on track to do that. In the upcoming farm bill, there is language that would ban states from passing laws on the sales standard, on animal welfare sales standards on goods. Right now, the default path is that that will pass as part of the farm bill in the next few months.

Dwarkesh Patel 00:44:09

If advocates are able to pass these laws or ballots at the state level, and it's popular enough that they're passing, why is it that at the national level they can't make a ruckus about this and prevent this from getting added to the full farm bill?

Lewis Bollard 00:44:24

The first problem is structural. At the state level, they've had to use ballot measures to get around entrenched lobbies. In this case, things start out in the House and Senate Agricultural Committees, which are heavily dominated by agricultural interests. The majority of House members signed a letter against this in the last Congress, but the vast majority of them are not on the Ag Committee. The Ag Committee gets to decide what's in this bill. The House Ag Committee just hosted a hearing on this. They only invited lobbyists for the industry. They didn't bother to invite a single opposing witness to their hearing.

We're also seeing that the industry is much better organized and funded on this effort than advocates are. The industry is constantly flying out a bunch of big industrial pork farmers, claiming they speak for the entire industry, telling the legislators this is their number one priority and absolutely has to be done. By contrast, animal welfare groups are not getting the same hearing. Legislators are not taking them as seriously as they take these Ag lobbies.

Dwarkesh Patel 00:45:21

Shouldn't there be some political constituency that's formed by the pork producers who are using more ethical standards and who are themselves being undercut by these Iowa farmers? Why aren't they getting flown out to these congressional hearings?

Lewis Bollard 00:45:38

That's exactly right. There is a large constituency of family farmers who support these laws because it has created a new market opportunity for them, where they can sell their already higher welfare meat and not be undercut by the industrial stuff. The problem they have is that they are far less wealthy and organized than the industrial pork interests. They don't have the money to just fly themselves to D.C. They can't stop farming. The people who are actually doing family farming can't just go to D.C. and hang out for a week because they need to be farming and looking after the pigs on their farms.

Dwarkesh Patel 00:46:11

But with the meat lobbyists also, given that it's a commodity business, you would think that there wouldn't be that much surplus that they can dedicate to political lobbying. Everybody here is not doused in cash. We can't subsidize a couple plane tickets for these family farmers. What's going on?

Lewis Bollard 00:46:30

There are people who are funding some of these family farmers to go to Washington, D.C., but we could see a far bigger effort. That voice is being hugely neglected in the debate.

The other thing I'll say on the money the pork industry has is that yes, it's a commodity business but it's also an oligopoly. You've got a very small number of firms that process the vast majority of pigs, and they do seem to make outsized profits. So they don't make the kind of profits you would expect. Across these industries, we constantly see price-fixing scandals and other antitrust scandals because it's a very small number of companies. It only requires minimal coordination for them to make greater profits than you would think they could.

Dwarkesh Patel 00:47:07

That might be good for animal welfare in the sense that if they can extract greater surplus, it makes it more possible for them to potentially invest in animal welfare. Not that they're necessarily doing it, but it would make it possible.

Lewis Bollard 00:47:20

It completely makes it possible. That's right. The absurdity of this is that the egg industry has been saying, "We can't possibly afford this transition to cage-free eggs." They, over the last few years of high egg prices, have made insane profits.

Dwarkesh Patel 00:47:35

How much?

Lewis Bollard 00:47:37

Cal-Maine, which is the biggest egg producer, its share price has doubled over the last few years. It's because the price elasticity for eggs is very inelastic. You can just keep cranking up the price on even a very small reduction in supply, and you can then take all that surplus. They've been doing that. As a result, you see a whole lot of these industries are actually flush with cash.

Dwarkesh Patel 00:48:00

But is it on the order of hundreds of millions, billions?

Lewis Bollard 00:48:03

It depends on the company. A lot of the egg producers are actually relatively smaller. It's the Tyson Foods and things that are on the billions. But they have the money to do these reforms. That is not the constraint. The constraint is the willingness to do the reforms.

Dwarkesh Patel 00:48:17

If the majority of House members have written this letter, apparently saying that this should be taken out of the farm bill… They're the people who are going to vote on this. So why is it still going to pass?

Lewis Bollard 00:48:28

The problem is this bill stopping states from regulating farm animal welfare meaningfully could not pass on its own. If it were put on the floor of the House and the Senate, it would lose. This is why they're putting it in the farm bill. The farm bill is this huge piece of legislation that includes all the farm subsidies, it includes all the food stamp assistance.

This is a bill that is considered a must-pass piece of legislation and is decided based on issues that most politicians consider far bigger than the issue of whether the state laws are wiped out. What the industry is banking on is that once they've got this in the text of the bill, people aren't going to sink the bill over this one provision, and it will sail through even though it's a deeply unpopular policy.

Dwarkesh Patel 00:49:12

I want your guide on how to corrupt the political process in the opposite direction. What insights do you have on how to actually have an impact on how Congress people or state legislatures vote?

Lewis Bollard 00:49:25

The good news here is we have public opinion overwhelmingly on our side.

Dwarkesh Patel 00:49:30

That's good. Ease the foot in the mouth I caused by saying the word “corrupt.”

Lewis Bollard 00:49:36

That's right. That's right. We don't need to be corrupt. The industry needs to be corrupt because they are trying to get politicians to do something that their voters strongly disapprove of. What we need to do is mobilize that base of support and show how real it is.

We need, for instance, to mobilize animal welfare advocates. We need to mobilize farmers who benefit from higher welfare standards. We need to provide them with an equal footing to the footing that the industry has provided to the very small number of factory farmers who have a stake in the system.

That requires the same things the industry is doing. It requires flying people to D.C. It requires getting people to go and talk to their politicians in their local district. Yes, it also requires money, because the industry is putting up so much money. Politicians need to see that there is also money on the other side of this issue.

Dwarkesh Patel 00:50:23

What would it actually take to… It's not clear what exactly you would do if you wanted to get this message in front of them. Abstractly, you can give money or whatever, but how does that actually transfer to political influence?

Lewis Bollard 00:50:41

My sense of what the industry does is that they get a whole bunch of their executives to max out on donations to politicians. The politicians then give them meetings. I wish this wasn't the way the system worked. I wish instead that politicians were actually just responsive to what the voters want. But given this is how the system works, I think that what people need to do is to bind together with a couple of other friends who care about this issue, max out on your donations to a politician, and then meet with the politician and say, "I really care about this and I'm watching what you do on this issue."

Frankly, you don't even need to just start donating. There are a lot of people listening to this who probably already donate significant amounts to politicians. If they started saying to those politicians, "By the way, this is something I really care about, and I'm watching what you do on this issue," I think you would start to see the political dynamic change.

Dwarkesh Patel 00:51:31

You wrote in one of your recent blog posts that the meat lobby spends on the order of $45 million in any given election cycle. They seem to be able to have influence on the topics they care about, which would be astounding and make all of us in tech jealous.

There are probably people listening to this podcast who could spend on the order of that kind of money on politics. But the ability of tech to have an impact on the kinds of issues that they care about is quite minuscule compared to the meat industry. What's going on here? What's the political economy of meat here?

Lewis Bollard 00:52:02

It's a real puzzle. This is an industry that accounts for less than 1% of Americans. It’s trying to defend wildly unpopular practices, and doesn't even get that much money. Yet they somehow have this total lock on the legislative process where they can stop any animal welfare legislation from passing.

There are a couple of things going on. The first thing is, it's not just them. They are fighting alongside the entire agriculture industry. There are allied industries, like the insurance industry and pharma industry, that have a big stake in factory farming.

It's not just the money. They appeal to this mythos of the American farmer. People think the American farmer is the good, hardworking, salt-of-the earth person. They sell the image of this person out in the fields, tending to their chickens and their pigs. They don't realize these are factory farmers. And they're extremely well organized, they have a very formidable lobbying presence in Washington D.C. and across state capitals. They have effectively used that to block any kind of regulation.

Dwarkesh Patel 00:53:01

You're telling me that tech bros aren't as politically sympathetic as a salt-of-the-earth farmer?

Lewis Bollard 00:53:06

There's this “children's kids book” rule of politics, which is that you should never mess with a character in a children's book. That's the police, that's the doctors, that's the farmers. I don't think there are any tech bros in the kids books yet.

Dwarkesh Patel 00:53:19

The front-end developers have yet to grace the covers.

Lewis Bollard 00:53:22

That’s right.

Dwarkesh Patel 00:53:23

You should describe the sort of franchise hierarchy type structure of a lot of these meat companies. Yes, you would anticipate that the Perdues and Tysons of the world would want a particular thing to happen in terms of political processes.

But the farmers who are indebted to these companies often have an adversarial relationship. Why are they able to form an effective political coalition with them?

Lewis Bollard 00:53:46

This is a great point. Most people don't realize that the way these factory farms are structured is that you have these giant corporations like Tyson Foods or Smithfield. They mostly don't own their own farms. Instead, they have these contract farmers who are essentially indentured laborers. They have a huge loan hanging over their head and they're farming. Why would those people support this?

The answer is they often don't. The agribusiness lobbying associations have done a very good job of pretending they do. They present themselves as representing the farmers. But if you look at their boards, if you look at the people who are actually leading these organizations, it's made up of people from the giant agribusinesses and the very largest industrial farmers. They do not have small contract farmers on the boards of these organizations. It really is a bit of a bait and switch where they claim to be representing those family farmers, but they're not.

Dwarkesh Patel 00:54:36

What is the reason that these contract farmers are willing to work with these large businesses? People will often say things like, "Uber is bad for Uber drivers," but I trust Uber drivers to know what's best for them. Why would these small farmers be working with these companies in the first place if it's uneconomical for them?

Lewis Bollard 00:54:52

It depends. For some people, it is just the least bad option they have, especially if someone just has a little wee bit of land and they want to preserve that land and they don't have other skills they can use.

I was chatting with this guy, Craig Watts, who was a chicken contract farmer for Perdue. He told me that when he got into the business, they made all these exorbitant claims to him. They said you're going to be making over $100,000 within years. They said just get out this loan, and it's going to be incredible. They told him all the things that could go right. Then he got into the business, and they slowly started eroding the payments to him. They slowly started paying him less and less. They slowly got to a point where he was making less and less money. He wanted out, but by that point he couldn't get out because he had this giant loan hanging over his head.

You've got a bunch of people who are stuck in the situation. There aren't easy alternatives because normally in one area, there will only be one processor that has a slaughterhouse in that area. There's no effective competition going on. Also, often you're locked in these long-term contracts as well. There is an element of people being locked in this, and then there's an element of people just not having better choices.

Dwarkesh Patel 00:55:54

What is the alternative use of that land? If you didn't work with some centralized processor, is the alternative use of that land for farming? If you've inherited some land and you want to figure out what to do with it, what can you do with it?

Lewis Bollard 00:56:06

Ideally, we would see pasture-based farming in those places. It doesn't require that much land, for instance, to have a pasture-based chicken farm. The problem is you would need to find a processor that you could work with. Normally, that just doesn't exist. Normally you've only got the giant players in an area and they say, "We just want commodity production. We don't want to fund you to do this pasture-raised stuff."

You get locked into that contract. Oftentimes, people who are doing pasture-raised production have to create their entire supply chain by themselves. They literally have to build their own slaughterhouse and create their entire supply chain around that, which drives up costs massively.

Dwarkesh Patel 00:56:41

Why is that? Because there must be enough consumers, even if it's not a majority of consumers, that there's some economic incentive to set up the economies of scale and supply chains that would make it easier to set up such a farm. Why doesn't that exist?

Lewis Bollard 00:56:55

There are people who are trying. Niman Ranch, for instance, has done this with independent pork farmers. There was a big effort to do this by Cooks Venture with pasture-raised chicken. Unfortunately, they just went out of business. The reason they went out of business is because there is such huge mislabeling across the industry that it's very hard to separate out what's actually better.

For instance, much factory-farmed chicken in the US is sold with the label "all-natural." We know from surveys that people think all-natural means the chickens were outside. It actually means nothing. But if you're trying to sell your product as pasture-raised next to a product that says all-natural, and people think it means the same thing and your product costs $2 more, you're not going to get very far. So long as we have this rampant mislabeling, it's very hard for the other players to get ahead.

Dwarkesh Patel 00:57:42

There's normal bananas and there's organic bananas. People are willing to pay quite a bit more for organic bananas. I feel like “pasture-raised” should be in a similar embedding space as organic. Organic is a huge industry even though it has dubious medical benefits, et cetera.

Then the problem is not that if there were accurate labeling, you'd think there might be consumer demand to make this a viable, much larger industry. It's just that it's very hard for consumers to identify which is which.

Lewis Bollard 00:58:14

Yes. You actually see that in the egg sector in the US. Within eggs, there is clearer labeling. Cage-free actually means something, pasture-raised actually means something. You can't just put the all-natural label on. What we see is that the pasture-raised egg sector is growing rapidly.

Even then, it is still handicapped by the fact that supermarkets use this as a price differentiation tool. They know that wealthier consumers prefer pasture-based eggs and are also less price-sensitive, so they mark them up heavily. The price you see is way inflated beyond the actual cost difference. Yet still, that is a rapidly growing sector.

Dwarkesh Patel 00:58:48

This is one thing I wanted to ask you about. One point you've often made is that you have to understand that meat and agriculture generally is a commodity business. In a commodity business, you'd expect all margins to be competed away.

I think you said in one of your blog posts that it costs 19 cents more for a dozen eggs to be cage-free, but often chains will charge on the order of $1.70 more for cage-free eggs. If it's a commodity business, why is it possible for supermarkets to extract this extra margin?

Lewis Bollard 00:59:14

This is the non-commodity part of the industry. The broader context on those retailers is that almost all the top US retailers have made pledges to stop selling eggs from caged hens.

A lot of them were meant to do that by this year, and a lot of them have not done it. Walmart and Kroger have not followed through. What they say is, "Our consumers don't want cage-free eggs because they're way more expensive." And it's true, they're way more expensive. They're selling them for like a dollar to two dollars more per dozen. When you look at the underlying production costs, it's only 19 cents difference.

What we see is these retailers are using this as an opportunity to get a big markup with less price-sensitive consumers and are in the process massively hampering their ability to fulfill their commitments. By contrast, Costco went 100% cage-free. They followed through on their promise. What we see is they are now selling cage-free eggs for the same price as Walmart sells its caged eggs.

There is that competitive pressure. Once cage-free becomes the new baseline, it does become the commodity market and you do see those margins competed away. It’s the same thing in states where they've banned the sale of caged eggs. Cage-free eggs now cost the same thing as the caged eggs cost next door. You do see that competed away once it becomes a commodity. It's until it reaches that point that you're seeing these crazy margins.

Dwarkesh Patel 01:00:29

Interesting. If these companies are already making these commitments, in many cases following through on them, to move towards more ethical ways of procuring meat, procuring eggs, etc…. I think I learned from you that McDonald's has made these commitments or that Chipotle has made these commitments. I didn't learn from McDonald's. What is the reason that this is not a more prominent part of their own advertising, given how much consumers care and how universally popular animal welfare is?

Lewis Bollard 01:01:01

The very best companies are advertising this, like Vital Farms or NestFresh eggs. They are out front focusing on the animal welfare benefits because they're pasture-raised and it looks amazing.

The fundamental problem for the large-scale companies is they have just made things less bad. It's still really good what they're doing. Moving from caged to cage-free is incredible. But there are two problems. One is, their consumers already thought they weren't using caged eggs. So if they advertise, "Hey, we're cage-free now," everyone's like, "What? What were you doing all this time? You didn't tell us you were using caged eggs."

And people still might think that even the new reality is not as good as what they thought things should be. They still would rather the animals were going outside, which they're not. And in a lot of cases, there are these phase-ins over time. So McDonald's is like, “In 10 years' time, we're going to get rid of the caged eggs.” You don't want to advertise that too loudly because then people are going to be for the next 10 years, "I'm eating caged eggs, and I didn't know that previously." That is just the unfortunate dynamic. Because this dissonance is so great between current practices and reality, that merely getting rid of the worst practices is not enough to create an advertising claim.

Dwarkesh Patel 01:02:05

Given how fast you're able to secure these commitments from different corporations, from retailers to restaurants, etc. It seems like corporate campaigns are even more successful than policy. Corporations are much more receptive. Obviously Perdue and Tyson are corporations as well. But the rest of the actual industry of getting food to consumers just seems incredibly receptive to these kinds of pressure campaigns. Maybe that's a lever of change that's especially salient.

Lewis Bollard 01:02:40

It's been phenomenally successful with these consumer facing brands, like the retailers, the fast food chains. Advocates have been able to secure over 3,000 corporate animal welfare pledges now globally, including from all the biggest retailers, all the biggest fast food chains, affecting hundreds of millions of animals.

The reason for that is twofold. The first is, there's a totally different structure from the structure in place on the legislative side. On the legislative side, if you want to pass an animal welfare reform in Congress or in any state legislature, it goes to the Agriculture Committee. The Agriculture Committee is dominated by a bunch of people who are in the pocket of big Ag, and they kill the bill. It never even gets out of that committee, let alone getting to the whole legislature. If you go to a company, you go to someone who is a decision maker who is not being lobbied by industry, or if they are being lobbied, is far less susceptible to that lobbying than they are.

I also think companies have just proven more responsive to consumers than politicians are to their voters. I think politicians have decided that they need to be responsive on the 10 issues their voters care most about. Maybe it's fewer than that. But on low salience issues like this, they can just ignore what their voters want and do what their donors are telling them what to do or what's easier to do.

Whereas I think what corporations are finding is, actually, if consumers are really outraged about this, then we need to act. Maybe this is higher on the list of salience for consumers at a retailer, because they're not worried about what their taxation policy is. For a retailer, actually, what is the quality of the goods you are selling? That is a pretty critical factor.

We know from surveys that when it comes to sustainability, animal welfare is the top thing people care about. So for all this talk we see from companies about climate change and prioritizing climate change, both the McDonald's and the Tysons and so on, they've all said this is the thing that consumers actually care about.

Dwarkesh Patel 01:04:28

Then what is the reason that the animal welfare movement has gotten so wrapped up with that? You go to most landing pages for animal welfare stuff and it'll be, "We're improving animal lives and we're making farming more sustainable. We're addressing climate change." That just seems really strange to me. Okay, we're torturing tens of billions of animals a year, but then also we're reducing emissions. We'll figure out some other way to reduce emissions, right? Whatever. How did this become the same issue in the first place?

Lewis Bollard 01:04:56

There's been this weird conflating. There's even been this very cynical exploitation of the climate issue by producers to not do animal welfare reforms. So something that Tyson Foods will say is, “We can't move to these higher welfare breeds because they would have a slightly bigger carbon footprint, because they eat a little bit more. Also, if you let the animals move around a lot, they expend more calories and that's got a bigger carbon footprint.” It's this total absurdity.

I had a conversation with the SVP for sustainability at one of the largest meat companies. What they told me was, "Yes, we know from internal surveys that animal welfare is actually more important to consumers. But we are far more responsive to what the fast food companies and the investors are telling us. And the fast food companies and the investors are obsessed with climate."

Dwarkesh Patel 01:05:43

The ESG stuff?

Lewis Bollard 01:05:46

The ESG stuff. I think they've all made these targets that they need to implement. Those targets are getting much higher priority than the targets they made on animal welfare.

Dwarkesh Patel 01:05:53

But then why do animal charities… It's not just a cynical attempt by the meat industry. If you go to animal charity websites, they'll often also emphasize sustainability on their landing page. I understand other people's psychologies are different, so I don't want to project the way I think about it.

At least whenever I see that, I'm like, "Oh wait, are you actually optimizing for the thing that makes this a really salient issue for me? Or are you just going to optimize for carbon footprint rather than this incredible amount of suffering that this industry produces?" So why are they doing this? Why have your friends roped sustainability into this area?

Lewis Bollard 01:06:34

I think a lot of people care about multiple things, right? They care about animal welfare and they care about sustainability. It is true that in certain cases, these things go hand in hand. Alternative proteins are both better for animal welfare and have a smaller environmental footprint. They are more sustainable.

This is not always the case. One thing that is wild to me is where you have people out there telling people to switch from beef to chicken because it's better for the climate. Literally that switch is 23 more animals per year you'll be consuming, costing several years worth of suffering in these factory farms, for a pretty marginal climate impact.

I do think there is often this tendency that climate just gets total precedence. It's just seen as obviously more important than any number of animals suffering. I actually think that that is more of an elite narrative than it is what regular people think. I actually think regular people are just pretty horrified by animal suffering and do prioritize that.

Dwarkesh Patel 01:07:29

Lewis, thank you so much for coming on the podcast and thank you for the work you do. You are allocating the largest amount of philanthropic funding in this space. You're a cheery fellow but I'm sure, day in and day out, this is not pleasant work to do, to learn about these gruesome details and how we can make the situation better. But it's awesome that you're doing it. Thank you for coming on, and thank you for your work.

Lewis Bollard 01:07:54

Thank you very much. And thank you for both being willing to take on this tough topic on your podcast and for making such a generous donation match. I'm really excited about the impact you can have there.

Dwarkesh Patel 01:08:03

Cool.

Lewis Bollard 01:08:04

Awesome.

After my last lecture series with Sarah Paine ended, I still had so many questions. I knew we’d only scratched the surface of Sarah’s scholarship, so I immediately invited her back for another series: she graciously agreed, and we’ll be releasing the results online over the coming weeks and months!

This first lecture is focused on the balance of power in East Asia at the turn of the 20th century. Specifically, how did Japan (population 47M) defeat China (400M) and Russia (130M) to become Asia's dominant power?

For me, the most interesting thing was that Japan's surprise attack on Port Arthur at the beginning of the Russo-Japanese War (1904) helps us understand why Japan might have thought Pearl Harbor would work.

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Google’s Veo 3 helps us visualize the hypothetical scenarios that often come up during our interviews. Veo’s ability to generate both video and audio—all with incredible realism—makes it perfect for bringing our content to life. If you have a Google AI Pro or Ultra plan, you can try it in Gemini today by visiting gemini.google.

• Hudson River Trading is one of the world's top quantitative trading firms. Responsible for around 15% of all U.S. equities trading volume, HRT powers their trades with cutting-edge deep learning models. Their in-house AI team does fundamental ML research and then applies it to some of the most competitive markets in the world. If you’re interested in joining them, you can learn more at hudsonrivertrading.com/dwarkesh.

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – Japan’s Meiji reforms

(00:14:44) – Trans-Siberian railway & Japan’s 3-year window for empire

(00:29:58) – The most important battle in the Russo-Japanese war

(00:48:38) – China’s implosion: imperialism, civil wars, and opium

(00:59:31) – Was Russia on track to dominate Asia?

(01:14:20) – Pearl Harbor (1941) vs surprise attack of Port Arthur (1904)

(01:34:03) – Why big countries still lose wars

(01:46:56) – Grand strategy for small countries

Sarah Paine 00:00:00

I'm going to talk to you today about one of the two great generations in modern Japanese history. They are the Meiji generation, named after the Meiji Emperor here. That generation transformed Japan into the first and only non-Western modern power in that period. The second great generation of the modern Japanese is of course the post-war generation that transformed their country into a global powerhouse.

I'm going to ask a question, or try to both ask and answer. What caused the reversal of the balance of power in Asia in the period that I'm going to talk to you about? It's a really consequential question about why these tectonic changes take place in the international system. Historically China had always been the dominant civilization in Asia from time immemorial. Then upstart Japan winds up doing things, or China winds up doing things, and it reverses and has profound effects. It's a very relevant question in our own day when there's an ongoing reversal of the reversal, when China's on the comeback and threatening to put Japan back in its box. So it's really interesting to ask why? How do these things happen?

So that's the background of what I'm talking about. But if you think about China back in the day before Japan trounced China in the First Sino-Japanese War, the Chinese believed that there was only one civilization, theirs naturally. They believed that of course it's the best. Because there's only one that makes it easier to be the best. But in addition, if you think about all levels of human endeavor, Chinese institutions were imitated throughout the East. It was the richest country on the planet for many, many years. Incredible achievements in science, philosophy, you name it.

Also there was another assumption that people didn't make a U-turn on the path to civilization. It's always forward towards Chinese civilization. Well, Japan by Westernizing is taking a U-turn on the road to civilization. It's dumping Chinese civilization. Already we have at least two civilizations out there. Then when it trounces China in a war, it suggests to the Chinese that they can't be better than the Japanese, at the military things at least.

This effect on China was far more devastating than the Opium Wars. The Chinese could write those off, the losses there to “a bunch of crazy Europeans, they're irrelevant to us.” But when Japan did this, it basically detonated the Confucian underpinnings of Chinese civilization. The Chinese have been trying to find a suitable replacement ever since. For a while they thought it was communism. Maybe they still do.

So I'm going to ask a question, why did the Asian balance of power change? Now, spoiler alert, I'll give the answer. I'm going to say, clever decisions in Tokyo. But I'm going to use a particular framework to answer this that I have found really useful. I learned it from teaching at the Naval War College where students are required to have a counterargument in papers. This is what I learned from doing this. So I'm going to have an argument, which is a thesis, and then I'll have some data supporting it. But I'm not going to quit there. I'm going to find the second best argument, the counterargument. The absolutely best alternate explanation, but not one that I think is the best one. I'm going to give you the best one. That's my thesis and I'm going to go into that. It's incredibly valuable, particularly in our own fraught political times where we need to hear each other out. You need to hear out the counterargument of what the other side is saying.

What you'll often find out in a counterargument is that there are actually some very valid points in it. It leads you to think, “Oh well, maybe I need to adjust my own argument.” So changing your mind's a good idea if the data comes in. Also, it gets you away from monocausal explanations where you come up with one cause you think “that's it, time to quit.” If you're thinking about the counterargument, you might get other causes as well.

Also, if you're going to do something like on a job and you have to recommend a course of action, your thesis, you had better anticipate what the counterargument's going to be. Then the third part of this is the rebuttal. Because you had better come into that meeting with your boss with the rebuttal in your back pocket so that you can deal with people who are saying you're wrong.

The rebuttal cannot be a repetition of the original argument because you know what? That's annoying. Don't do that. It’s really annoying. The most effective ones are coming at the problem from a completely different direction from either argument or counterargument that then shores up your argument. My other direction is, I've got a Sino-Japanese problem and I'm going to come around with a Russia angle.

So this is my game plan of what I'm planning to do and the analytical reasons for doing it. Okay, so I have a thesis which I'm going to give to you in emails and written work and lectures like this one. You really help people if you explain exactly what you're up to and you do it succinctly at the very beginning. So here it is.

The Japanese leaders Westernized their institutions. They integrated multiple instruments of national power into a coherent strategy. Then in the Russo-Japanese War, they quit that one exactly at the culminating point of victory for maximum gains. Together, these three things overturned the balance of power in their favor. That's my thesis. Short, sweet, you've got it. Whether you agree with it or not, we will get there.

Okay, so now I'm going to go to the first point. You should start with a topic sentence. I'm going to start one. Why am I doing this? These are all sign posts to orient you to my argument so that you can absorb it. Also, if you don't like it, you can see very clearly the parts that you don't like and we can get into a fun conversation.

Sarah Paine 00:06:20

My topic sentence for the Westernization part is that Japanese leaders concluded that in order to parry the threat of the Industrial Revolution, and of all these imperial powers coming at them, they needed to Westernize their institutions in order to protect their national interests. That was step one. So that's my topic sentence.

Okay, so what's going on? The Industrial Revolution started in England, or Britain more generally in the late 18th century. It spread to the continent after the Napoleonic Wars died down at the beginning of the 19th century. By the mid-19th century, it had reached Asia. It's profoundly disruptive to traditional societies whose traditional security paradigms no longer work when they're facing the weaponry of the industrialized age coming at them.

What the Industrial Revolution does and why it's so revolutionary is it produces compounded economic growth. Traditional societies are pretty much pretty stable. But when you compound economic growth, the difference in power and wealth becomes stark between those who do and those who don't.

It's also based, not only on technological changes—where you've got all these fancy armaments and railways and telegraphs—but it's also based on institutions. What are institutions? They're how we organize each other. When you think of institutions, you think of the buildings where people are. But that's not it. It's the people in there who are working on a shared project together, a shared area of activity. This is one of the hallmarks of Western civilization. This is what the Romans figured out about institutions and laws, that this is a way of really harnessing people, and it's profoundly powerful. I'll go into all of that.

Japan's looking at the world with this incoming industrial revolution, or the powers that have benefited from it. It's watching its neighbor China being defeated twice in war. They're horrified, not just appalled. They’re looking at it and going, “maybe we’ll be next?” And they're right. The United States does unto Japan what Britain and France did unto China. What's that? The treaty port system.

What it meant is that trade in Japan and China would take place in designated treaty ports, that the West would set tariffs on this trade, and that Western citizens in China or Japan in these treaty ports would not be subjected to Chinese or Japanese law, but home country law. When Chinese and Japanese citizens were in Europe and the United States, they most certainly were not dealing with home country law, they were dealing with US or Western law. So it was not reciprocal in any way.

In addition, each one of these treaties had a most favored nation clause in it which said, the one who's negotiating this treaty, the most favored one, whatever they negotiate will be given to everybody else. So it meant whatever one could negotiate accrued to them all. It meant that China lost sovereignty and Japan lost their sovereignty when these treaties went in.

So the Japanese, unlike China which fights war after war with these Westerners trying to defeat them militarily, and it's unsuccessful. The Japanese say, “Whoa, whoa, whoa, whoa, we're going to assess what the nature of the problem is.” They sent fact-finding mission after fact-finding mission to Europe primarily, but also the United States. This is just the most famous one, the Iwakura Mission, which was off to the West in the United States as well in 1871. They're studying not only Western military institutions, but a whole array of political, economic, legal, social, educational - the works - to understand the basis for Western power and the problem that is hitting them.

They arrive in Europe at a really interesting time. It's when Otto von Bismarck is just finishing up the third war of the unification of the Germanic states. The Japanese think, “Ooh, this might be quite a model for us.” Why? Because Prussia transformed itself over a succession of three wars from the weakest of the five great European powers to second only to Great Britain. It did so in part by unifying the Germanic states into modern Germany. The Japanese are thinking, “Wow, this might be relevant to us because we're divided up into all these feudal domains that we have just tried to glue together. What are the lessons to be learned here?” As they're thinking about this and watching what Bismarck is up to, they come upon thinking about institutions and technology.

I'm going to use the words in the following sense. Modernization means adopting the most state-of-the-art technology, whatever it is, not just military technology but all manner of technology. Westernization the way I'm going to use it means adopting Westernized institutions. I don't mean just military institutions. I mean everything from whether you westernize your educational institutions or political whatever it is. The question is, can you have one without the other? Can you modernize and have all the fancy gadgets and things without having the Westernized institutions that the societies that created these things had?

If you think about it, this dichotomy is still with us. There are a lot of fighters in the Middle East and North Africa who are more than happy to use state-of-the-art technology, but the last thing they want is Westernized institutions. The Japanese—when they posed this question back in the day asking whether you can have one without the other—they decided the answer was no. They didn't particularly like Western culture, but they believed that in order to not only use and import state-of-the-art technology, but become an independent producer of it, you've got to do some degree of Westernization.

So they get home. They set themselves a policy objective which is to protect Japanese national security and sovereignty in an age of accelerating imperialism. They come up with a two-phase grand strategy to do this. It's going to start with a domestic phase of Westernization, Westernizing your institutions. When they're done with that, they're going to have a foreign policy phase which is going to be about starting an empire. Why do that? Because they look at all the powers of their day and think “What's a great power look like in those days? Well, it has an empire.” So they go, “Well, we're going to have an empire, all right.”

This is the domestic phase. These are known as the Meiji reforms in honor of the emperor who reigned in this period. It's between 1869-1890. It's a whole generation. If you look at them, only two of them pertain to the military. There's the draft and then creating the general staff. Then if you look at the two that start it all, they start it at the top of the social pyramid with the feudal domains. Those are the power brokers of Japan. They're getting rid of all those and then they go right to the bottom of the social pyramid, which is children. They decide that they need to have compulsory elementary education because they don't believe you can have a modern country, a strong country, without a literate population.

But if you look at the rest of these things, you're getting a Bank of Japan, you're going to be having something running your currency and other things. You've got a cabinet, a higher education, you're going to have a professional civil service constitution, a parliament, you're going to have a court system that looks like a Western court system with laws that look an awful lot like Western laws.

As a result of doing all of this, the Westerners had no excuse left for having a treaty port system because this mirrors what's going on in the West. So Britain, which is the precedent setter in these things, the superpower of its day, it renegotiates its treaties with Japan on the basis of juridical equality. The other powers follow suit and do it. This happened in Japan a half century before China got rid of its unequal treaties.

Sarah Paine 00:14:44

The domestic phase is over the moment Japan signs that treaty with Britain. The foreign policy phase has to do with China. Japan believes it needs an empire and its neighborhood is a mess. China is imploding for various reasons, which I will get to, and Korea's even worse. China, because it's having massive civil wars throughout China, can no longer fulfill its suzerain role to stabilize Korea. And the Korean royal house is busy mailing package bombs to each other. I kid you not, they're blowing each other up.

What Japan is terribly concerned about is that Russia might try to fill this power vacuum. Why would Japan think that? Well, it’s the Trans-Siberian Railway. Russia decides in 1891 that it's going to build a Trans-Siberian Railway to exactly what? There is no Russian population out there. Japan understands exactly what it is. It's a bid for an empire in Asia. Because once Russia completes this thing, it's going to overturn the Asian balance of power because Russia is going to be able to deploy troops where nobody else can.

Therefore, treaty revision happens on the 16th of July, 1894. That's when it's signed on the dotted line with Britain. Nine days later, Japan fires the opening shots of the First Sino-Japanese War. The Japanese fight three wars of Russian containment. I'm going to talk about the two that went well for them today. The third one's a whole other topic. The first one's the First Sino-Japanese War, when little Japan defeats the greatest land power of Asia, China. Incredible. The second one, which I'll get to a decade later, is the Russo-Japanese War when the Japanese defeat, sorry spoiler alert, Russia, the greatest land empire of Europe. Amazing that they can do this. The third one does not go nearly as well. That would be the Second Sino-Japanese War from 1931 to 1945. That morphs into World War II. That ruins the Japanese, but it's a different topic.

The First Sino Japanese War, to let you know what happened in it. It's comprised of two pairs of key battles. There are other battles as well, but this is a good way to understand it. The first battle is at Pyongyang. The Japanese defeat the Chinese army, which takes off and retreats all the way over the border river, which is the Yalu, back into Chinese territory. So Japan has actually achieved its war objective, which was to remove Korea from the Chinese sphere of influence. Battle number one, they've already done it.

The second battle occurs within the week, the same week in mid September 1894. It's the Battle of the Yalu, where the Japanese navy trounces the Chinese navy, which, believe it or not, in this day both countries had state-of-the-art navies. Japan trounces it and gets command of the sea after that. That's terribly important for Japan. For Japan to reach the theater here, it's got to cross the sea. If there's a hostile navy out and about, it can sink troop transport, supplies and other things. It's very important to get rid of the hostile navy. The reason it gets command of the sea is because the Chinese decide they're never going to engage with the Japanese navy ever again and they duck into port. The Japanese are going to solve that problem for them.

There is a second pair of battles which are fought over the winter of 1894-95. China only has one naval refitting station where you can actually fix large ships. That's at Port Arthur. They will take it by land the same way they're going to take it in the Russo-Japanese War. The Chinese fleet, what's left of it, flees to Weihaiwei to hang out in port. Japan lands an army on the Shandong peninsula there and also it blockades with its navy. Then the army turns the landward guns on the ships in port and they sink them all. And that is the end of that war.

Here's what Japan got out of this war. What I've got is a very simple framework: domestic, regional, international. This is a way to help you remember what I'm going to tell you. Three-part frameworks are helpful for getting information to other people. Domestically, victory in this war validated a very controversial Westernization program. All those Meiji reforms which sound so great in retrospect, actually the Japanese population didn't like them. Who wants their kids being sent to elementary school if they were working on the farm before? Who wants this Westernized curriculum? Who likes westerners anyway? People are wearing all this Western clothing and stuff. It's crazy land. Why would anyone like that?

Once Japan wins this war and trounces China, a lot of Japanese have second thoughts about this. They're quite proud of their achievements and it vastly increases the prestige of the military, particularly the army. This is going to have bad follow-on effects for civil military relations because it's going to increase military power over civil power but it takes a while to play out.

Regionally, Japan's replacing China as the dominant power and Japan's getting the beginnings of its empires, Taiwan and the Pescadores. Internationally, Japan becomes a recognized great power. What's my proof? It would be the 1902 Anglo-Japanese Alliance, which is Britain's only long-term alliance between the Napoleonic Wars and World War I, alliance with Japan.

However, this war gets the eye of Mordor turned onto them because Russia is going, “Whoa, rising power in Asia. Potential two-front war problem for us with Europe in the west and whatever the Japanese think they're doing. That triggers a Russo-Japanese arms race and Russia, the eye of Mordor, turns from Europe to Asia. That's going to be problematic.

Sarah Paine 00:20:32

So now my transition sentence, I've done part number one. Not only did Japan Westernize its institutions in order to overturn the balance of power, but it also mastered grand strategy and integrated multiple instruments of national power. Here we go on that one.

Marshall Yamagata, who was the writer of war plans for the very successful Sino-Japanese war, predicted another war within the decade. The Russo-Japanese war came right on time. In the meantime, Japan prepared for war. It integrated such instruments of national power as diplomacy, intelligence, military, economics. I'm going to go through each in turn, starting with diplomacy. Here you have Sun Tzu, who's China's big guru, Art of War, who's talking about how it's really important to disrupt alliances. In modern terminology, that would be isolating the adversary. That might be a good thing to do.

That's the purpose of the Anglo-Japanese alliance. How does that work? Its terms say that if more than one European power comes to Asia to fight Japan, that means Russia plus one European buddy, then Britain is going to weigh in on Japan's side. Britain is the number one power in Europe, so why would you ever want to help Russia out? It won't go well with you if Britain is on Japan's side. This alliance goes into effect from 1902 to 1907. It's a five year event. It opens a window of opportunity for Japan to sort out its empire in Asia. But the Trans-Siberian Railway, when it gets completed, is going to threaten to close that window.

Here's why. The Trans-Siberian Railway in those days, it's not north of the Amur, it's actually straight through Manchuria. This Chinese Eastern Railway, it's Russia's bid for empire, trying to control Manchuria. It was unfinished. It hadn't been double tracked. That means you're always having to push trains off so other trains can pass them in the other direction. It's missing its Lake Baikal link. Don't think lake, think Switzerland. Lake Baikal is about the size of Switzerland. The Boxer Rebellion—the Al Qaeda of their day and I'll get to them—had destroyed much of the track, really upsetting the Russians.

As a result of all of this, at the beginning of the Russo-Japanese War which began in 1904, the carrying capacity of that railway was only 20,000-40,000 men per month to the front. By the end of the war, the last battle, it's 100,000 men per month. If those numbers had been available at the beginning of the war, Japan would have faced numerically superior Russians from start to finish and would have been in a world of hurt.

So Japan has got a window of opportunity that it's worrying about sorting things out in. In addition, Japan engages in a really big military buildup. It gets a really big indemnity from the First Sino-Japanese War and it spends it. That spending was finished in around 1901, meaning it's about ready to go to war. At the time the war breaks out, Russian naval assets in Asia were about ¾ those of Japan, but Russia was scheduled to surpass Japan's naval assets by about 1905. Again, you could see this window of opportunity threatening to shut. So if you look at it, Japan's window of opportunity of getting its empire, if that's what it thinks it wants… You've got to have treaty vision in place. You've got to isolate Russians, make sure that there's going to be no other power interfering in these things. You’ve got to have your rearmament program. But look, this window is very short. It's going to close in 1905-ish.

When you think of windows of opportunities, what they mean is whatever it is you plan to do has to be completed before it slams shut. If you're on the wrong side of the window, which is what happens to Japan in the Second Sino-Japanese War, you are in a world of hurt. In addition, what it means is that actually time is on the side of your adversary. It is a sign of weakness, not strength. But the Japanese are looking at the Russians who are procrastinating. The Japanese are telling them, “Hey, we will trade recognition of your dominance of Manchuria if you'll recognize our dominance in Korea.” The Russians didn't want to do anything. They're procrastinating and trying to go beyond this window. The Japanese are thinking, “We’ve got to sort it out before that happens.”

Another element of national power are psyops, as the US military likes to call them, psychological operations. The Japanese were engaged in a really wide array of them both at the front in Russia and across the Russian Empire. At the front, the Japanese were secreting in all kinds of postcards for their Russian recruits there, showing the great life of the POW and rather posh Japanese accommodations, as opposed to the really bad life of getting disabled or killed in the front.

Meanwhile, I think there were only three European countries, including Russia, that lacked a legislature in this period. I think Montenegro is one of them and maybe the Ottoman Empire might be the other one. Japan had a legislature. The Russian population's sick of it. The war wasn't going well. They start hitting the streets in the Russian Revolution. Japan wants to advertise that to the troops. You want things to be stirred up in Russia so that Russia has to pull troops back into European Russia. So they're doing all of that.

Then there’s this gentleman. He was a colonel back in the day, Colonel Akashi, but he's a general by the time this picture is taken of him. He's working in the Japanese legation in Stockholm. He's busy cutting checks to Finnish and Polish revolutionaries who are part of the Russian Empire and want out, trying to stir things up there to have Russia forced to pull the troops out of Asia.

Then the Japanese have this gentleman in there, and employing a lot of other people. Yuan Shikai goes on to be key in overthrowing the Qing dynasty and becomes China's first president. But back in the day, he's running reconnaissance missions for the Japanese and telling them what the Russians are up to. When little detachments of Russian troops try to go out and about, these people are harassing them, which doesn't help Russian morale.

Also, the Japanese are being really good about purchases from Manchurians. They aren't just taking things from people, they're actually paying. So they're triggering an economic boom in Manchuria, which means locals like them. Then the Japanese also figure out how to tap into Russian fleet communications so they know where the Russian fleet is, which is most convenient. So this is the information element of national power.

Then there's economics. Two-fifths of this war for the Japanese side is paid for with loans. If they don't get the loans, they can't wage the war. In fact, one of the reasons Russia has to call it quits at the end of the war is when it tries to raise a final loan. It failed. No one will pay for it. But the Japanese loans depend on battlefield success, as do interest rates. So if you are successful in the field, the interest rates go down. Japan is doing quite well with all of this.

So if you sum it all up, the Japanese use diplomacy to isolate their adversary with this UK alliance. They use all these psychological operations to promote revolution and desertions. They're using the military instrument to fund their rearmament. Then they've got the economics going with all of these loans. The US military is really partial to a Reagan-era acronym of DIME. D is for diplomacy, I is for information, M is for military, E is economics. It sounds like a bad cheerleading routine. It's inadequate. It may be a place to start. It's cute and all that stuff, but cute does not mean complete. Think about it. One of the most important factors in this war bar none, certainly for Russia, is railways. I don't see an R in there or anything. By all means, this is better than only looking at military factors, but it's incomplete.

Sarah Paine 00:29:58

So, I've talked about Japan's Westernization. I've talked about its mastery of grand strategy. Now here's my third reason for how they overturn the balance of power. It has to do with pegging the culminating point of victory in the Russo-Japanese war. Now for a commercial break, I'm going to give you some terminology: “culminating point of attack” and “culminating point of victory.” They're different. Culminating point of attack is an operational term. If you do not reach your culminating point of attack, it means you could have gone further.

The culminating point of attack applies to a single battle or a set of battles, which would be called a campaign. If you don't go far enough in your battle, you could have taken more territory, whatever it was you were after. If you go too far… Imagine you're going deep into whatever territory it is. Your lines are ever more extended. Your enemy's lines are probably being shortened. Your supply problems are getting worse, theirs might be getting better. If you go way too far, your enemy will launch a counterattack that will send you much further backwards than if you'd been a little more cautious about how far you went.

So that's an operational term. The strategic term is the “culminating point of victory” and it concerns the objective for which the war was fought. Japan is fighting this war in order to protect its sovereignty, and become a great power to do that. If you don't reach your culminating point, you could have gotten greater winnings. If you go too far, typically what will happen is you will trigger a third-party intervention. What may have been feasible before that third party joined the party may no longer be. So this is the point of the terminology. Commercial break is over and let's get going on the Russo-Japanese war. Here's a nice map of it.

It starts out with a Japanese surprise attack on Port Arthur. Lüshun is the modern name, I'm using a traditional name. It's on the Liaodong Peninsula. This is the main Russian base. The Japanese have to get those ships in base and sink them, even better. Their supply lines are in danger if that navy's out and about.

Simultaneously they land an army in Korea that's going to go northwest into Manchuria. If you look at the railway line, it goes Port Arthur all the way to Harbin up there. That's the east-west junction of the Trans-Siberian Railway. In those days that's the Chinese Eastern Railway. I've only listed a couple of the major battles, but basically it's going up the railway system from Port Arthur upward and you got the Battles of Liaoyang and Mukden.

Japan has got only four armies until the very end of the war when it gets a fifth. As long as Port Arthur has ships in it, one army has to be stuck there besieging it. Japan's theory of victory in this thing is to have an annihilating battle. That's what Bismarck had done to the French in the Battle of Sedan. So the Japanese really need to get that army out of Port Arthur so that it can concentrate on these other battles. What happens is Russia keeps losing the battles, but it has an orderly retreat moving ever further northward, extending Japanese lines. That's an overview of how the war goes so you can orient the rest of the conversation here.

If you look at the invasion routes that Japan uses and in the First Sino-Japanese War and the Russo-Japanese War, they're remarkably similar because guess what, the geography hasn't changed. If you want to send an army into desired locations, it may well have to take similar routes. The British, the French, the Americans, and the Japanese had studied very carefully the First-Sino Japanese War. But apparently the Russians didn't waste their time on it because here's what's up.

If the Russians had studied it carefully, they would have known the Yalu River is lethal, to send an army across that if there's an army waiting for it on the other side. The Russians also would have realized that this Fenshi Pass and Motien Pass—Motien means literally “scratch the skies”—that if you're prepared there, you're going to ruin an army coming through. Of course the Russians aren't prepared there.

They also would have known that the Liaodong Peninsula—it has a very narrow neck. If you cut the narrow neck right where Dalny is, a big commercial port, you're going to get that port which is connected to the railway. You're going to be able to supply your armies going right up north. In addition, it means you're probably going to get Port Arthur, the big naval base as well, because it's basically been turned into an island and you'll probably get that. So the Russians should have figured out that but they didn't. The Japanese literally blast through these locations.

Here is the key to reducing the fleet in Port Arthur. Since the Russians wouldn't sortie, they're just sticking in harbor, the Japanese have to figure out a way to do it. Here's how it works. They have to get a bunch of these 11-inch howitzers in place. Japan didn't own many of these. They're really heavy. It's really hard to transport them with horses and things. They've got to put them behind the hills in Port Arthur. Then they've got to take the high point. It takes them a while to figure out that they needed to do this. It’s a 203-meter hill. You put a spotter on 203 Meter Hill. This thing starts firing shots and the spotter starts radioing in how you have to adjust the line of fire in order to hit the ships in harbor. So that's what's going on here.

Here you can see the narrow neck there above Dalny. You can see it's a big port. We've put a lot of ships, but it’s not a very protected port. The star is where Port Arthur is and that is a protected port where you want to put naval vessels. You can see it takes the Japanese two months to work their way down the Liaodong Peninsula to get themselves in position to blow away the ships in port.

Meanwhile, the Japanese armies are having a hard time while this is going on. It could be argued that maybe they even reached their culminating point of attack in the Battle of Liaoyang. Who knows? Liaoyang takes place less than halfway through the war. It's in early September 1904. And Japanese munitions, numbers of officers, numbers of horses… It all goes critical. They don't have enough. They get away with it. They win the battle but the Russians have an organized retreat northward. At the Battle of Shaho, which happens next in October, the Japanese supply system almost collapses but they get away with it.

So bad things are going on at land. Japan doesn't have an infinite supply of soldiers. They also don't have an infinite supply of artillery, howitzers and things. Here are their gun emplacements. Notice the green ones are army gun emplacements around Port Arthur. That makes sense. What's all the purple? They have so few guns, they're pulling them off ships in order to reduce Port Arthur. All these things they really need up in Manchuria. They need General Nogi who's running all of this, and his army, up in Manchuria. There you can see the circle is where 203 Meter Hill is that they need to take as well.

General Nogi, who is commanding the 3rd Army which is in charge of the siege, is really desperate to reduce the fortress as soon as possible. He runs four different really costly infantry assaults on a fortress where you're killing lots of young Japanese men. The first two occur right prior to major battles, Liaoyang and Shaho respectively. Because he wants to win those battles, win Port Arthur, so then he can take his guns and troops to fight at Liaoyang and Shaho. Well, it's not to be because he loses both of them. The assaults are insufficient.

It's only after the second one that he realizes the importance of 203 Meter Hill. The problem is Nogi gets 45,000 Japanese soldiers killed doing this. 45,000 soldiers in that day is an entire army. I've already pointed out that Japan only had four armies until the very end of the war when it tried to cobble together a fifth one. Here's what 203 Meter Hill looks like from down below. Here's what it looks like from up top, the kind of view you get. When there was a truce in early December 1904, and there was some talking between Japanese and Russian officers, Russian officers said, “You will never capture 203 Meter Hill.” To which the Japanese officer replied, “We'll purchase it in blood.”

And they did. Their gamble paid off because all these ships, you'll notice they're listing. It's because they ain't sailing anywhere. You can see the 203 Meter Hill. They take it at the end of November and within the week, they are now sinking all of these battleships. They're gone within a few days. The Russians give up at Port Arthur and the 3rd Army is up and heading into Manchuria where it'll be there for the Battle of Mukden, which is a huge battle. It's got what, 500,000 troops?

Even so, Russia can muster 125,000 more troops than Japan can. Japan in this battle is just taking anybody, young boys, old people. Whoever they can put into that army, they're putting into it. You could argue that they're well beyond their culminating point of attack, but for incompetent Russian strategy. If the Russians had run one more battle against the Japanese, the Japanese supply lines would have collapsed. It's unclear how far down the Liaodong Peninsula they would have had to retreat.

Sarah Paine 00:39:45

So this is when Japan's war termination plan goes into effect. The Japanese realized from the very beginning that they had a high-risk, high-reward strategy. In contrast to World War II, they had a really carefully prepared exit strategy. When Prime Minister Itō Hirobumi is convening the cabinet and they're going to make the decision to fire the first shots in this war, he is already lining up a Harvard grad, Viscount Kaneko, who was an acquaintance of the President Theodore Roosevelt of the United States. What they want is to have Viscount Kaneko work on getting Roosevelt ready to do mediation at the end of this war. The negotiations are going to be held in Portsmouth, New Hampshire, negotiated by another Harvard alumnus, Baron Komura, who's going to do all of that. Meanwhile, Ito sends his son-in-law, who's a Diet member but also a graduate of Cambridge University, off to Britain to keep the Anglo-Japanese Alliance solid. So the Japanese are very aware that you have to have a way out of this thing.

So it's in the Battle of Mukden that Field Marshal Yamagata decides it's time to call the American card at this moment. He said, “Look, the enemy is never going to request peace unless we have invaded Moscow or St. Petersburg” something he knows to be impossible. He said, “Look, the enemy still has powerful forces in its home country. We have already exhausted ours. Second, while the enemy still does not run out of officers, we have lost a great number since the opening of the war and cannot easily replace them.” At this moment, the Russian army was three times the size of the Japanese army. That army in theater was increasingly comprised of crack soldiers, not the colonial kind of soldiers they'd started out with.

So here's Field Marshal Yamagata. In the first Sino-Japanese War, the government in Tokyo was afraid he was going to march on Beijing and do regime change. They pulled him out of the theater and gave him a sinecure to make sure he didn't do that this time. This time, he's a wiser man. He goes, “We must now be prudent.” Here are other members of this brilliant Meiji generation. The Chief of Staff of the Manchurian army said, “Look, if you start a fire, you’ve got to put it out.” And here is Field Marshal Ōyama, who is the commander of Manchurian forces. Before he sets out to take command, he tells the navy minister, “I will care for fighting in Manchuria, but I'm counting you as the man to tell me when to quit.” So the navy's going to be the fire department, apparently.

To tell you what the navy was up to, they took Elliott Island to turn it into a cruiser base. That's how they're running the blockade operations on Port Arthur, keeping the Russian fleet in until they can eventually sink it there. In addition, the Imperial Japanese Navy is busy laying mines. A lot of ships go down to mines.

This is an interesting story. Nicholas II, tsar of Russia, doesn't like it that his fleet is just sitting in port, not doing anything. So he sends Admiral Makarov out there to assume command. I've had the misfortune to read Makarov's major literary contribution, which is all about naval strategy. He's a real hero in Russia. But then I read his book. It's unbelievable. He said, “If you face a big ship, what do you do? A more powerful ship, you attack. If it's a smaller ship, what do you do? You attack.” It's one size fits all. I mean, how unanalytical.

Here's what happens to him. He comes into the theater and within the month, another spoiler alert, he's dead. How did that happen? He followed his strategy. That's how it happened. What he does, he arrives, everyone's really excited because he's getting the guys out and thinking they're going to be doing things. But on April 12, the evening before he dies, there are a bunch of Japanese ships laying mines out at the harbor's entrance. He mistakes them for Russian ships and tells his people not to fire on them. Oops.

The next morning, when he sorties with the whole fleet to take the offensive… Well, they didn't sweep the harbor entrance for mines. His ship hits a mine and he goes down with a ship. Alright, that is a tactical error and he's a dead man because of it. However, there's a much more important strategic error that follows. Two weeks later, Nicholas II decides to send his Baltic Fleet, named for the Baltic where it is, all the way around to come and relieve the siege at Port Arthur in order to avenge all of this thing. Okay, what's the problem with that plan? Well, that's the problem with that plan, right? It's a long way to go. There is no air conditioning for the sailors. They come from northern climes. By the time they arrive, they're mutinous. Their ships are covered with barnacles, which means they don't move very fast. Meanwhile, once Port Arthur has fallen, the Japanese immediately refit their navy so they were good to go.

So the way this blunder plays out… Well, if you think about Britain, Britain will do these worldwide adventures, but they had a complete basing system across the globe so that their ships had coaling stations, they could refit, put more water on, give people some R&R time, relaxation and things. Also the British were very careful to fight on their own terms as ever possible. You don't want to fight on the enemy's terms. Well guess what, you're going to be fighting right next to Japan with barnacle ships and mutinous crews. That's not going to go well for you.

The Japanese know because…Port Arthur falls while this expedition is taking place. So then the purpose of the expedition is no longer there. They should have taken a U-turn and gone home. But no, no, no, they're going to keep on going. There is only one other possible location for them which is a far inferior base up in Vladivostok. The Japanese know there are three ways to get to Vladivostok. If you're coming up from the south, you're either going to go the long, long way around through La Pérouse Strait and it's awfully narrow and it's a long way around. You can go through the Tsugaru Strait which is a little wider, but it goes by a bunch of big Japanese army bases and also a very large port of Hakodate. You want to do that? That's kind of risky. So the short way is to go through the Korea Strait which is broken in the middle by the Tsushima Islands and the Tsushima Strait is to the east. It's 20-25 miles wide at its narrowest point. That's the short way which is what the Russians take.

Guess what? They have a welcoming party there of all these newly refitted Japanese ships that know exactly what they're up to. The Battle of Tsushima is one of the most lopsided naval battles in human history. The Japanese just either sink or commandeer basically the entire fleet, proving the Russians had no idea of how to do naval strategy. The navy had been a product of a bunch of aristocrats. The one who ran it was a guy who was known for, what was it, fast women and slow ships? It did not work out well for them.

While the negotiations are being held on whether to hold negotiations, President Roosevelt suggested to the Japanese that they try to take Sakhalin Island. You can see the tip end of the Japanese home island, Hokkaido and then there's Sakhalin. Why? Because it's much more valuable to the Russians than the Japanese and it'd be a good tradeback item in the peace negotiations. Because for Japan it's a nice fishing ground. But for Russia it's sovereign Russian territory. Horror ceding that sovereign Russian territory to anyone, let alone an Asian power given the prejudices of the day.

To sum up what Japan got out of this war, it got its immediate war objective, Russian troop withdrawal from Manchuria. That's what they wanted. They get this Japanese sphere of influence in Korea. That's what they wanted. But remember at the very beginning they wanted to trade Japanese predominance in Korea for Russian predominance in Manchuria? Let's look at what they really got. They got the southern half of Sakhalin Island, Russian territory. They got southern Manchuria. It's actually arguably the valuable half, the southern half, because it has Port Arthur and Dalny and all the railways that Russians had invested there. Then it confirms the outcome of the First Sino-Japanese war that Japan is indeed the dominant power of Asia. So I rest my case. Japan did it. Smart decisions in Japan are what overturn the balance of power in Asia.

Sarah Paine 00:48:38

Okay, but I promised you a counterargument. What would a smart person say who disagrees with me? Here's a perfectly good counter argument. It's great you think the Japanese are so clever, but actually China faced a perfect storm of catastrophes. Because China imploded, that's why it leaves Japan on top in Asia. The primary factor is that China's falling apart and this perfect storm of problems. I'll go through three. It's always a good number. People can remember three things. You have the civil wars afflicting China, in an age of accelerating European imperialism. Also the Manchus who ran China, they're only 2% of the population, were suffering from all kinds of dynastic decline, which I will get into.

China had reached its preindustrial limits to growth. Its population kept on going up and up, but its agricultural productivity just could not feed people. So people are trying to farm really marginal lands. Either they're too vertical or they don't have reliable rainfall. You're getting massive soil erosion doing these things. You're also getting a lot of famines. Famines are both the cause and the effect of civil wars. On this map you can just see, I've named some of the big rebellions on this one. It's to give you a sense that these rebellions affect all of China. It’s not just a little here and there, but a lot everywhere.

Now I'm going to give you a table. This is a simplified table which only goes 1845-1895. It's part of a much bigger table that would cover the entire 19th century. And that table is an oversimplification. So the point is this is not business as usual. In the red box, that is the height of these rebellions, 1851-1878. The biggest rebellion there is the Taiping. It's estimated that 20 million people died in the Taiping Rebellion. To put that figure in perspective… People don't know how many people died in all these things. China didn't know how many people they had, much less how many people they lost. But to give you a sense of it, in World War II it's estimated that 55 million people died. So you're talking 20 million just in the Taiping. I have no idea how all this adds up, but it's huge.

The Chinese like to talk about these as being rebellions or uprisings. Give me a break. They're civil wars. A whole bunch of them want to overthrow the dynasty in Beijing. A whole other set of them want to secede from the empire. Often these minority people who just want the Han, the predominant people, to go away, or the Manchus in this era. Some of these provinces are devastated for generations.

So that's point one, these civil wars. Point two is that this coincided with an era of accelerating European imperialism, where Europeans and also the Japanese are carving out massive spheres of influence for themselves. The Chinese are not going to have full sovereignty over their country for a very, very long time. The story gets even worse. This happens to China because China loses a succession of regional wars. It loses the First Opium War, the Second Opium War, the Japanese snag the Ryukyu Islands. Then in the Sino-French War, China loses control over Indochina. Then in the Sino-Japanese War, they're losing their tributary of Korea. The Ili crisis is up in Xinjiang, which goes a little better for them. This is a mess. This is not business as usual. Think about one country being afflicted by this much trouble.

Then it comes at a time when the Manchus are in real trouble themselves. They had ridden to power on these fabulous cavalries and tremendous operational success for a people who are only 2% of the population that they dominate. They transform China into the richest polity around the globe by the 18th century. But here's the problem. With their tremendous success in their military campaign. They're overextending China financially. Inside the round circle there, that's China proper. These are the core provinces of China that produce all the income. The Manchus came in from Manchuria up there, but they also take Mongolia, Xinjiang, and Tibet. Not Manchuria so much, but the other ones, they cost a lot of garrisoning. It costs a lot of money to garrison these places. It looks like a great success, but it's going to have problems later on.

Then here's what happens to the Manchu emperors. They no longer ride at the head of their armies. The Qianlong Emperor, who's considered to have ruled at the height of the dynasty, what's he spending his money on? All sorts of interesting architectural programs. That's why people think he's great, he builds beautiful things. But these emperors are isolated. They don't have command experience, expensive spending habits.

Then it gets better. So they got all these tributaries. Some of their tributaries produce really high quality opium. The emperors decide they want to sample the goods, and the people who are preparing it for them also want to sample the goods. Then the Manchu banner forces that are the Praetorian Guard of the Manchus are also sampling the goods. Then you go, why are the last Qing emperors incapable of producing offspring or have hardly any? They have huge harems, so what's their problem? Well, if you're really totally high on drugs, you can ruin yourself. So that is what goes on. No wonder they're losing these wars, if some of your key thinkers are not thinking very clearly.

So the counter argument to my argument says, “Look, it's nonsense. The Japanese may have been really clever, but the real big event is this massive series of civil wars, European imperialism, and the collapse of Manchu minority rule. That would be the real thing.”

Sarah Paine 00:54:45

Now, if I leave my argument here, I've just shot myself in the foot because I've just decimated my original argument with a pretty good counterargument. That would be bad. That would hurt. So you need to do a rebuttal. It's really good to come from an unexpected direction. So I'm going to try my direction, which is going to be Russia. I’d say, “Look, Russia is the catalyst. It's taking advantage of the collapse of China, and it's going to catalyze things in a way that Japan is going to leverage. It's going to be Japan leveraging these things. That proves that smart decisions in Tokyo answers the question.”

So what's going on with Russia? Russia has always tried to expand its territory. It is today the largest country on the planet. Why they need more territory, nobody knows. But that's what they're after. And these days what they're taking are those two large areas. By the time the Bolsheviks finish working their magic, Russians have taken from the Chinese sphere of influence —the Bolsheviks are beyond today, but it gives you a sense— more territory than the United States east of the Mississippi. That's a lot of territory. But in this day, while the Opium Wars are going on and while the Taiping and Nian Rebellions, the two big ones, are happening simultaneously, the Russians go to the Qing and say, “Hey, we'll deal with those foreigners for you. We'll be an intermediary, but just sign on the dotted line for these treaties giving us all this land.”

The Chinese were vague on geography and they didn't believe in treaties. They thought they'd get it back later. So Russia gets all this stuff, does exactly nothing because we know about the treaty port system. The British and French got all of it. Russia got this stuff too. Once Russia gets this stuff, that's why they want the railway, because they want to integrate it there. So the Trans-Siberian Railway is the blue thing, but understand that the blue part that's above the green part, that wasn't built until World War I. Russia built the green part, which is the Chinese Eastern Railway, the name of it, in order to stake out its plans for empire.

Also notice it wants the warm water port, that's the orange part going down to Port Arthur, because Vladivostok, prior to the age of icebreakers, is frozen solid several months of the year. So Russia's building this railway. To save on construction costs, it's more direct going straight across Manchuria. Contain Japan, stake out its claim to Manchuria. The Japanese get it. Well, that was earlier in this lecture. They see this coming and they don't like it.

Then in addition to what's going on, the Chinese don't particularly like all the imperialists messing with their country. And so the Boxers are another rebellion. These folks want all Westerners out of China and they want to kill any stragglers. They're all over China, they go into Manchuria. The Russians send 100,000 troops, far more than any of the European powers or even Japan which sends a lot, to the rest of China to defeat the Boxers, which they do. Then the other Westerners in Japan remove their troops, Russia doesn't. This is the thing that gets Japan's attention, because Russia's got 100,000 troops in Manchuria that won't leave. That's when Japan is sick of Russia procrastinating and we're going to wind up getting into the Russo-Japanese War.

So guess what? We're back to my original explanation, right? The Japanese finessed all of this and they wound up on top. It's by being very thoughtful in their grand strategy that they reverse the balance of power. I'm going to end this thing with General Nogi and a poem. General Nogi is the one who oversaw the four assaults on Port Arthur. He lost both of his sons in this war. One in the Battle of Nanshan, which is down the Liaodong Peninsula, and then his other, his favorite, under his command in Port Arthur.

When the war was over, he asked the Meiji Emperor if he could commit ritual suicide, and the Emperor said no. So when the Meiji Emperor died, he and his wife did. This poem helps explain why:

“Imperial troops, a million strong,

conquered an arrogant enemy;

But siege and field warfare left

a mountain of corpses.

Ashamed,what face can

I show to old parents?

How many men have returned

this day of triumphal song?”

Wars bring much sorrow. I leave you with that. Thank you very much for your attention.

Dwarkesh Patel 00:59:31

I loved this lecture and this topic especially, because it's about a conflict that many of us weren't educated on. But it adds context to the other ones which are obviously super famous, like World War II. It adds context to the development of these powers, which are the most powerful in Asia.

I'm not going to keep wasting time. I just want to jump into asking you a bunch of questions about this. One way you can explain Russia's loss here is that Japan had better tactics. Given the fact that they had deployed similar amounts of men to the front, Japan with their better tactics, was able to win in these different battles.

But maybe the more important thing to explain is why Russia didn't deploy more men or didn't deploy its greater resources. Russia's population is over 130 million people at this point. Japan's is 47 million or so. You might say that the Trans-Siberian Railway isn't completed by this point. But in 1905, it actually is complete. So Russia could have continued the war. It has all these fresh troops it could send. Why didn't they have the same will, or the same ability, to muster their massive resources, that Japan brought to bear?

Sarah Paine 01:00:44

One is that the railway is a bottleneck. Japan's starting the war deliberately while that bottleneck is bottled. That's one factor. Russia has massive troops. Also, the Russians are arrogant. They think Japanese… Initially, the Russians think they're going to do regime change in Tokyo. That's their plan. How they plan to get to Tokyo is a mystery, but never mind. So they don't realize what they need. It's a fairly common problem in wars, a gross underestimation of the other side. That's what's going on in Ukraine right now. Putin missed that one. He's not the only one. Hitler underestimated the Russians in World War II.

Then you go, “What's going on with Japan?” With those Meiji reforms, they have an educated population. Russian soldiers are a bunch of illiterates. And the Russian soldiers are going, “What? Why are we fighting in Manchuria? Tell me precisely why I'm here getting killed.” The Russian soldiers had no buy-in. You can ask what exactly did the Russians want there. Apparently, there were some royal favorites who thought that they were going to get some lucrative timber concessions on the Korean border. Okay, let's go into the logistics. I've never heard there was a shortage of wood in European Russia, ever. And they're going to ship it all the way back? This is a non-performing economic model.

In addition, Nicholas II is an incompetent tsar. He's hiring the guy who's running the navy who is also incompetent. Their navy gets no training whatsoever. Great, you have expensive ships, but no one's trained in how to use them, whereas the Japanese have.

For the Japanese, it's the value of the object, one of these concepts. How valuable is it to win for Japan versus Russia? For Russia, it's already the biggest country on the planet and there's nothing really that exciting out in Asia from their point of view. Maybe for Nicholas II, because he wants to be a great tsar and add to the empire, that would be a nice thing. But when the Germans come after him in World War I, I think he's going to care about the other things more.

For the Japanese, rightly or wrongly, their government and educated people consider it existential. They're looking at China going, that's our future if we don't fix things. They're thinking the solution for Japan is empire, because in those days that seemed to be the way powerful countries ran things.

It's a combination of legitimate logistical bottlenecks. Japan, since they're going to be starting this thing, has massive preparations. The Russians aren't planning to fight this thing. They're thinking, “I determine whether wars begin or end.” Excuse me, they don't.

Dwarkesh Patel 01:03:52

So I guess you're saying Russia doesn't care about this conflict as much as…

Sarah Paine 01:03:56

Russians. Tsar Nicholas II might, but his population is fighting it.

Dwarkesh Patel 01:04:00

So then he doesn't have the state capacity to mobilize his…

Sarah Paine

That's the whole problem with Russia. This is the power of institutions. Japan clearly can mobilize. Boy, is it doing well with the loans and all sorts of things. It's using these institutions. Whereas with Russia, remember, it doesn't have a legislature. The Tsar doesn't have a cabinet in the same sense. He has ministers, but they don't ever show up at his house at the same time to sit around a table. He's just doing them one at a time.

Then the Romanov family has lots of first cousins. Everybody's having lots of kids, and so you've got a million first cousins. They are all deployed throughout the ministries, basically being the spy system for the Romanov family for what's going on. Are any of these nice rich boys particularly competent? No.

Dwarkesh Patel 01:04:53

I want to understand, in World War I, they were able to mobilize many more men. Obviously there was incompetence in World War I as well, but the sheer amount of resources the Tsar was able to bring to the Far East during this conflict…

Sarah Paine 01:05:07

Look at a map and you'll see the railway grid is much more extensive in European Russia. This is where Count Sergei Witte, their finest minister of late tsarist period, he's the one who's trying to do his version of a Meiji Restoration in his purview, which is the Finance Ministry, financing all these railways. So there are many more railways. This is where Russia's population is. This is where their historical security threats come from. Russians can get on board with protecting European borders.

Dwarkesh Patel 01:05:38

Once the Trans-Siberian Railway is finished in 1905…

Sarah Paine 01:05:42

Just the part in Manchuria. The one that goes on their side of the boundaries not completed until World War I.

Dwarkesh Patel 01:07:09

I want to double click on what part of the Meiji reformations exactly helped Japan mobilize for this war. Okay, the population is literate. You have institutions for having political say. How exactly does that help you have more people go to the front?

Sarah Paine 01:07:27

I don't know the answer, but the Japanese… Talk about a culture that's about detail, about hard work, about service to the emperor. I've done in your earlier podcast series a whole series about bushido. They're imbued with service to the emperor and if you go fight wars, you win them. That's part of their culture.

Russians also don't like being messed with when people invade them, but this time they're invading other people in an irrelevant part of the empire. It's at a time when Russia is trying to industrialize itself, but they haven't put enough money into their education system. So they have a bunch of illiterate troops running around. Let's face it, how can you read any of the manuals or do anything?

Some of their generals, one of them didn't know what a howitzer was. One of the guys who's planning stuff, he can't read maps. The level of incompetence of having royal favorites in your officer corps, they make these decisions. Also, there's split command in the Russo-Japanese War. There's General Kuropatkin, who is the professional who actually had fought against the Ottomans. But those are colonial wars, and he'd won some of those. He's a professional. But then Admiral Alekseyev—who's the illegitimate son of, I can't remember which Tsar, and a favorite uncle, I believe, of Nicholas II—he's out there and no one knows who's actually in command, except the royal favorite is probably the better bet.

So Kuropatkin wants to not engage the Japanese until they're way inland, because he wants to extend their lines and then clobber them. All these aristocrats who don't know what they're talking about say, “No, no, no, no, no, we're not going to let these racial inferiors do whatever. We're going to take them on immediately.” Okay, try that. Kuropatkin's strategy probably would have worked. Just bring them inland and let them enjoy it, like with Napoleon Bonaparte. You want to get to Moscow, have at it. Then try winter in Moscow.

Dwarkesh Patel 01:09:35

One really interesting takeaway from this lecture—and I'm curious if you agree with this—is that when we think about this period in history, we often think of Japan as the rising power in Asia. But it seems like the takeaway here is Russia was actually on the path to be the dominant power here. Their population is so much bigger. If the efforts to modernize had worked, they would have had a 4x bigger population. Then they could have had the modern industry Japan has.

Were it not, later on obviously, for the Bolshevik Revolution and the Civil War and collectivization, Russia would be dominant. Japan, seeing this, acts early. I guess I didn't put it in those terms before I heard this lecture. Would you agree with that?

Sarah Paine 01:10:18

I'm going to give you a tangential answer, then you tell me what I'm missing. The sorrow of Russia is that they have this incredible land, natural resources, and the works. They're really hung up on empire, having lots of territory. They live in a difficult neighborhood. Witte is telling Nicholas, forget about the Asian adventures. We need to build more railways in European Russia. He was right. Because if they could have mobilized better in World War I, you might get some different outcomes. Also, you might have avoided the food riots in St. Petersburg, which toppled the Romanov dynasty. If you'd actually provided enough food to St. Petersburg, maybe you don't have those riots because you have the railways running and things.

So Russia is a story of misinvestment. Look at it today. If Putin had taken all of his oil money and invested it in the transportation grid… Believe it or not, Russia still just does things by train. Its road systems are appallingly bad. If Russia actually cleaned up its legal system, which under the tsars it hadn't done either, everything was always royal monopolies… If they had cleaned up these things to allow individuals to make their own investments and make money, it'd be a completely different story. But tsars couldn't stand merchants. Communists hated them even more. There are cultural differences. The Russians aren't known for being meticulous and hard working the way Japanese are. Russians are known to be hard drinkers. They're also known for being brutal.

Dwarkesh Patel 01:11:56

Let me take another different stab at your thesis.

Sarah Paine 01:11:59

Go for it.

Dwarkesh Patel 01:12:01

It's not Japan's strategic brilliance, which mainly explains why they won the war, but rather the fact that Russia could have kept going even after the Baltic Fleet was sunk, even after they've lost in Mukden. But what changes is in 1905 you have this massive peasant revolution. You have mass strikes in the cities. You have Bloody Sunday in January. So the Tsar’s government almost collapses to the point that he has to at least give the perception of instituting major reforms.

Sarah Paine 01:12:33

Unclear whether it was going to collapse.

Dwarkesh Patel 01:12:35

This is maybe a tangent but when I had Stephen Kotkin, the Stalin biographer, on…

Sarah Paine 01:12:39

He might know more about it..

Dwarkesh Patel 01:12:41

So Japan cannot claim credit for the fact that there were these long-standing issues in Russia which almost brought down the government in 1905. This was a lucky break that they got which forced Russia and the Tsar to reallocate their efforts from the fight.

Sarah Paine 01:13:01

That's a good argument. He might be right. We'll never get definitive answers because you'd have to be in Nicholas II's brain. It's an autocracy. What he decides matters. I dare say if he had been Peter the Great, the guy who tortured his only son to death and suffered numerous setbacks, I doubt that he would have had any problems dealing with rioting Russians. He would have figured out who are loyal units and then just blown away whatever's getting in his way. He would have done his next battle and then when he won that battle the Russian Revolution would have gone away because they would have gone, “See we did it. We got these people.”

So Kotkin's argument is a very good one. But to me, Nicholas II’s incompetent leadership is the bigger thing. Then you're arguing about how important the rebellion is and how deep rooted it is and you'll never know. Kotkin could be absolutely right. One of the reasons I gave the counterargument is to get used to understanding ambiguity and appreciating that other people could be right. You may not know.

Dwarkesh Patel 01:14:20

To be clear, I don't want to put words in Kotkin’s mouth so I'll just put them in my mouth. I won't blame him for any errors I make.

Another thing that is super interesting in this period is that it potentially helps explain why Japan thought that Pearl Harbor might work. Superficially these two situations seem similar. In both cases you begin with this imminent threat that an adversary of yours is going to gain massive military leverage over you. In the 1904 war with Russia, it’s the fact that the Trans-Siberian Railway would be finished. In World War II, it's the worry that after Japan invades French Indochina, America would do an oil embargo on Japan, giving them one to two years of oil runway before their entire empires collapses and grinds to halt without oil.

There's this worry that they have to act now, otherwise they're about to lose their leverage. Then there's this initial surprise attack, of Port Arthur in 1904, and then obviously Pearl Harbor. Even if you look at other battles like Midway and Mukden, we'll have this sort of annihilating battle where we will expend our blood and just push them back. I’m curious to get your take on this interpretation.

Sarah Paine 01:15:38

In both cases, Japan is seeing a window of opportunity. It has to operate in this window and get it over with. If you're doing that, that's a high risk strategy to get away with it. This goes back to the "they were just lucky" argument.

Another thing is that, you can look at why certain wars turn out. Another concept is a cooperative adversary. What's a cooperative adversary? It's not one that wants to cooperate with you, but it's one that doesn't play its cards remotely well. If you think about literally playing cards, if you're playing a game with a little child, they're trying to win. But they're a cooperative adversary because they just don't know how to play whatever the game is. You could argue that Nicholas II is a cooperative adversary, and that's required. The United States is not remotely a cooperative adversary, and it goes really badly for Japan.

Also, there’s another piece. When you run one successful war, quite often you think it was easier than it actually was and less risky than it actually was. This country does Gulf I where the allies pay for the whole thing. No Americans die, hardly at all, loads of Iraqis die. They're out of Kuwait within days. Wow. Then people are complaining we should have marched on to Baghdad. We did that. That didn't work out so well. But after you do one of these things, you think you're good to go.

In Japan's case, between the Sino-Japanese War and the Russo-Japanese War, it shifted the balance of civil and military power in Japan. People thought that it was the military officers that did the right thing and that the diplomats lost the peace. In the Russo-Japanese War, they're saying that they didn't get a big indemnity as they did in the last war. You're getting the whole Russian railway system. That's an indemnity in kind, but that's not what they're seeing. It shifts the balance of power.

In addition, this Meiji generation, the prime civil leader, Itō Hirobumi, is assassinated by a Korean revolutionary, because the Koreans don't much like the Japanese in their empire. He dies. It's a decade-plus before Yamagata does. That gives Yamagata and military friends another decade to insinuate the roots of military institutions, and civil institutions are not being built.

Then you can go, “Look at the Japanese, they're incompetent because their institution building was inadequate.” Well in one generation, you can't do it all. Institutions take generations to really sink roots. They did as much as they could in a generation, but their leadership was bereft compared to… This is why the Meiji generation is truly a brilliant set of leaders, and what comes after, their children and grandchildren, are not.

Dwarkesh Patel 01:18:41

This is also another very important point. The fact that the civilian leader of the Meiji generation dies before the military leader reinforces the contingency of history. It didn't have to be that he died first, but then that sort of changes potentially the whole trajectory of history.

Sarah Paine 01:18:57

Individuals, trends, the whole thing matters. There's no saying. Maybe if he'd stuck around, everything would have been the same. Since you can't rerun the experiment, this is where, again, counterarguments are useful. On so many human endeavors, you just don't have complete data. When you're doing things in the here and now, maybe the data will come in a little later, or it just doesn't exist. Yet you have to make decisions.

If you're not willing to change your mind, you're just going to double down on bad decisions. It's incredible the number of people who don't want to change their minds. That's why I made a big to-do about giving the argument-counterargument-rebuttal framework, because it helps get you out of the rut so that you can change your mind. You're going to tell me that Kotkin's got this thing and I'm wrong. I'm game. Maybe he's right. I don't care who's right. I care what's right.

Dwarkesh Patel 01:19:51

Going back to the question of what was different between the Russo-Japanese War and World War II, and what Japan's miscalculation was. One of them has to be that, in the case of Russia, they attacked an adversary where it took one year of conflict to push them to the point of significant destabilization of the government. Whereas, if you thought the same thing would happen with America… “We're going to attack Pearl Harbor and there's going to be all kinds of internal conflict.” They just misjudged the internal level of cohesion and functionality of American institutions. They thought the same thing that happened in Russia would happen.

Sarah Paine 01:20:29

They're missing the whole war. Their war is with China. Instead of isolating the adversary, they bring in a whole new slew of adversaries. They're having trouble with China. That's the Second Sino-Japanese War. We get all excited when we're involved, but it starts in 1931 and their objective soon becomes an unlimited objective. They eventually want to do regime change in China.

Whereas in the First Sino-Japanese War, Yamagata is thinking about doing regime change and the government pulls him out because it's, “Buddy, don't do that. We're going to have all kinds of foreign powers intervening if you try.” In the Russo-Japanese War, they're not trying to do regime change in Russia. But in the Second Sino-Japanese War, they absolutely are trying to do regime change. They get more and more frustrated.

Their solution to that one is to attack all the colonial interests in Asia. That's when Pearl Harbor happens. Then they're bringing in us, the British, and that brings in the British Commonwealth. Australia and New Zealand do significant fighting, and then the Netherlands. It's a different event.

Dwarkesh Patel 01:21:31

I want to go back to the Meiji reforms and the Iwakura Mission, which was sent to Prussia to understand what Bismarck did right. I think it's funny because Sergei Witte, who's the finance minister from 1892 to 1902, models himself as Russia's Bismarck. He wants to do the kind of industrialization that Germany went through, in Russia. Yet the Meiji generation succeeds in Japan in doing this kind of thing.

In both cases, there's an authoritarian government that wants to have this kind of top-down reform. Why does it work in Japan but not in Russia?

Sarah Paine 01:22:07

Russia just has a phenomenal illiteracy rate. It's an empire, which gets in the way. If you look at how many non-Russian people they're trying to dominate, they're perennially overextended. They're perennially occupying places that don't actually make them much money. In fact, they probably cost them more money to hang on to. They’re perennially doing that.

I believe the Japanese are known to be incredibly hardworking. I believe the Chinese are known to be very hardworking. I have never heard anyone say the Russians are known to be hardworking. Maybe I missed something. I've heard they're like artists and they have the russkaya dusha, their soul. Okay, great. That's like a big excuse for saying, “We may fail at everything, but our soul is superior.” There are these other things that are at work.

With Germany, it's the center of Europe. It's a highly developed place, and it already has high levels of literacy. There's a lot more to work with in Germany. It already has better infrastructure. For Russia, it's huge when you're trying to build railway systems, just enormous. Think about the Trans-Siberian. If you just took all that track, just length, and put it all over European Russia, you'd really have something there.

Dwarkesh Patel 01:23:31

This is a quote from Sun Yat-sen, who was the first and short-lived leader of the Chinese Republic in 1912. He says, “We regarded the Russian defeat by Japan as a defeat of the West by the East. We regarded the Japanese victory as our own victory.”

Now, we know that just 10 years prior, China and Japan had fought a war together. We know that in 20 years, there's going to be a brutal invasion of China by Japan where millions of Chinese will die. Help me understand why during this period, at least to the Chinese revolutionaries and reformers, it seems like Japan is the power to emulate.

Sarah Paine 01:24:13

First of all, he's the founding father of modern China. It's one of the few things that the Taiwanese and the PRC agree on, that he's the founding father of modern China and it's positive what they think about him. Remember when I told you in the Russo-Japanese War how they're paying for everything in Manchuria, that the Chinese actually thought that it would be a good thing.

Also understand that when the Manchus lose… In Chinese, the war is called the Qing-Japanese War, the Manchu-Japanese War. The Japanese are defeating the Manchus, who you may not particularly like if you're a Han nationalist. You can relate to that. The Japanese got in and out of that thing. That's the key. The really, really, ugly, brutal stuff is going to happen in the Second Sino-Japanese War where the Japanese do things that are unforgivable. The problem for the First Sino-Japanese War is that it's detonating Confucianism. What's it going to replace that with? Sun Yat-sen is going to do nationalism and work it that way.

I think that's the big answer, but you're quite right that as things went on, as it sinks in that Japan is the dominant player in Asia… Because first you have a war and that's happened, but you're focusing on all your civil wars and other stuff. It hasn't sunk in that this is not an exception. It's going to go on for a century. I suspect that it's gradual for the Chinese to get really upset about Japanese in particular. It's what they did in the Second Sino-Japanese War. It's unbelievable, because when they start losing, they get brutal.

Dwarkesh Patel 01:26:11

Something that wasn't emphasized in the lecture, but it seems quite important in explaining why Japan was able to be so dominant in this period, is that despite the fact that it has a much smaller population. China at this time is 400-450 million people. Russia is 130 million. What's going on in Japan? The Meiji reforms, what specifically are they enabling?

If you look at the fraction of GDP that China is spending on defense… The country sort of imploded, so it's less than 1%. Russia is also not spending nearly as much of its output on defense. Japan in the preceding 10 years is spending 5% of output on defense. In 1904 it’s spending 10% of output on defense. That ability to marshal… It's a smaller place, but you can marshal the whole of resources towards this defense effort. It seems very important to understand.

Sarah Paine 01:27:07

It is, but there's another piece. China is not really a unified state in this period. If you think about what people's loyalties are, it's to province and place. It's not to the whole of China. They're all part of it, but it's the place.

I'll give you an example. China has three fleets. The one that I've been talking about is the northern fleet. In the Sino-French War, the southern Fujian Fleet that’s fighting in Indochina in those days, wanted the northern fleet to come down and help them. The northern fleet says, “Hell no.” In the Sino-Japanese War, when the northern fleet wanted the southern fleet to come help it, the answer was, “Hell no.”

It also has to do with how these forces are being financed. Provincial governors are financing it. They want to use these forces for the province. They don't want to send it out of province. In the traditional world where China is just worrying about some barbarians on horses way off and wherever, that works well enough. Because if they come riding in, you can probably eventually deal with it in Province 1 and it'll eventually work. But all of a sudden, these Western powers come and then the Chinese very foolishly fight in the treaty ports. What they should have done is said, “Hey, come on in 25 miles and see how you like it. Alright, let's try 50.” The British would have had real trouble with that. It's a real failure of strategy that they can just own.

They love to complain, the Chinese today, about foreign imperialists. They don't look at their own gross incompetence. This country we're dealing with, Britain's a lot closer to us than it was to China. Their population was larger than the American colonists. George Washington and friends gave them a run for their money. The Chinese, what's their problem? They should have been able to deal with it.

It's a case of lacking leadership, lacking institutions. These things matter. Even if you have all the resources in the world… So this country needs to watch it. We may have all the resources in the world, but if we don't organize ourselves appropriately, or if we do a whole bunch of own goals with our own leadership, it will cause problems.

Dwarkesh Patel 01:29:33

Talk about the long-run institutional effects of history. You just mentioned how the provincial governments controlled a large share of Chinese resources. You go from that period under the Qing Dynasty to the Republic, then you have the Warlord Era, then you have the Japanese in control of parts of the country, and then you have the Communists or Nationalists, and then Communists. But after five different regime changes, you have this dynamic in China.

Today, in the US 50% of government spending is done at the federal level, 50% is done at the state level. In China, it's 85% at the local level, 15% at the national level. A big problem recently has been the fact that the local government has financed a lot of construction that wasn't economically valuable. But you can go back more than 100 years to this period and the same problem of provincial governance still comes up again.

Sarah Paine 01:30:33

Under the Qing, local areas had to finance all of the government. The central government financed hardly anything. That was always a problem and led to massive corruption. Because if you don't have a budget, then you have to have some way of getting graft or something from people to pay for things. This is enduring and it's a different institutional setup of how you're going to run things. Institutions matter. They channel decision making. They channel how money can or can't be made. It's consequential.

Dwarkesh Patel 01:31:13

Part of the reason that Russia loses is that this is not the main priority for them. But, if that's the case, before the war, Japan produced an agreement which would allow Russia to maintain their sphere of control over Manchuria and Japan to maintain their sphere of control over Korea. Russia refuses this and that's why Japan decides to do the surprise attack. Obviously, after the war, Russia is in a much more disadvantageous position. Do we understand why they were not game to consider this before the war?

Sarah Paine 01:31:52

It's a gross underestimation of the enemy. They're thinking these people are a bunch of nobodies. Nicholas II had an unfortunate incident in Japan. Who knows whether this influenced him or not. He does the royal tour while his dad's still alive. He goes all the way around, winds up in Japan, and a samurai there or somebody tries to assassinate him with a sword. Nicholas always has a scar up here. I imagine that didn't leave happy memories for the Tsarevich.

Dwarkesh Patel 01:32:28

It is a pretty cool story, though. “I almost got assassinated by a samurai and here's my scar.”

Sarah Paine 01:32:32

Yes, whatever. Anyway, that's what happened to him. It's the feeling that there's no need. It goes back to grossly underestimating other people and the problem with racism. You think you're so special and other people are so unspecial. Wrong.

And the Japanese are amazing. Think about it. They're in the Group of Seven of major economies. Its major economies and major democracies. They're the only non-Western power there. They don't get there out of charity. It's out of their own achievements. It's an enormous achievement.

What's tragic about World War II is if they hadn't done that, people would be looking at the Meiji reforms in order to model what they're going to do to get over the hump and develop. In fact, a lot of the things that Deng Xiaoping does in terms of investments and things rhymes with the Meiji reforms.

If you look at the four little tigers, originally Singapore, Hong Kong, Korea, and Taiwan, a lot of their development parallels a lot of what the Meiji reformers did. These are Japanese ideas, but they're not getting credit for them because of the brutality of World War II. You're thinking of that and thinking, “If I say nice things about Japan, I'm endorsing this other stuff.” No, they're two separate topics, different generations.

Dwarkesh Patel 01:34:03

I want to better understand the degree to which the level of technology in a country matters for its military capabilities. At this time, countries are just buying armaments from more advanced countries. In the 1894 war, China bought two German battleships and many other vessels. Japan has no battleships. In the 1904-1905 war with Russia, both Japan and Russia had British ships because of the alliance that Japan had with Britain. Russia also had French ships. So if they're able to procure the most advanced weaponry from the West, does it matter what their level of technological development is?

Sarah Paine 01:34:50

You have to learn how to use it, and you have to maintain it. The Russians had this fancy fleet that their aristocrats wanted to buy. So they're good at cutting checks, but they aren't doing any training on it. If you look at the number of days at sea for the Russians, it's laughable. It's cold, so all winter, no one's going anywhere. Whereas the Japanese are constantly deploying their forces and they're learning, doing all kinds of practice and things. You can imagine if you don't do any practice, it's not going to go very well for you.

The counterargument to this technology is that if the value of the object is high, and if you're an intervening power messing in somebody else's country, it may not matter on the technology. Or it matters, but these people will never give up. In the American Revolution, the British had us on technology. It's incredible. They actually could supply, in an age of sail and horses, things by sailboats and keep it going. But we were just totally obnoxious and we would never give up.

This is the problem with the United States in Afghanistan. Talk about a backwards place. It doesn't get any more backwards. Or in Iraq, the locals just never gave up. They may not have the technology, but “It's our house and we want you out of here. We're never quitting.” This would be Russia's problem in Ukraine. It’s going to go on.

Technology is important, but so is the value of the object. What is the relative value of whatever you think you win is? If you're an Iraqi, it's a lot higher for you than it is for some American intervening there. If you're looking for Japan and the empire, it's a lot higher for Japanese who are looking at China and thinking that's going to be their fate. Maybe they're misidentifying the cure, that's a whole different matter. But that's how they look at it. For Russia, it’s, “Who even cares?” There's nothing this part of the world actually offers Russia other than more lines on a map that looks bigger.

Dwarkesh Patel 01:37:05

Speaking of winning, after the war, as you mentioned, Theodore Roosevelt mediates the peace treaty between Russia and Japan, and he wins the Nobel Peace Prize for this.

Sarah Paine 01:37:16

At least one president gets one.

Dwarkesh Patel 01:37:18

The Treaty of Portsmouth is heavily criticized in Japan in the following years for not giving them the concessions they feel are warranted as a result of their victory. I don't think there are any reparations. There are many territorial disputes that don't end up landing in Japan's corner. Potentially, a lesson to learn here is that if you give too relaxed peace terms, then the person who wins will feel aggrieved and they'll go at it again.

Although for the other lesson, you have the Treaty of Versailles, where it seems like you should learn the opposite lesson. If you have too aggressive peace terms, then the person who lost will come back and have this sense of being aggrieved. This is a general problem in history. It feels like there's an equal and opposite lesson.

Sarah Paine 01:38:12

They’re different types of wars. One's a war for unlimited objectives, which is what World War I eventually becomes. We were trying to do regime change all over the place. Whereas you're doing a limited war for Japan. You also have to have a feasible objective. If Japan had demanded more at those peace talks, Nicholas II would have gone back to war and he would have slaughtered them because he had all these crack troops that were just sitting there in the theater. The Japanese literally don't have the men. Russia's got three times the army now. It's there, and they can bring them in at 100,000 a month. We just saw Mukden where the Russians had 125,000 more men than the Japanese. If they can ship at 100,000 a month, wait a few months and it's going to get really ugly.

It wasn't feasible. There's another piece. You're going to go fight a war and then you lose a lot of people. That's why the Japanese are feeling they should have gotten more. You've got General Nogi who's committing suicide over this. The poem at the very beginning I showed you, the Meiji Emperor also has this feeling of sorrow. With wars you start losing. This is Putin's problem. He's lost so many people. He didn't feel sorrow, but he needs to get revenge that's equivalent to make up for that. The longer these wars go on, the harder it is to stop them.

On Versailles, it's a different story because there is an unwillingness to enforce the peace. There is also a feeling, particularly in France and also Britain, to get really even with the Germans. The United States was an irresponsible power in those days. One of the reasons you get World War II has to do with the Great Depression, which is not strictly related to the Versailles peace terms terminating the war. If you hadn't had the Great Depression, maybe the Versailles settlement would have been good enough. We'll never know. One of the reasons the Great Depression gets so bad is because this country doesn't want to act as the lender of last resort, which is what Britain had done in the past but couldn't after World War I. It had been bankrupted.

Our country's irresponsible. These problems are on the other side of the sea. We think, "Okay, America first and whatever," and we're going to raise tariffs and things, and all we do is tank the global system. Whereas if you're the lender of last resort, you're going to put money into the system as it's collapsing to prevent it from collapsing so that it can rebound. That's why it's important to do that. It's expensive, but boy, a Great Depression is a lot more expensive. There are different problems.

Dwarkesh Patel 01:41:05

Is there a lender of last resort today?

Sarah Paine 01:41:07

Isn't that the problem? That's what the IMF and things are about. But it used to be that this country was willing to take on these burdens of last resort in the area of nuclear weapons, a nuclear shield. We're trying to shed these things right now. We shall see. At bare minimum, it'll make us an irrelevant power instead of an essential power because Europe is going to reorganize itself. They have the institutions to do it. The United States used to be the leader of the West. It's going back home to Europe. Europe's going to lead the West. They invented it. They have an extensive set of institutions to do this. They're already starting to marginalize us by putting out feelers to set up trade agreements with Asian countries, with South American countries. It'll take them a while to do this. Once they do it, we will be on the outside of it if we don't change our minds. They're going to get stronger institutionally.

What is it Churchill said something about allies? "The only thing worse than working with allies is trying to survive without them." You're better off with lots of friends going after bullies in the world than you are by yourselves. We'll find ourselves by ourselves. Other countries have tried to tangle with China all by themselves. Japan did it when China was an undeveloped country. It did not work well. You want to have friends when you do this, but we're making pivotal errors and we will pay for them.

Dwarkesh Patel 01:42:51

Maybe this is a subtle but recurring disagreement between us that has come up in different Q&As, but I do want to find the crux.

Sarah Paine 01:43:00

That’s fine. I could be wrong

Dwarkesh Patel 01:43:05

It seems like the quality of institutions is central in your thinking of how well a country will do or how central it will be in the future. For me, it's much more, "How big is your economy? How big is your population?" If you think that the center of the Western alliance shifts to Europe, maybe you think their institutions are good and that's why it'll happen. But it seems to me America is just such a big and powerful country economically that it would be difficult for Europe to displace us, even if we make these mistakes in terms of our alliances and diplomacy.

Sarah Paine 01:43:44

Let me ask you a question. Why isn't Argentina, which is a really big country with tremendous resources, a great power?

Dwarkesh Patel 01:43:51

Its economy sucks. I think, if its economy was better, I think it would be better.

Sarah Paine 01:43:54

But why does its economy suck? Excuse me, not my choice of words.

Dwarkesh Patel 01:43:58

But that’s a result of mistakes in domestic policy with Peronism, not with its mistakes in foreign engagement. I agree that domestic policies matter a lot.

Sarah Paine 01:44:10

Okay, but Peron is a populist leader. Populists, by definition, aren't people who are institutionalists. They're like a "Me, I'm so special" kind of person. Aren't you actually telling me that Argentina has inadequate domestic institutions and they failed?

Dwarkesh Patel 01:44:26

Agreed. This is great, because then we're identifying the real crux. Maybe the crux is we both agree that institutions matter. I think they really matter because they shape the trajectory of economic growth.

Sarah Paine 01:44:44

Look at Russia. Lousy institutions. That place should be rich.

Dwarkesh Patel 01:44:48

So I think Europe's institutions are still bad because the rate of economic growth in Europe is quite bad. Whereas you think they might be better than I think they are, because they're good at diplomacy.

Sarah Paine 01:44:58

I suspect that this crisis now, because it's existential with Putin, will force them to clean up these institutions. It will take them time. Maybe I'll be dead wrong… or maybe I'll be wrong, dead. Who knows how that all works? But I suspect, because usually an existential threat is clarifying. You can see Ukraine trying desperately to clean up its institutions. It has a lot further to go than Europe does. I suspect that Europeans are going to clean up some of this. It'll take them time, but they have to, because the piggy bank here is no longer willing to fund their defense.

So there’s merit. The Europeans need to pick up the tab on their defense, and they should have done it earlier. Without a doubt, there's merit to that. But the way we're doing it, we're going to cut ourselves out of the action. You want to put the screws on, but you also want to keep yourself part of it. We're overdoing it on some fronts. Also, you're doing it in such an insulting way.

I'll give you some Machiavelli on this one. He's the founding father of political science. He talked about motivating people with fear, love, and hate. He's talking about it in The Prince. He says, "Forget about love because you can't make people like you. You can do everything and people still don't like you." He's a big man in saying, "Make them fear you because you can really motivate people," but he says, "Don't make them hate you. If they hate you, they will fight you."

If you're running around serially spitting in your allies' face, they are going to hate you and it won't work out well. It's a bad strategy. Don't do it. The clickbait of insulting someone who comes to your house, that's very temporary, but the hatred of it, and also other people viewing it and going, "This is what qualifies as American diplomacy?" Those memories are long-term. It's unwise.

Dwarkesh Patel 01:46:56

Are you game for some audience questions?

Sarah Paine 01:46:58

Sure. Uh-oh, what are they? No advance notice.

Dwarkesh Patel 01:47:00

Anonymous asks—

Sarah Paine 01:47:01

Anonymous, brave person.

Dwarkesh Patel 01:47:05

"What is a good grand strategy for a minor power that wants the status quo to continue? For example, present-day South Korea probably does not want China and the US fighting over Taiwan."

Sarah Paine 01:47:14

Its allies and alliance systems. Smaller powers have been tremendously influential in the rules-based order, on how it's developed. They've provided some of the really good ideas over the years. It's "cooperate" because if you have allies… This is what the United States doesn't get. We're big, but if we're going after everybody else, they're bigger. If you're small… This is how Europe is important. There are a lot of small countries or medium countries, but when they get together, they get powerful. When they add in Japan, they get the equivalent GNP of us. Then you start adding other people in and they're bigger. Yes, it takes a long time to negotiate whatever the agreement is going to be, but once you agree, it's powerful. That's the answer.

This is how we have as much of a rules-based order as we do, which allows everyone in the audience to take a little piece of plastic called a credit card, go to any country in the world, pretty much—I wouldn't go to North Korea, but most countries—and you use that credit card. If it says you have enough money to buy a Mercedes and there's a Mercedes in the shop, the chances are the credit card will go through and you can buy your Mercedes. It's just a piece of plastic. This is the rules-based order that most of us are into. It's because we agree with each other on these things.

Dwarkesh Patel 01:48:35

Denny Trimcev asks, "What made Japan a more fertile ground for Western ideas than China? Was China just too big for meaningful institutional change in the same time scale?"

Sarah Paine 01:48:45

This is my guess because who knows? China is suffering from the incredible success of its civilization. It had been the dominant civilization of Asia forever and ever for excellent reasons. If you look at the achievements of China, they're amazing.

For Japan, it had always been in China's shadow. The Japanese had always tried to learn from others. The Chinese also learned from others. A lot of their institutions come from the Manchus, but they pretend that, "We sinify these people," and so they're actually Han Chinese achievements. Not so much, actually. The myth lives on that anything that's worth knowing emanates from China. Hubris. This is a human condition, and the Chinese get a fatal case. What I'm wondering about is whether this country has got a fatal case right now. “We know everything. We can do everything.”

Dwarkesh Patel 01:49:42

This leads in very nicely to the next question. Lydia Dean asks, "In the spirit of the Meiji delegates sending out to explicitly study the West, where would the United States do well to send delegates today?"

Sarah Paine 01:49:54

Delegates today? Our allies. Start by apologizing to them. That would be step one. I can't see that ever happening. We have all these international organizations. Use them. We're in the process of firing the State Department. I believe when WikiLeaks leaked a lot of their memos and things and the New York Times was reporting on it, what it revealed was incredibly accurate reporting of whatever place they were assigned to, and describing things in really accurate detail that would enable US policymakers then to make an accurate assessment and determine to do what's next. But that's not what we're doing.

Dwarkesh Patel 01:50:47

This is actually a very interesting take you gave to me offline once about the WikiLeaks leaks of the State Department documents. You were like, "Oh, wow, they're kind of on it. These seem like good memos."

Sarah Paine 01:50:58

I technically wasn't allowed to read them because they were classified and I was working for the Navy. So I didn't go out of my way to read them. But I'm reading the New York Times and other things that are reporting on them. From what's there, you just see that it's accurate and it's in-depth reporting. Of course, we're angry that they're all out there because it's revealing who's talking to whom and whatever. But what it showed is a very professional staff.

I don't know why we got some axe to grind with the State Department. It's also really tiny. Working at the War College on student solutions to the world's problems, they would always say quadruple the State Department's budget because it's really tiny compared to the military budget. There's an understanding that what much of the US government does, if it's doing what it's supposed to, is it's preventing catastrophes. If you're sending your State Department out there, you're trying to prevent an Indo-Pakistani nuclear war. We've sent diplomats at different times to prevent that from happening. Who knows why it didn't happen? Maybe it wasn't the State Department. But all over the world trying to prevent these things…

If you think about the government, it's about setting all these standards to prevent disasters. I’ve talked to you offline so this must be repetitive for you but, you immunize people to prevent disease. Can I prove who didn't get what? No, I can't. I can talk about people who got bad reactions to it. But then there's a failure to understand. There are some bad reactions, but let's look at the much greater number of people who are alive in our own day. Earlier, when Roosevelt was president, Roosevelt had polio, and people were scared to death of getting polio. They understood the importance of these things. Everyone was lining up to get the polio vaccine when Dr. Salk invented it.

There are other things, like having enough meteorologists to predict storms so you can get people out of summer camps when you need to get the people out of summer camps, having appropriate spending of government funds to have the appropriate sirens and things. People are looking at the expense of doing these things but they're preventing disasters. It's the negative objectives. You can't measure the things that fail to happen. But I assure you, by averting a catastrophe you're saving a lot, probably a lot more than whatever you're spending on the budget. The US government needs to be much more efficient. I get it. But tearing the whole place down is going to be a mistake, and we will get more catastrophes.

Dwarkesh Patel 01:53:35

Alright, next month we're doing a podcast with you and Elon together.

Sarah Paine 01:53:39

Oh, Elon. Oh, lucky me. That's a whole other thing.

Dwarkesh Patel 01:53:45

Maybe we should save that for the next lecture.

Sarah Paine 01:53:50

The different varieties of American hubris.

Dwarkesh Patel 01:53:55

Welcome to San Francisco, Sarah.

Sarah Paine 01:53:57

One variant of hubris A is, I'm an American. I don't need to consult any other country. It's just whatever Washington does or doesn't do. That determines how the world works. It's half-court tennis. It ignores what the other team's doing, and it doesn't work out well.

Then there's Silicon Valley, or whatever it is, a different variant. It's what I call Type 2 hubris. “I'm incredibly talented in one area” which he is, he’s incredibly talented or whatever. He's made himself a household word. Not everyone can do that. But you're thinking, “Because I'm good at one thing, this is transferable to everything else.” I got news. There's all kinds of expertise and it's important. But if I take a really good hairdresser and make them an auto mechanic, or the auto mechanic a hairdresser, it is not going to work out well. You have to line these things up correctly. This type 2 hubris is thinking, “I'm smart at one thing, and therefore I can do everything.” These people are making decisions in areas where they have no expertise, and it will not work out well.

Dwarkesh Patel 01:55:00

Great place to close.

Sarah Paine 01:55:01

Yes.

The Stephen Kotkin episode. Kotkin is arguably the world’s foremost expert on Joseph Stalin and has written a massive 2-volume biography of Stalin (with a 3rd volume in the works).

No other individual had more of a profound impact on the 20th century than Stalin. He held the power of life and death over every single person across 11 time zones, and he killed tens of millions of people, utterly consumed by an ideology aimed at building paradise on Earth.

And he was one half of the biggest and most consequential military confrontation in history (even if Hitler didn’t prove to be his match).

Watch on YouTube; listen on Apple Podcasts or Spotify.

• Lighthouse is THE fastest immigration solution for the technology industry. All they need is your resume or LinkedIn profile to tell you which visas you’re most eligible for, and they’ll send you this eligibility document for free, no commitment required. Get started today at https://www.lighthousehq.com/ref/Dwarkesh.

To sponsor a future episode, visit dwarkesh.com/advertise.

(00:00:00) – Was the tsarist regime the lesser of 2 evils?

(00:23:45) – The peasants brought Lenin to power, then he enslaved them

(00:37:38) – Why did so many go along with enforced famine and the Great Terror?

(01:02:26) – Today’s leftist civil war

(01:13:01) – Doesn’t CCP deserve credit for China's growth?

(01:35:13) – Why didn't somebody just kill Stalin?

(01:52:45) – Overcoming the pathologies of communism with tech: USSR vs China

Dwarkesh Patel 00:00:00

My guest today is Stephen Kotkin, who is a senior fellow at the Hoover Institution and author of two-thirds of his three-volume Stalin biographies. The first one, Stalin: Paradoxes of Power. The second one, Stalin: Waiting for Hitler. Thank you for coming on my podcast.

Stephen Kotkin 00:00:14

Thank you for the honor of the invitation.

Dwarkesh Patel 00:00:16

Let's begin with the tsarist regime. First question, how repressive was the tsarist regime actually? Presumably the motivation behind the revolution is to get rid of this autocracy. But you have these examples. Lenin's brother tried to kill the tsar. Lenin himself is writing these long manifestos about taking down capitalism and overthrowing the government. He and people like Stalin are in exile in Siberia, living off government money, robbing banks, and small shenanigans. Honestly, it sounds more forgiving than many countries today. How bad was it really?

Stephen Kotkin 00:00:50

You have to put yourself back in the time period to judge the level of repression based upon what norms were, what other regimes did, rather than take the 20th-century regimes as the guide and go back. We need to widen the aperture a little bit here.

This is the tsarist regime's problem. It needs to be able to compete in the international system. That means it needs a modern military and modern industry to underwrite that modern military. So it needs armaments, it needs steel, it needs chemicals. For that, you need workers.

So you want the workers only to work in the industry. You don't want them, for example, to have a labor movement or to go on strike, or to have ideas about how politics should be organized. Similarly, with the intellectual side, you need the engineers. You need the engineers in order to design and build the modern attributes that you need to compete as a global power. But you don't want those educated people to have their own ideas and values about politics, about whether you'd want an autocratic government like the Russian regime has, or whether you'd want some other type of government.

So all of these countries in the modern period have this dilemma: importing modernization but keeping out the political side, the value side that goes along with that. They need to have some way to repress and control the working-class organization, movement stuff, and the university-educated intellectuals. That's a problem we still have today. The Iranian regime now has that problem. The Chinese regime in Beijing has this problem. The Soviet Union had that problem. Contemporary Russia has that problem.

How do you bring in modernity—meaning you have tanks, you have airplanes, or you have AI—but keep out, for example, separation of powers, freedom, property rights, all the things that undermine your dictatorial rule. The tsarist regime was a quintessential example of this fundamental dilemma.

Modernization is not a sociological process that kind of just happens. It's a geopolitical process. You modernize because you need to compete in the international system. If somebody has ships made out of steel, either you also have ships made out of steel or they're going to show up at your door, like we did to Japan, and tell you that they're in charge now.

Dwarkesh Patel 00:03:45

This was one of the most interesting takes in your first volume, that modernization is not this inevitable process but is instigated by ruthless geopolitical competition. Do you think that that still applies in today's world? Yes, there are pockets of conflict in the Middle East or in Ukraine, which would motivate the key powers there to want to have modern militaries and modern technologies. But through most of the world…

For example, the odds that if France falls behind technologically if their AI is worse, Germany is going to take over, is just sort of unthinkable. This dynamic where in order to ward off colonization or other great powers, you need to stay at the cutting edge of technology and also have up-to-date political processes. Is that still a drive that moves countries forward?

Stephen Kotkin 00:04:31

If you have an autocratic regime, it is existential for you every day. You want to compete. France can compete or fail to compete, and its political system is not at risk. No one's going to say the regime is illegitimate because someone else beat them in AI. Students are going to protest in the streets. That's not going to mean that the regime is going to fall. There may be a change of government, but the system remains.

You have this dilemma for the authoritarians. Think about Peter the Great. I need to compete against the great powers, so I need to have a navy. To get a navy, I need to have the industry that supports a navy. I need to have the officers. I need to have the technical skills. So I need to have all of that to be able to compete, but I have this autocratic regime. How do I retain the social structure, the hierarchy? It’s non-elected, non-legitimate in some ways based upon modern understandings of constitutional order. How do I retain that while I'm importing these attributes of modern power?

That's the stuff that persists today for Iran, Russia, China, North Korea. They have to get very good at holding at bay those attributes of modernity that threaten their political regime while importing as much as they can of the attributes… It's two sides of the same coin. The thing that gets you the engineers also gets you the possible political ideas.

So the tsarist regime begins to repress the very thing it needs to compete in the international system. It represses the working class and it represses the engineers and the intellectuals, without which it can't be a great power, without which it can't compete, but with which its political system is threatened.

So the amount of repression is an important question. We would call this a vegetarian regime compared to the carnivores, like Stalin's regime or Hitler's regime, in terms of the degree of repression. But the dynamic of being compelled to exercise repression against the very people you need like oxygen, that's the dynamic that we see in tsarist Russia and that we still see today in a certain form.

Dwarkesh Patel 00:07:03

Is one of the key lessons from your volumes that you should be tripping over yourself in order to embrace a lesser of two evils? Does that apply to all the examples you give? This is maybe a general question about how much you can actually learn from history. For every seeming lesson, there's an equal and opposite lesson that you can also learn.

During the tsarist regime, in retrospect, we can say that the liberals and the constitutionalists should have cooperated with Stolypin or Witte. Even though it was an autocratic regime, they were actually doing these real reforms and there was growth. They should have continued that process. Or when the government fell in February 1917, the Provisional Government factions should have united to oppose the Bolsheviks.

But then there's all these other examples. In Germany, the conservative Weimar government allies with Hitler in order to fend off what they think is the greater evil, the communists. Given the events up to that point, it's a reasonable concern to have, given what the Bolsheviks have done in Russia. Where should we end up on this? Should you embrace a lesser of two evils whenever you get the chance, or no?

Stephen Kotkin 00:08:03

The Tsarist regime is undertaking this repression of people who have legitimate claims. That repression is quite severe by the standards of the day. It's not going to be everybody murdered or everybody deported to the wastes of Siberia, which we're going to see in the 20th century when we have a different level of communications and transport, different technology, when we have a different level of ideological commitment. Still, it's highly repressive. It's totally unjust and the claims of the people protesting are legitimate.

Stalin goes into the underground, not because he's looking for power. It's because he's dedicated to fighting the injustices of the tsarist regime. A lot of young people like him do the same. He's in the seminary. He's highly successful, unbelievably successful. He's great at school. That's been true for many years now, since he was in elementary school. He sang in the choir, got good grades, and did his homework. He's on track to be successful in society.

He gives it all up. He starts reading forbidden, otherwise censored underground literature and books. He learns about social injustice, not just through firsthand experience, but analytically. He never graduated from the seminary, which is the highest level of education that was available to somebody in the Caucasus. They don't have a university, because the Tsarist regime is afraid to allow a university. They need the university graduates, but again, they're afraid of the politics of it. Universities elsewhere, including in the capital, St. Petersburg, are constantly being shut down through these revolutionary episodes.

He goes into the underground and for 20 years of his life he's got no job, no profession, no source of income. He's in and out of prison, in and out of Siberian exile, constantly harassed by the police. If he escapes, they find him, they put him back. From the ages of about 17-18 to his late 30s, he's a penniless, jobless revolutionary, dedicated to fighting the genuine injustices of the Tsarist regime.

What he'll produce is a much more unjust regime than the one he's fighting against. This is known as perverse and unintended consequences. He's legitimately dedicated to revolution as he understands it in his day, and it's fighting against legitimate injustices. But the way he does that—the revolutionary methods that he uses and then the regime that he ends up building—turns out to be worse than the problem that he was addressing. This is a perverse and unintended consequence.

Your question is about whether revolution is a good thing ultimately, even if the injustices are there, and whether there could have been some solution that was more evolutionary that could get you to a better place. The constitutionalists, otherwise known as the Kadets, we would call them the classical liberals. They're the private property, constitutional order, anti-autocracy people fighting the tsarist regime. They're going to be the protagonists of the February Revolution in 1917. It looks like they could possibly be the solution because they're against arbitrary, autocratic, unjust rule. They're in favor of constitutional order. That's not how they behave once they're in power. But let's leave that aside.

Here's your problem. This was foreseen by the Interior Minister who put down the 1905 Rebellion Revolution, Pyotr Durnovo, who understands the liberals, the constitutionalists. They’re the classical liberals who want private property and constitutional order and probably, if they had their druthers, a constitutional monarchy like you have in the UK and Britain at the time. But Durnovo says to them, “You guys are fools, because if you take down the Tsar, you won't get a constitutional order. You'll get chaos and anarchy and you'll get a massive social revolution by the peasants predominantly, but also by the nationalities and the workers. You will bring on the opposite of a constitutional order. So instead of fighting against the Tsar, you should throw in your lot with the Tsar against the Stalins and the social movement that came from the investment in modernization.”

Durnovo is proved right. He's proved right not just in Russia, by the way. In many places in the world at the time, between like 1905 and 1925 or so, 1926 in the case of Portugal, you have constitutional revolutionary attempts to introduce constitutional order: Mexico, Iran, China, Russia, Portugal. They all fail. The constitutionalists take power for a brief period of time and then they're swept out by a more leftist, more social-oriented revolutionary process. So the constitutional epoch of the early 20th century turns out to be a bust.

It had to have happened beforehand, before the modern era. Why? Because when you institute constitutional order, like in England, in the US, to a certain extent in France which has a more complicated process, you do it before you're in the mass age. You do it before the working class and the peasantry are politically organized. You're able to introduce rule of law, constitutionalism based on a private property model where not that many people get to vote. The franchise is restricted, the vast majority of people can't vote, only property holders can vote. Only men can vote, not women. You have this restricted franchise, which is like a breathing space for you to introduce and get used to a liberal constitutional order that you can then democratize over time. Over time non-property holding males get to vote. Over time, women get to vote. Over time, slaves become citizens. They become people and fully-fledged citizens. They also get to vote and to own property legally.

You have this order that has all of these birth defects. It's very restrictive in the franchise. Some people are slaves, not even considered people. Yet over time you can get that right because the category of citizenship and the constitutional order are already embedded. When you have the constitutional revolution in the mass age—when you have the peasants and the workers and those for national self determination participating in the constitutional revolution—constitutional order and rule of law isn't enough for a lot of them.

Dwarkesh Patel 00:15:38

It worked in Taiwan and South Korea. There was an era of industrialization under, not a dictator, but an authoritarian government. Then they were able to transition to rule of law democracy.

Stephen Kotkin 00:15:48

There are these exceptions which have turned out to be false normative, or guiding, stars for us in every case. It happened in West Germany and Japan under American occupation, enormously successful. We turned Hitlerite Germany into our ally with their cooperation. It's just astonishing. And with Hirohito's Japan, the emperor stayed. Then with Japan's two former colonies, South Korea and Taiwan, it worked there as well. Hong Kong was on that trajectory until it was turned back when the lease ended to the communist regime in Beijing. But the number of countries that have done this is very, very few. The opposite has happened in most other cases where you've gotten an attempt but a failure to introduce enduring constitutional order in a mass society.

Dwarkesh Patel 00:16:48

But there's a paradox here. If you institute this sort of revolution, or changing of the guard, during the mass age, then you're going to get this sort of leftist revolution which is very antithetical to future prosperity and rule of law. On the other hand, if you don't change the regime, you will fail to… It's been pointed out that Chiang Kai-shek, when he did have control over China, should have done some amount of land reform.

You talk about how Stolypin, after the 1905 attempted revolution, attempted to put in this sort of agricultural reforms. But their success was mixed because the existing aristocracy obviously didn't favor it. So there is this paradox. If you don't change the regime, the existing stakeholders will not want the kinds of reforms which would make it possible to have a lower class that's bought into the system.

Stephen Kotkin 00:17:35

How do you bring the whole society, what we would call the masses or what used to be called the masses… How do you bring all levels of society into a political system as citizens? How do you build a polity which is inclusive of people regardless of how much property they have? How do you then provide opportunity to them so that they can rise up the ladder? That's the secret of success in the modern world, which a handful of countries have done from a non-democratic starting point. Other countries have done it by democratizing a liberal constitutional order.

Go back to the tsarist regime. What were the options for the tsarist regime? You have this very heavy absolutist regime. Autocracy is even a more absolutist version of what we had, for example in the French case, when the Bourbon dynasty could say, “the state is me” or “the state is I”. You have this Versailles-like absolutism, and aristocracy that forms around the absolutism that is the main beneficiary. Most everybody else is excluded from the political process and they are going to break through at some point. So how are you going to manage that breakthrough.

Again, in the French case it took more than a century to get this right. The monarch is killed and the monarch comes back and they have an emperor and they have a constitutional order and they have an elected president who does a self-coup and one republic falls and another republic takes its place. Eventually even the Vichy regime, the Nazi occupation regime, overthrows the republic. So it takes de Gaulle in some ways from above to reimpose a republic, the fifth Republic, later on. So it's a very difficult process, even when it works, as the French case tells you.

So we're not saying that it's simple and easy and that Tsarist Russia could have gone down this path. A lot of people say, if it hadn't been for the war, Tsarist Russia was on an evolutionary path. It was modernizing economically, so it might have been a kind of Taiwan story over time, where the dictatorial regime gave way, under economic success and political pressure, to a more benign regime. They institutionalized rule of law, private property, civil liberties, and an inclusive polity for everybody.

The problem with that is that the autocracy refused to do that. The autocracy had wanted no part of any evolutionary process. So in 1905, when the tsarist regime is compelled—under pressure of tremendous peasant revolt and worker strikes and a defeat in the war against Japan—to introduce some version of constitutional order, a kind of quasi-parliament, quasi-constitution, the Tsar regrets doing this almost immediately and is trying to push back against his concessions.

Once the lid was put back on because of repression, once Durnovo successfully repressed the political movement, the Tsar wanted to undo those concessions and go back to being an autocrat. An autocrat, autokrator, means a “self-power, a power unto itself”. So this is your challenge. How do you undo the autocracy and get to an evolutionary mode when the autocracy itself is committed to not allowing any political participation whatsoever?

So you have the leftist version of overthrow, where you end up with a radicalization in the leftist direction, and you have a rightist version of overthrow, where you end up with, “Oh, we're not going to have Leninism here. Let's prevent Leninism. Let's go with the radical right.” The traditional right invites, as happened in the German case and earlier in the Italian case, invites the radical right to power, thinking they can control the radical right, the fascism, the Nazism.

The traditional right is wrong. The radical right, once it's invited to power, institutionalizes itself. You get a leftist version of this and a rightist version of this. They're kind of codependent. Each uses the threat of the other to further consolidate their dictatorship.

This is the 20th-century version. The irony here is that you got the radical right, the fascist solution in the German case, and you got the radical left, the socialist solution in the tsarist case. The tsarist regime had a massive radical right. They had the Protocols of the Elders of Zion, that infamous anti-Semitic tract, which then makes its way to Germany but originates in the tsarist empire. The antisemitism is there. The right-wing movement is there, implanted in villages, the Union of the Russian People.

Russia has the fascism before Germany. Germany has the largest socialist party in its parliament, in the sense that it's not a majority, but it's a plurality. The Social Democrats in Germany are enormously successful at the ballot box. The right-wing movement with the antisemitism is enormously successful in the streets in Tsarist Russia. If you were alive before 1917, 1933, you would predict that socialism would be victorious and triumph in Germany, and the fascism would triumph in the Russian Empire. But it's the opposite. So that's just a fascinating, amazing paradox that I tried to deal with in the first two volumes.

Dwarkesh Patel 00:23:45

You say that in 1917 a leftist revolution of some kind was inevitable, but that it didn't have to be the Bolshevik October Revolution. Why was leftism inevitable in Russia?

Stephen Kotkin 00:23:55

You put your finger on a big part of it when you talked about Chiang Kai-shek and land reform. You have this peasant land hunger. The peasants are often without their own holdings. They work on someone else's property, or their holdings are so small that if there's a little bit of bad weather let alone a massive drought, they're on the verge of starvation. Subsistence level agriculture is not politically stable.

You want a class of people, a kind of yeoman capitalists, property owners who can expand their farms and can succeed and hire labor. Some of those hired hands can then get their own land and become a version of these yeoman farmers, sort of Thomas Jefferson-style or Stolypin, the great attempted reform of Stolypin after the 1905 revolution, which ended in his assassination.

You need to deal with the peasant land hunger so that it becomes a stabilizing political force. You have the peasants get the land and then they have a piece of the status quo and want to retain the system, versus the peasants not having the land and they want to overthrow the system to get the land.

In the Russian case there is the end of serfdom in the 1860s, again as a result of the defeat in the Crimean War, where there is a reform. They free the serfs, emancipation of the serfs. But the serfs don't get the land to the degree that could have happened, because the landowners are the political support of the tsarist autocracy. To take the land away from the land magnates and give it to the peasants is to go through this risky path where you're losing one political support, the landowners, before you've fully gotten the new political support. You're gonna go through this valley of hell potentially where all bets are off and you're not sure if it's going to work.

So the peasants don't really get the land as they could have in the 1860s. It became a problem that's not resolved right through 1917, 1918. So the peasants had their own revolution in 1917 and 1918, which was not about the socialist parties. It's not about the Bolsheviks, it's not about Lenin, it's about the peasants seizing the land. But that creates an intense radicalism that becomes the platform for the socialists in the cities to gain and hold power in the system.

You don't have that in the German case. In the German case, you have strikes and seizures of power in a few places, like Bavaria for example. You have a Bavarian Soviet Socialist Republic, but they're easily put down by the forces of order or the army and guess who's in the army? It’s the peasants. You don't have a peasant army ready to put down the revolt in the Russian case, because the peasant army is the one seizing the land. It's the one doing the radical revolution. So you lack the forces of order to destroy the leftist movement in the Russian case, because it is the leftist movement in the Russian case which should be the forces of order.

In the German case—and to a certain extent the Italian case which happened simultaneously, and there's also a Hungarian case here—you have leftist revolts in the cities, seizures of power like the Paris Commune of 1870-71, which happened in Paris and not in the rest of France. You need a peasant army that has a stake in the existing order to undo the city leftist revolution. You have this in the other European cases.

One case you don't have this is in the Russian case. Then you're going to get to the Chinese case later, which is going to be a variant of what happened in the Russian case, where you have a gigantic land hungry peasantry that's going to become radical for a time. Again, there are going to be perverse and unintended consequences.

The peasants are ready to destroy the existing order, not to bring communists to power, but to seize the land themselves. In the 1920s, the peasants were de facto, not de jure, landowners. They don't own the property in law, they own it in fact. But then Stalin's going to reverse the peasant revolution violently and re-enserf or enslave the peasants across all the 11 time zones, this gigantic Eurasia. The peasant revolution is going to be annihilated in blood. So the peasants have, through their radicalism of seizing the land, helped bring Lenin and his Bolsheviks to power in the cities which is going to be the death of the peasants owning the land and lead instead to the re-enserfment of the peasants. Something similar is gonna happen in the Chinese case.

Again there's this irony of history: perverse and unintended consequences. Stalin is fighting against tsarist injustice only to impose worse injustice and worse bloodshed and worse repression. The peasants are fighting on behalf of obtaining the land, only to then be expropriated and forced into these collectives and losing the very land, the land that they took in the seizures that brought these leftists to power in the case.

In central Europe—the southern German case, the northern Italian case, the Hungarian case—you don't have the endurance of the leftists in power. They're all thrown out. They're thrown out by the forces of order. They're thrown out by the right. So the traditionalism of the peasants—where they believe in God, they believe in law and order—is overriding because they already have a lot of the land in comparison to their Russian or Chinese counterparts. They can be part of the forces of order.

So you can get fascism in Central Europe. You can get the right wing dictatorships in Central Europe, the forces of order destroying the left. Whereas you get the leftist dictatorships in the giant peasant societies where you don't quite have the distribution of land. Now the peasants are complaining about land distribution in Italy and Germany, don't get me wrong, but relative to Russia and China they're doing well. So then you think about the Mexican case, the Iranian case, the Portuguese case, all of which are peasant societies as well.

There's how you integrate farmers. The whole world order rests on the backs of farmers. How much farmers till means how rich or poor your country is. Whether you have a surplus, as we call it, that the farmers can sell on the market after they consume what they need for their family's purposes or not, tells you how much wealth you have to then build an army, build modern industry, et cetera.

The world order rests on these hardworking, predominant in the population, peasantries. In some ways, the political system doesn't derive in deterministic fashion from them. Politics still matters and politics is never reducible to social relations. But failure to master, or mastery of the social relations of, the peasant land question is fundamental in some of the political outcomes.

The politicians have to be good at managing the peasantry's integration into that society, where you're trying to get an order in the mass age. You're beyond just the court society at Versailles, the tsarist court in St. Petersburg, or the men at the Constitutional Convention in the US. You're beyond that in the mass age. You have to be able to incorporate the masses somehow in a polity. It's really hard to do. So this dynamic of failure to master, or mastery over it, tells you a lot about the direction you're going to go.

Dwarkesh Patel 00:32:19

This answers one of the other questions I had for you. Why did we see these communist revolutions in peasant countries? It’s the opposite of the Marxist prediction that you would first need capitalism and industrialization before you would see the turn towards socialism. I guess the answer is that the private property which is engendered by capitalism and industrialization actually helps the peasants more, or helps them somewhat and buys them into the system.

This raises another question. If it's the case that all of this unrest is caused by the mistreatment of peasants in China and in Russia, you have the mistreatment of them to an extent unimaginable after the collectivization in 1928 where literally 100 million peasants were enslaved. Of course, there's some lack of cooperation with the regime. They kill half the livestock and so forth. But it doesn't break the regime, even though it's way more repressive and destructive than anything the tsar did. So if the peasants are the backbone of the regime's stability, why doesn't collectivization in China and Russia break the regime?

Stephen Kotkin 00:33:24

Terrific question again. You have a multi-pronged answer. Let’s do it this way.

On the one hand, you have a much bigger repressive apparatus, a much, much bigger repressive apparatus. The tsarist regime has a very small secret police, really small. The secret police for the tsarist regime is mostly following a handful of intellectuals. You had a few thousand university students in tsarist Russia in the mid-19th century, about 5,000 or so when this term intelligentsia got invented. You're going to have a few thousand more over time. But you're in the thousands, not the millions.

Dwarkesh Patel 00:34:09

Even there though, Lenin, Trotsky, and Stalin go off and they're sent into exile. They're not only living off government money, but while in exile and they're writing for Pravda. They're writing, “here's my manifesto on the fall of capitalism.” Even the intellectuals are not really repressed.

Stephen Kotkin 00:34:25

The sense of repression is more about following what they do rather than arresting them on a pretext and putting a bullet in the back of their neck. You have this tsarist secret police, the “Okhranka”. That's the pejorative nickname for them, the “Okhranka”. They're tasked with following these revolutionaries and infiltrating their groups and maybe sabotaging them from within. Something similar happens in the labor movement. Zubatovshina is the Russian term where you plant the leader of the workers' movement in order to make sure that it's controlled by the secret police rather than has a spontaneous or autonomous version that could get out of hand.

So you have a small police that's dedicated to surveillance and infiltration. You're reading their mail, which is something that's invented in France. The black cabinets are a French invention that the tsar's secret police borrow. You're following them and a lot of them get deported to Siberian exile, like happened to Stalin. Some get forced into European exile, like happened to Lenin. Lenin for 15 of the 17 years between 1900 and 1917 was in European exile. He's not even in Russia.

In fact, the Paris branch of the tsarist secret police, the Okhranka, which conducted the surveillance and infiltration in Europe - we have their entire archive right here at the Hoover Institution. It was supposed to be destroyed. The order went to destroy it in Paris after the revolution. Instead the guy put it on a boat and secretly had it shipped here to the United States. Now we have the Tsarist Russian archive, secret police for the foreign revolutionaries in foreign exile.

So you have surveillance and infiltration on a lower level. The main force of repression in Tsarist Russia is the army, not the secret police. You don't have a gigantic armed secret police. The secret police are kind of intellectuals. They're reading Lenin's tracts and they're writing summaries, like AI would do now, about what they contain and how to combat it and why the idea is wrong. They're sort of like pseudo-intellectuals, or in some cases intellectuals with degrees. They're not the thugs, the torturers and the thugs that we would associate with secret police.

That's built under Stalin in order to enact the re-enserfment, the enslavement of the hundred million peasants. It's that act. It's a kind of chicken and egg thing. How do you enslave the peasants without the gigantic secret police? But then when you enslave the peasants, the result is you have this gigantic secret police now that can do everything and anything. It's a process where the chicken and egg are happening simultaneously and they're building the secret police capacity, while enslaving the peasants, that they didn't have before.

Dwarkesh Patel 00:37:38

How do we explain this surplus of sadism during this period in Russia? Stalin recruits the twenty-five-thousanders who go out to the countryside and steal from basically starving people. They can visibly see, I'm sure, that they're stealing from a family that's going to starve without this grain. You have tens of thousands, maybe hundreds of thousands, of interrogators and torturers in the gulag system. They must know it's a cynical thing where they're making the person confess to a thing that they haven't done and they're employing torture to do it.

It wasn't just Stalin doing all these things. There were hundreds of thousands, maybe millions of people if you include informants, who are implicated in this whole ghoulish regime. Is this just a latent thing that is true in any society and Stalin was able to exploit it, or with some circumstance created this level of sadism?

Stephen Kotkin 00:38:29

Again, a fantastic question. We get this from Lev Kopelev, The Education of a True Believer. He was one of these people. He later then becomes a dissident. He gets forced into exile in Germany. He's a Germanist by profession. He writes this fantastic memoir, a couple of memoirs. One of them is called The Education of a True Believer, which covers how he was the one who did this, including in his native village.

Here's your answer. On the one hand, there's ideology and the importance of ideology. We may think that no one really believes the ideology. The ideology is too ridiculous. It's too disproven by facts in life. We're too smart and they couldn't have been as stupid. They had to be smart like us and not believe these crazy fairy tales about it. In fact, it's wrong. They do believe the ideology and they're young. They're young people.

A story of the evil of capitalism… You have World War I, millions of people die. For what? Why are those millions of people being killed? The flower of European youth. Then many colonial armies get drawn in because of European imperialism. Young people, the future of those countries, go to their senseless deaths. What's that about? It's about imperialism, it's about capitalism. So that's evil and we must overcome that.

So there's a way in which life experience, as well as the fervor of youth, leads to, “let's build a new world today rather than wait for tomorrow. Let's be impatient, let's eradicate capitalism, let's bring about socialism. Let's bring about socialism, meaning we end war and imperialism. Achieve abundance for everybody, so that it's not just the haves and the have nots, but everybody's got something. In the process, let's make my little life world historical.” So here I am, just a little activist with a red star on my cap and my life means nothing. Except I'm a participant in building a new world, in a world historical process that's going to end exploitation, that's going to end haves and have nots.

Dwarkesh Patel 00:40:50

I get it if you're a middle bureaucrat in the Communist Party, sure. Do you think that explains the motivation of an interrogator in a Gulag, that this is part of the goal of communism?

Stephen Kotkin 00:41:04

There's a big story here, which is about how we're building a new world. There are people against that because they're the bourgeoisie, or they're the fools who are doing the bourgeoisie's business on behalf of the bourgeoisie. They're duped into false consciousness.

Dwarkesh Patel 00:41:22

But in many of these cases, they know that they're the ones orchestrating this show trial, this cynical game. They know that they just picked up a random person in the dead of night.

Stephen Kotkin 00:41:30

They know that there are enemies out there. They know that this process has people who are against it. That's a given. Who are the people who are against it? They're masking their true feelings. They're hiding behind professions of loyalty. When in fact, when the hour of crisis comes and there's a war, they will be the saboteurs behind the lines, the fifth column.

It might be that some of them are innocent that you're arresting, but some of them are clearly going to be guilty as well. To get the guilty, you have to somehow manage to deal with victimizing people who are likely innocent and you may know are innocent. But you also know to your bone marrow that some people out there are enemies. It's hard to identify and find them. So you're overcompensating a little bit to make sure that you get every last enemy. Again, it's a crazy idea to us. It makes no sense to us, but a lot of things make no sense that people believe in.

There's this young kid who's really adept at social media who just looks like he won the primary for Democratic candidate for mayor in New York City. One of the things he wants to do is freeze rents, rent control, because he wants more affordable housing. He's a complete idiot in terms of facts, because the way to get more affordable housing is to build more housing.

If supply massively increases and it exceeds demand, the price has to go down. It's proven again and again and again. What rent control does, or freezing of rent, it inhibits the building of new housing. Because who's going to build new housing when you can't make money off of it? So rent control is what produces the lack of affordable housing in the first place. He sees what's the problem as the solution now.

Now, is he a fool? No, he's a really bright guy. He's very well-educated. He's read everything and anything. He's been to university. He's talked to a lot of really smart people. You'd say, “how could he be so foolish to believe an idea that's obviously falsified by empirical reality?” But again, it's an ideological belief. He wants to allow people who can't afford Manhattan to live there. That's a good idea.

Life should be more affordable. There should be more places like Queens, where you can come in as immigrants or you can come in as lower class, front end of the social ladder, like my family did for example. My father worked in a factory. You should be able to get some housing for your family, work hard, and rise up. I agree with that 100%.

But he's got an ideological approach to how to achieve that, which to me is completely foolish. If I were as smart as him… how could he possibly hold that idea in his head? But ideology is pervasive. It's pervasive, flying in the face of empirical reality. We could give many examples. I'm not picking on this guy in New York. It's just a recent example. I don't know him. I've never met him. Whatever. Maybe there's a more complicated story there. I'm just saying that we have to take ideology seriously because it's deep and it can be enduring even in the face of empirical reality.

Dwarkesh Patel 00:45:18

There's ideology, but there's a very specific thing to these Marxist regimes. They might believe in class conflict and that you need this revolution and so forth. But there's also this sense that you cannot contradict the party, you cannot contradict the vanguard. Even in 1924, when Trotsky is getting condemned by the party, or whatever that was, he gets up to give a speech to the Party Plenum, and he says, “Look, for all of my thoughts, let there be no mistake that the party is always right and party discipline is always important.”

So there's not only the sense where Mamdani would say, “Oh, I want these specific policies implemented” but there's also the sense that loyalty to the party, and eventually to Stalin even when it seems to contradict my understanding of socialism, is absolutely paramount. One way to explain that is that they were just genuinely afraid of Stalin and they thought this was antithetical to their understanding of communism. Another is that part of the ideology is this theocratic understanding that the party's always right, even if it seems like a single individual is manipulating it to their ends.

Stephen Kotkin 00:46:28

Why is Marxism-Leninism especially so attractive to young people and to intellectuals? Why? We have this history, which is a bloody mess. Millions of people die, and they die because of the enactment of this ideology. How could people continue to adhere to an ideology like that during the murderous time period, and even more after the murderous time period when we can look at it dispassionately?

Here's part of the answer. Young people are attracted to impatient, quick, total transformation of the world, eradication of war, eradication of social injustice. There's a simplicity to the ideology. It's a total package. It gets rid of everything bad if you just follow the precepts. Sure, things happen that shouldn't have happened. There are some surprises, there are some downsides. But are you pro-capitalism? Are you pro-imperialist war? Are you pro-landowners having all the land and the have-nots having nothing? There's this constant threat where if you contravene the ideology, you're in bed with the very evils that the ideology is trying to overcome.

You become an accomplice in the persistence of the things that you're dedicated to overthrowing. It's not just that you're loyal to the party. You're loyal to the outcome that the party is dedicated to achieving. You know that there are going to be mistakes and costs and bad things will happen along the way, but is imperialist war better than that? No, the answer is imperialist war has got to be worse.

The other reason, which is even deeper than that, is that Marxism-Leninism empowers the intellectual class and the lumpen-intellectual class. In a market system, you get to do what you want. You want to open up a family business, you want to take a loan and give it a try? You can do that. Nobody can stop you. It might be that it's hard to get the loan in some neighborhoods. It might be that the loan interest rate is… You've got to work much harder than you thought, etc. But you get to make the decisions. You get to decide what to do, when to do it. You can work for somebody else, you can put out your own shingle.

In these kinds of systems, it's the intellectuals and the lumpen-intellectuals who make those decisions. They use the state as an instrument to overcome the injustices of the existing society. Again, the injustices are real. But that empowers them to be in charge. The beauty of Marxism-Leninism, and why what we used to call the Third World loves this, is that they get to be in power. It empowers them across the board. They get to make the decisions on the economy. They don't have to submit to elections. They don't have to have a mandate. They don't have to legitimate their rule beyond the ideological building of a new world, of overcoming injustice.

What we see again and again is young people being impatient for evil to end, but also empowering themselves to be in charge. They love the state. They love the state as an instrument for social justice, social engineering. They love to empower themselves as the decision-makers because, after all, they're the intellectuals. They've studied the theory. They know better than others. Workers and peasants and the downtrodden, the lower classes, sometimes have false consciousness. They don't understand why, for example, we have an imperialist war. They get sucked in. Bread and circuses fool them. They have this false consciousness. But I know better, and I can be in charge. I can get us to a better place.

Even along the way, bad things are going to happen. Some people who are innocent are going to die or be arrested. But this is the march of history. This march of history is to peace and justice. Who is going to stand in the way of that? Especially when it empowers you personally so that… You could never do this in the private sector. Nobody could afford you this kind of power in the private sector and in a decentralized political system, in a federalized political system where nobody accumulates that much power. Social engineering is always coercive, always coercive. The issue is how much of the coercion is necessary that we accept in the tradeoff to right some of the things that are obviously wrong.

Dwarkesh Patel 00:51:36

All the anti-Marxism, I totally agree with. But still, just purely analyzing the system there's still something to be explained. There are many ideologies. You have this line that you can't explain Stalin by saying that he was beaten as a kid or he's a Georgian or whatever, because many other people are Georgians or beaten as a kid and they turn out not this way. There's many different kinds of ideologies, and very few of them end up as amenable to dictatorship as Marxism.

Another thing that's really confusing here is that all of these Old Bolsheviks who abet the system and whom eventually Stalin purges, whatever you might say about them, they're not weak men. They were willing to face down the Tsar. They were able to organize the revolution against the Tsar. They're willing to live in exile, to potentially get shot… I guess not shot by the Okhrana, but whatever. They're willing to go through hardships for their beliefs.

At the same time, you might think they might just go along with Stalin's doings because this serves what they think is the end goal of communism. But we know that after Stalin died, Khrushchev, who was one of the key people in the regime, gave a Secret Speech where he said that, “No, Stalin was going against it. He was destroying the building of socialism and the building of Marxism-Leninism.” So people did believe that Stalin was actually going against this end goal that they had. At least Khrushchev believed that. In many cases, they themselves are being implicated and they know they're innocent.

In many of these cases, there's this period in between when they're a dead man walking—because Stalin has started putting the feelers out that this person is a Trotskyite or something—but they're still in their positions of power. They're still the editor of Pravda or in charge of the military or something.

It's mysterious why these people who were able to… They're not cowards, they were able to organize a revolution against the Tsar. But they are not using this period that’s like a chicken with its head cut off in order to organize some sort of defense of themselves. Maybe the next time they have the Party Plenum, instead of just confessing or giving the obligatory speech where you're castigating yourself, you just say, “No, I think Stalin's leading the revolution wrong. I'm going to die either way, but I might as well say this.”

The same thing happened in China. Liu Shaoqi, when he's a dead man walking as the premier under Mao during the Cultural Revolution, he doesn't use that opportunity to go up… It’s very strange.

Stephen Kotkin 00:53:57

There are very few, but there are some people like that. They're arrested and executed, all of them. Very few of those kinds of people are going to survive. The ones who publicly decry the failures of the system and its perverse and unintended consequences, the ones who decry the dictatorship as opposed to the freedom that Marx predicted would happen, there are some people like that. They're extremely courageous. They're known to us. I'm not the only one, but I feature some of them in my book.

But again, you're creating a new world. It's going to be messy because it's about class struggle. And class struggle means there are going to be winners and losers. The bourgeoisie have to go. They're inherently evil. They're an evil class. They have to not just be retired to a farm somewhere. They have to be eradicated, liquidated as a class. They're going to resist and therefore there are going to be enemies everywhere. You're not going to know who they are. So are you on the side of the enemies?

Dwarkesh Patel 00:55:08

But what about the cases where they're implicating themselves? They know they're not an enemy.

Stephen Kotkin 00:55:12

Again, the party is a larger cause than they are. You're building a new world. Your life can contribute to that or not, and it's insignificant.

Dwarkesh Patel 00:55:23

What fraction of confessions by high-level party members do you think were not coerced out of a sense of fear of their own lives or their family's lives? I guess they knew they were going to be executed. So it would have to be for their family's life or to avoid torture vs. a very sort of Ozymandian, "I will sacrifice myself for the…"

Stephen Kotkin 00:55:44

We're in the realm of psychology here, DK. Human psychology is a complex subject. Figuring out human psychology is a big challenge, even with everything we know now, let alone what we knew then. So it's simultaneously everything and anything. We have to get rid of the binaries. They didn't believe, they were cynical, and they sacrificed themselves because they were cowards. Or because they knew that, they were forced to make that sacrifice to preserve some of their family members or whatever.

There are elements of belief and elements of, let's say, suspension of disbelief, and elements of cynicism and knowledge and understanding simultaneously in almost everybody. They coexist. We think of them as contradictory, but humans can hold contradictory thoughts simultaneously without too much trouble. We could give many examples. You've had people like that on your show, for example.

So the psychology is not as surprising in some ways. What's surprising is that this whole thing succeeds. It doesn't collapse of its own internal contradictions. It doesn't undermine itself. If you're murdering a high percentage of your upper officer corps… If you're murdering your intellectuals, your scientists, your cultural figures… If you're murdering your loyal party elites centrally and in the provinces… And if you're murdering the police who are carrying out all of these murders…

The thing about Stalin's terror is the police are also murdered during the terror while they are doing the murdering. You're doing all of that and the whole thing doesn't collapse. To me, that's more interesting in some ways than the complexity of human psychology that holds these contradictory thoughts and fails to go for self-preservation in some cases or fails to say, "I'm going to die anyway, I might as well go down fighting," or whatever the metaphor might be.

The fact that the system is able to undergo this level of self-disruption and come out the other side, that's pretty astonishing. Hitler does not murder his upper officer corps. He doesn't like them, he retires them, and they get a pension. He doesn't murder the Gauleiters or the Nazi party officials. He doesn't murder the intellectuals. Some go to prison, some go into exile if they're lucky. Some definitely are executed, often for acts that they've committed. Sometimes it’s just because they had an enemy in the system who wanted to enact revenge against them.

But for the most part, Hitler is attacking what we would call his real enemies, that is to say, people who are opposed to his regime, either in thought or in action or both. Stalin is attacking those people, but he's also attacking loyalists. He's taking down, in really big numbers, system loyalists, people who would walk through fire for him. One of the things they do is to walk through the fire of their self-immolation on behalf of the cause. This belief in the new world, in the better world, in transcending capitalism, in getting peace as well as abundance on the planet, in building paradise on earth, needs to be understood as absolutely fundamental to everything that we're talking about.

We often talk about Nazi racial ideology. How could they believe that stuff? That stuff is obviously ridiculous, you say. Goebbels, who helped enact the regime's ideology about a master race, had several deformities, right? Clubbed foot, walked in a brace. Yet, he's helping preside over the murder of people because they are disabled. They're singled out solely for their disabilities to be sent to the gas chambers or to be sent to imprisonment. And he's one of them. So we say, "Could he really have believed this? I mean, wasn't he a cynic enacting this? How could he not understand that these are real people? Because he's one of them."

The answer is, “Yes, he was a Nazi.” These people were communists and he was a Nazi. And there were a lot of them, and they had doubts. They suffered bouts of doubt. A lot of events contradicted the official ideology. Innocent people, innocent family members, themselves innocent, went to the gallows, got the bullet in the back of the neck. So you say, God, the belief. They couldn't have believed this stuff. And yet they did. The main thing that we know from the archives that were formerly secret, that we get into when they're declassified, is that the Nazis were Nazis and the communists were communists.

So here we have this problem with socialism. Socialism means many different things. A communist party is building socialism. Why? Because their view of the world is: feudalism, capitalism, socialism, communism. First they have to destroy capitalism to get to socialism, and then socialism can eventually get you to communism. So the communist party must first build socialism. How do you build socialism? It doesn't exist. So how do we know what it looks like? How do you get there? They don't know. The only thing they know is that it's not capitalism.

So let's destroy capitalism. That will be the step to get us to socialism. Capitalism has markets, we'll have planning. Capitalism has private property, we'll have state property or collective property. Capitalism has bourgeois parliaments where they vote and they claim it's democracy, but it's only for the property-holding bourgeoisie. So we will have a dictatorship of the proletariat. Everything. Capitalism has wage slavery. That is the story.

Dwarkesh Patel 01:02:23

And real slavery.

Stephen Kotkin 01:02:26

Exactly. Everything is to eradicate capitalism to get you to socialism. It turns out that doesn't deliver freedom, it doesn't deliver prosperity, and it doesn't deliver peace. It delivers massive state-ization. Because once you eliminate private property and individual choice, the state is now responsible for everything. It delivers ration tickets and the Gulag.

You get a bunch of socialists that break from this. They say Lenin is wrong. Eradicating private property, markets, civil liberties, and parliament is a mistake. We have to accept private property, markets, capitalism, and parliaments because that's the only way to get to freedom. Otherwise, you get to the Leninist dictatorship, total state-ization, Gulag, and ration tickets.

These people are denounced as revisionists, like Eduard Bernstein in Germany, for example. The Swedish Social Democrats, they say, “We accept capitalism, markets, and private property. We want to redistribute the income because it's tough for some people to make their way in the system. The system produces inequality. Let's make it more equal with social engineering redistribution. But we keep capitalism, we keep markets and private property, and we keep democracy, voting, rule of law, et cetera. We'll evolve towards full socialism and eventually communism, but we will not do it the Leninist way.”

There's this huge break in the socialist movement between those who are real revolutionaries and want to overthrow, eradicate capitalism to get to the just and prosperous and peaceful future, and those who want to use the existing system and evolve, embrace, and accept it. The left has a civil war, a civil war on the left which is still going on. It’s between those who say capitalism is evil and must go vs. capitalism has a lot of problems, but we need it in order to have peace and prosperity, in order to have freedom. We need to manage it better, redistribute.

This civil war on the left, which arises in real time, the critics of Lenin's revolution, the German Social Democratic Party, people like Bernstein and the rest of them, they are critics in real time of this. Yet, some of the critics who are what we would call the social democrats of Europe… Lenin was also a member of the Social Democratic Party of Russia, but the communist thing makes this divide between those who are serious about destruction of capitalism and those who are, “revisionists”, this denunciatory term.

What happens is that some of the people who are in the revisionist camp begin to flirt with the “capitalism is evil” analysis. They begin to truck with the communists that they've broken from and are in civil war with. You get left-wing social democrats who are closer to Lenin than they are to right-wing social democrats like Bernstein and the rest who are pro-capitalism but pro-redistribution. This is confusing to people because not everybody is a communist.

Some people, like in Sweden, accept private property and markets. But some of the people in Sweden seem to go back on that promise of accepting it, and argue that if we don't get rid of capitalism we're still going to end up with an evil system. This civil war on the left never gets resolved, it's ongoing, and the right uses this confusion to paint everybody as anti-capitalism. The left gives them ammunition by talking about the evils of capitalism, even when they've come along to accept private property and markets.

You have this deep and fundamental problem for the left. The tragedy of the left that it's never able to overcome, even to this day, where it comes out and says, “No more anti-capitalism ever, that is over. That leads to death, bloodshed, Gulag, ration tickets, war that actually is worse than the original problem it diagnosed.”

Think about Marx. Marx says that you get rid of private property, markets, capitalism, and you're going to get freedom, you're going to get abundance. You don't get that. People say, “Marx wanted freedom, he didn't want Stalin's dictatorship. It's not Marx who's the problem, it's Stalin who deformed Marx.” We see this argument all the time. Stalin is a deformity, whereas Marx was about freedom.

Think about a nuclear bomb. You're going to do a nuclear bomb. You're going to nuke a population, but you don't want to kill any people. Your goal is to nuke them, but nobody dies. That's what you say. You're going to get rid of capitalism, you're going to nuke them, but instead everybody's going to live. You give that order to your generals. You say, “Nuke them, but everybody lives, nobody dies.” They nuke them, and everybody dies instead of everybody living. You say, “I never said to kill the people, I said that they should live.”

But once you nuke capitalism, you're going to lose freedom, you're going to lose the ability to have politics, you're going to end up with some version of a Leninist system. The ideology is going to drive that to the doubters. You're going to get a second wind where you get Khrushchev, as you said. He comes into power, he denounces Stalin's crimes. He doesn't praise capitalism, private property, and markets. He doesn't undo collective farms. He doesn't undo state ownership of property. He doesn't undo the planning system. He just undoes Stalin's personality. He's trying to subtract… The de-Stalinization is to take away Stalin. It's not to take away any of the other attributes of the system. It's a second wind that it was Stalin who was the problem, not the system that was the problem.

So here we have the experience of going through the horrors, then having those horrors publicly denounced within the party. The Secret Speech is not published in Soviet newspapers, but it's discussed at party meetings in all locales. Within the Party, there's a public dimension to this. All party members become familiar with Khrushchev's denunciation of Stalin. They experience the horrors in many cases firsthand. They enacted the horrors in many cases themselves. They then see this denounced as horror, and they get facts that they didn't know. They get a big picture view back onto this.

Instead of saying, this system is evil, we made a big mistake, we have to undo state ownership of property, we have to undo collective farms, we have to undo the dictatorship… Instead of saying that, they say, “We get a chance to do it right this time without the evil Stalin who messed it all up.” The Khrushchev thing, the revelation of the horrors, the denunciation of the horrors, ironically gives you the second wind of belief in the system that's going to last right through Gorbachev, who's a Khrushchev-era baby.

We have this with Xi Jinping. If you know the story of his father and of his own upbringing, they suffer massively through Mao's regime and the Cultural Revolution. They're purged, they're humiliated. Yet, instead of saying, “This system is horrible, if I ever get power, I'm going to undo this system, which was so unjust to me and my family.” Instead of that, “They say, let's make this better. Let's not have the bad things that happened under Mao, but let's keep all the good things, supposedly good things that happened, including the Communist Party monopoly.”

That to us looks like the problem, and it wreaked havoc in their lives, in their family's life, but to them, that's the solution. This is a paradoxical element of communism where its failures don't become discrediting for so many of the people. They instead become a kind of second wind once you acknowledge and denounce them. It's never the system at fault, it's Stalin at fault. It's never the system at fault, it's Mao's mistakes or excesses as they're called, that are at fault.

Dwarkesh Patel 01:13:01

I guess in China's case, they actually did reform the system and they didn't just discredit the Cultural Revolution. They said no, much of the planning and state-owned enterprises was a mistaken idea. But I do have a different question.

Stephen Kotkin 01:13:14

Wait, that's not exactly the way you described it. You have a point. You're onto something, DK. But it needs to be qualified. What happens in Deng Xiaoping's case is that the communists have unwittingly destroyed the planning system. They have sent down to the village people who do economic planning. They've sent them to manual labor. They have smashed them in the face because they wear glasses, in many cases, and therefore they're putatively intellectuals. They've undermined their ability to continue the economic system as they had it.

Dwarkesh Patel 01:13:57

But if that's the reason why they weren't able to do planning, shouldn't Stalin's purges and then World War II have also had the same effect on the Soviet Union?

Stephen Kotkin 01:14:06

Not the political system. Deng Xiaoping never takes down the political system or the ideology. So you still have today the communist monopoly. Communism can fail at everything. It can starve the people, it can kill the people. It only has to do one thing to survive: suppress political alternatives.

So during that peasant resistance to Stalin in the collectivization episode that you referenced earlier, there's no political alternative. There's no other place for them to go and say, "We don't like the injustices of the tsarist regime and we don't like what communism is doing. Therefore, there's something else that we can go to that's an alternative." Communism has suppressed all the alternatives. It's either return to tsarism or keep communism.

In the Chinese case, you have something quite similar. They allow economic liberalization, in part because they have no choice, but they don't allow political liberalization. They're able to “reform” by enabling the people to generate wealth, jobs, prosperity through market behavior. It's mostly the peasant class in China. That then leads to family-owned businesses, which then leads to larger businesses. Society, not the Party, creates the miracle in China. The party tightens its grip because the ideology of the party is that when the socioeconomic base has a lot of market in it, it's a threat to the party's rule. The party has to be even more vigilant against the capitalists in the society.

It turns out that you get to Jiang Zemin, who is Deng Xiaoping's handpicked successor. Jiang Zemin sees that the private sector is becoming dominant in the country and that the party's monopoly on power is under threat. Jiang Zemin decides he's gonna do something called the Three Represents. He's gonna bring the millionaire capitalists into the party. He's gonna make them party members.

Instead of the party being against capitalists, the capitalists are gonna join the party, and this is gonna somehow increase the party's leverage and control and transform the psychology and behavior of the capital. Of course, it fails. Instead, the party members are in cahoots with the millionaires. They begin to form their own businesses by expropriating other people's property. The party begins to go dissolute in an anti-Marxist fashion in terms of private property, wealth accumulation.

So Xi Jinping comes along, predictably. He looks at Jiang Zemin's solution, co-opt the millionaires into the party, and sees that it failed. Not only did it fail to transform the behavior of the private sector people, it infected the behavior of the party people. Instead of bringing the capitalists into the party, he's going to force the party back into capitalism. He's going to push the party into the private sector more strongly than it was before. Board directors, party officials. CEO, party official. Private sector people who don't cooperate? Destroy them, make examples of them, including in the tech sector, so that people get the message that the party is the boss here.

You have a kind of natural progression where you open up the system economically in order to drive jobs, prosperity, wealth, because you've destroyed. People say the Communist Party brought 700-800 million people out of poverty. No, the Communist Party put those people into poverty. Why are a billion plus people in poverty? Because of the party's rule. It's the people themselves, they lift themselves out of poverty.

The Communists have to reassert their control, their Leninist monopoly on power, because the very thing that has rescued them—the diligence, entrepreneurialism, ingenuity of the amazing Chinese people and of that society—is now a threat to Communist rule.

Dwarkesh Patel 01:18:35

I agree with the mechanism by which the growth happened, but I don't think it's the case that it was their inability to have true Marxist communism which led to liberalization. If you look at the creation of these special economic zones, the imperative at a national level that you must have growth, and then Deng's Southern tour… Jiang Zemin, he tries after Tiananmen to clamp down on opening up. Deng says, "No, we must open up. If you don't, we’ll remove you." All of that is a sort of positive. Maybe positive is the wrong word, but…

Stephen Kotkin 01:19:20

Policy driven, that’s true.

Dwarkesh Patel 01:19:23

It's a special effort you have to make towards economic liberalization. It didn't just happen by default. People really had to push for it. The alternative story, which it seems you're saying, is “No, it was just that they physically could not enforce communism anymore.”

Stephen Kotkin 01:19:36

That's how it started, DK. Let's go look at the facts. Why do you have a Special Economic Zone? Why can't every zone have market relations?

Dwarkesh Patel 01:19:47

Agreed, but then the creation of that zone had to be a proactive action.

Stephen Kotkin 01:19:49

You’re doing this grudgingly. This is a grudging allowance of certain behavior. You look at the decrees. You can trade onions, but you can't trade potatoes. Okay, you can trade potatoes, but you can trade only on Tuesdays and Thursdays, not Monday, Wednesday and Friday.

Dwarkesh Patel 01:20:07

You can only have three employees per corporation.

Stephen Kotkin 01:20:10

Right, so the decrees are all grudging with very few exceptions. The society is forcing more and more concessions onto the ideologues in the Marxist-Leninist system. At the beginning, it's launched by the party's grudging acceptance that the society is gonna rebuild and not starve, through its own hard work. They've been through a couple of famines here, really big famines.

They don't have the state capacity to reimpose the system immediately in the economic sphere, so they grudgingly make concessions, very few but some, in the market sphere. Gradually, that expands over time as more and more people push against the system's restrictions. It's a policy-driven story in part, but not as the lead. It's a policy-driven story as the following of the entrepreneurialism and the hard work of the society.

Dwarkesh Patel 01:21:15

But that pushback is coming from within. This is your thesis in Uncivil Society, right?

Stephen Kotkin 01:21:19

Yes.

Dwarkesh Patel 01:21:20

It is coming from within the system. Because they could have… In 1976, they're like North Korea, literally. North Korea still exists. There's no reason…

Stephen Kotkin 01:21:29

North Korea doesn't have the Cultural Revolution where it annihilates its state capacity in a Maoist frenzy in order for Mao to, in his mind, undo his rivals, unbalance and destabilize them.

Dwarkesh Patel 01:21:44

But they have famines.

Stephen Kotkin 01:21:47

Yes, they do. But they still have the mechanisms of economic control and they have a massive black market. You’ve got to remember that communism doesn't have legal markets for the most part. It has restricted legal markets, again grudgingly. Household plots… But for the most part, it has a lot of illegal market activity, including in the state sector. The state sector gets an order to produce certain numbers of large quantitative output for the military-industrial sector, but it only gets allocated 25% of its ball bearings. It has to assign its supply department on the black market to go out and find the other ball bearings that it's not assigned from central planning.

You get a massive black market in the system, not just at the level of little people in the village, but at the top level of the military-industrial complex to make the system work. So when market behavior is grudgingly accepted, what that does is it brings market behavior out of the shadows into a legal or quasi-legal realm. You're not inventing market behavior from scratch. You're surfacing it in some ways. So party officials and industrial officials have market behavior in their firms to grease the system and to meet their output quotas.

Dwarkesh Patel 01:23:16

I agree with your general point that how any nation gets wealthy is not by the government but because of the thrift and entrepreneurialism and hard work of individuals. But that's also true in Western capitalist countries. In those countries, we also have a lot of stupid policies…

Stephen Kotkin 01:23:32

Yes, as we sit here and speak.

Dwarkesh Patel 01:23:33

When we say America is a capitalist country, what we could say is the government or all the bureaucrats, they'll try to put in all these regulations. It's only grudgingly that they will accede to… We could point to a bunch of stupid policies in America that are equivalent to where they try to outlaw the potatoes and the onions, but they could only outlaw the potatoes. So any “capitalist society”, is just a case where the government had to cede some amount of control. We give credit to capitalist countries in the West for saying, at least the government wasn't maximally stupid.

Stephen Kotkin 01:24:07

Fair point, DK. We often exaggerate the role of policy in all realms because we do policy ourselves or we talk to policymakers. We have a bias towards the causality of policy.

America has been 25% of the global economy since 1880. We've had no income tax, we've had income tax. We've had high income tax, we've had low income tax. We've had tariffs, we've had fewer tariffs. We've had all sorts of regulations, we've had deregulation. For 150 years, more or less, we've been around 25% of global GDP, 5% of the population, through every imaginable variety of policy regimes.

That doesn't mean that policy is inconsequential. It matters for a lot of players in the system. It matters for those who get the policy turned in their direction, the subsidies or the tax breaks, or the taxes on their competitors or whatever it might be. There's a lot of gaming of the system and it does matter. But in the larger picture of things, you can't create the wealth of the United States over those 150 years, that global economic dominance, and you can't strangle it, in the policy realm. You can affect it, but you can neither create nor strangle it.

We have to understand in the communist sense that incentives matter. When you create incentives for officials to increase GDP and to increase job creation, and that's how they get rewarded, you're going to get a lot of that behavior. The Party will do that, not immediately.

Remember, immediately they're kind of flat on their back. They've had this Gang of Four. Deng Xiaoping has come back from having been purged. They're on the verge of another potential famine. Per capita GDP under Mao is $200 during the Cultural Revolution. $200 is the annual per capita GDP of a billion people, or slightly under. You think, “that's insane”. That's where all the people are in poverty when I was saying the regime put them in poverty.

They're a little bit flat on their back, which creates an opening. It creates this grudging dynamic of “we're going to hold power and we're going to allow economic entrepreneurialism to take place, but we're gonna control it. We’re going to control it with special economic zones…”

Dwarkesh Patel 01:26:44

In any country in the world today where there's a lot of poverty, the reason the poverty exists is also because of policy. To the extent that poverty has been removed, it is because of some combination of human capital and policy got less stupid. If we're going to complain about a country like Bangladesh being poor—and there are many poor countries in the world… Maybe we're going in circles here. A different question I wanna ask is…

Stephen Kotkin 01:27:14

No, you have a point. We're not disagreeing. I’m just trying to say that we give too much credit to the Communist Party for what's happened in China and not enough blame for what's happened in China. This is part of the dynamic of us seeing communism as potentially successful.

We criticize these fools who thought that the Stalin regime was not going to kill them, and was not going to produce famine. Yet we have this narrative that the Communist Party produced an economic miracle in China. I’m sorry, the Communist Party took advantage of the economic miracle in China. It played a part in it, expropriated the hard work of many people, and stole the businesses. A lot of those local officials just stole the land and stole the businesses from people who'd created a success.

This is the thing that the Party did that's really important. Deng Xiaoping first went to Japan in early '79, before he came to the US and met Carter and put on the cowboy hat, that gigantic ten-gallon hat that was bigger than he was. He was like five gallons, the hat was like ten gallons.

He goes to Japan. You're looking at Japan, DK, and it lost the war. It was literally incinerated in the American use of atomic weapons. It was destroyed, it lost the war, and it's rising to be the second-largest economy in the world. What happened? How was that possible? How could Japan rise from the ashes, literally, when China won the war? It was on the winning side and it has a $200 per capita GDP.

Deng Xiaoping looks this over, and he says the answer is that Japan is partners with America, not with the Soviet Union. Deng Xiaoping is going to divorce the Soviet Union economically, and he's going to marry the US. Deng gets most-favored-nation status in 1980, thanks to Jimmy Carter. The communist regime in Beijing gets most-favored-nation status, which has to be renewed every year and is renewed every single year until 2000-2001, when they're admitted to the WTO. That's a Clinton initiative that happens right when Bush is going to come to office.

The secret sauce is, you have to manufacture and export to the American domestic market because the American middle class is insatiable. They will buy anything as long as the quality is high and the price is low. Japan did this. Japan's two former colonies, Taiwan and South Korea, followed in Japan's footsteps. China's going to do this too. We're going to use this Japanese model and the American middle class, and their insatiable overconsumption, is going to create the Chinese middle class. So this is what the party does, this geopolitical reorientation from a Soviet economic model to a Japan-style export-led partnership with the US domestic market and middle class.

They have a couple of tricks that are really important. They have Hong Kong. Hong Kong is a British-controlled, rule of law, international financial order that allocates capital based upon risk and return, not Communist Party dictates. Gorbachev's Soviet Union, they have nothing like Hong Kong. The only reason China has Hong Kong is because after World War II—when Truman announced that Chiang Kai-shek and the Nationalists were going to accept the Japanese surrender in Hong Kong—the British sent their boats in and took Hong Kong back themselves, so that when Mao defeated Chiang Kai-shek he didn't have Hong Kong. The British had Hong Kong, and they created this international financial system that Mao's successors would be able to use. Had the British not done this, there would be no Hong Kong and there would be no Chinese miracle in the Deng Xiaoping and after period.

The other thing they have is overseas Chinese who know the culture, speak the language, and are going to do the FDI, again routed through Hong Kong. So you have Taiwan. Ironically, the failure to win the civil war 100%… We think of the Korean peninsula as divided. We think of this partition of the Korean peninsula, but China is also partitioned. There's also Taiwan, still partitioned to this day. Taiwan is the FDI that's going to come in, through Hong Kong, routed through into the special economic zones on a risk-reward, capitalist market basis, not a communist basis.

Furthermore, they have the Japanese war guilt because the Japanese committed those atrocities in China. So the Japanese are going to make up for what they did by helping rebuild China. Again, the FDI and the tech transfer, like it's coming from Taiwan. So Taiwan and Japan, this partitioned China, and this war guilt, through the British Hong Kong is going to go into the special economic zones, manufacture things like the Japanese, and then export them to the American domestic consumers. It's going to be t-shirts at the beginning and then they're going to, like the Japanese, and then they're going to climb the value chain until it's got the highest value-added products. Then supply chains are going to change as a result, and nothing is made in one place anymore, and the world gets very complicated.

The point being is that Deng Xiaoping did that. That was intentional. That's the credit that the party deserves and the party never gets because it's a story not of the party's rule alone, but of British Hong Kong, Taiwan, Japan, United States domestic market, etc. So how do you get rich in the modern world? You sell to America. Those countries that are partners with America and that are able to compete.

The Chinese deserve credit. They can manufacture higher-quality, lower-cost products that American consumers will buy. American consumers are not forced to buy Chinese products. They're just better, and they're cheaper. So they buy them because it's a market and there's competition. Bangladesh can do this in textiles, you referenced Bangladesh, which is how I got launched on this reverie that I hope is now ending.

My point being is that the East Asian miracle is Japan selling to the American domestic market, followed by South Korea and Taiwan doing the same trick, and then followed by Deng Xiaoping's Communist China, the same exact trick, filtered through British Hong Kong. The problem with the Chinese one is that they're not allies, former enemies who are allies like Japan. They are former allies who are now enemies who have done this magic sauce. Now we're in the pickle that we're in as a result of this. But the formula… This is where the communists deserve the credit that they never get, whereas they get credit for things that they didn't do.

Dwarkesh Patel 01:35:12

Suppose Stalin had lost the succession battle in 1924 and somebody else is in power, but he's still on the Central Committee or the Politburo. It's 1930 and suppose the other person is also ruthless and is one by one getting rid of every single person in the inner circle. What would a Stalin-type figure have done if he found himself on the periphery of somebody else's regime?

Stephen Kotkin 01:35:39

Counterfactuals are really critical for historical thought. A lot of historians are pedantic about this. They say we're against counterfactuals, it's just speculation. But every single one of them is a practitioner. Why? If you say that Stalin caused collectivization, that means without Stalin there's no collectivization. If you say Hitler caused World War II, you're doing the counterfactual. You're saying no Hitler, no World War II.

Dwarkesh Patel 01:36:12

Just to put a finer point on my question, I mean it not just in the sense of whether collectivization would have happened, but more in the sense of, how would he personally have avoided the fate of Bukharin and Kamenev and Zinoviev? Potentially, he’s thinking, “I'm going to get purged someday, I don't want to be the toady to somebody else.” How would he personally have navigated the sort of power struggle at being what Zinoviev was to Stalin or Bukharin was to Stalin?

Stephen Kotkin 01:36:39

This is a question about how do you become Stalin? Could there have been another Stalin besides Stalin? A lot of people will argue that you have this formative period when you're growing up, your parents, your schooling, the influences of your peers, and you become a certain personality. Part of it is genetic, and then a lot of it is the environment, and you have this then personality. So you have to understand how the person formed, whether it's Picasso as a painter or Stalin as a dictator. Then if you understand their personality, you'll understand what they do in power.

The problem with that analysis is that Stalin is not Stalin when he first gets into power. It's the experience of being in power that makes Stalin, Stalin. It's the building of the dictatorship within the dictatorship, and it's the enacting of that kind of power that makes Stalin who he is. It's sitting in that chair. It's being in the Kremlin, running a Leninist regime, and being responsible for Russian power in the world against Nazi Germany, the UK, the US.

People say about Xi Jinping now, “You know, Xi Jinping has made a lot of mistakes. If he had just kept to Deng Xiaoping's policies, China would be much better off. We'd still be in a kind of détente or partnership with China. Instead, we're at loggerheads and there's potential war.” The problem with that analysis is, “What would Xi Jinping have done if he were the number one guy under Deng Xiaoping instead of Deng?” Maybe he would have done Deng's policies just like Deng did. More importantly, what would Deng do if he were alive today instead of Xi Jinping? Would he do what Deng did in the 80s and 90s, or would he do what Xi Jinping is doing today?

In other words, how much is the personality and how much is the system? How much is formation before you get into the position of power, and how much is the circumstances and responding to those circumstances and the exigencies of the moment and the way the system operates and the place the system is and what the larger context in the world looks like?

Here you have a Communist party that seizes power. As I said, unlike the case of Bavaria, southern Germany, northern Italy, Hungary, it holds power. It doesn't just seize power. The Paris Commune, 1870-71, they seized power in Paris. Then they were destroyed, put up against the wall and shot. They seize and they hold power.

But they're in this peasant country, and the peasant has the land de facto and they're Marxists. They believe that the base, the socioeconomic base, the class relations, determine the superstructure or the politics. In Marxism, the base, the socioeconomic base, gives you the superstructure. Politics depends on it, it's an outcome of what the base is. So you have a de facto capitalist base.

Dwarkesh Patel 01:39:57

I don't mean whether they would have done collectivization. I mean how would he personally… Because he wants power.

Stephen Kotkin 01:40:02

I get that. He makes this decision. All of them want to get rid of capitalist relations in the countryside. Every single one of them wants to do that. They're all communists, they're all Marxist Leninists. But they don't think it can be done. They think if you try it, you'll fail. He goes and tries it. He creates even more destabilization than they had predicted. But he just powers through and gets there to the end and succeeds. Most of them are grateful that he's pulled this off because they thought it couldn't be done.

Your question is, “who else could have become Stalin in that position?” Who else among the Marxist-Leninists could have been the guy who says, “Whether we can do this or not, we have to do this because we can't have a socioeconomic base that's capitalist and have the communist regime survive.” They were all ready to say that, but they weren't all ready to do that. Moreover, after he did it, they had criticized him during the process while he was doing it.

He was the only one in his mind who was Marxist-Leninist enough to get it done, and they were all carping at him. This is where you begin to see the paranoia, suspicion being magnified, where he then, in a few years after collectivization is more or less finished, he's going to go after them. So you needed a person who could have felt in their head that this was not only necessary, but doable. Someone who would undertake those risks, power through no matter how much famine and resistance and upheaval and criticism there was. Someone who would then come out the other side of that as the victor with this gigantic secret police that was really small but got really big in the process of doing the very thing that people said you couldn't do. Then having all of that and not destroying any rivals and going from dictatorship to despotism…

So you would need a person who was capable of being Stalin in that group, not from outside that group but inside that group. Then that person would either not use that power to destroy everybody else—not yearn for despotism but be satisfied with dictatorship where others exercised power and their domains—and yet still be able to hold on to the system. Was there such a person in the circle? If the answer is no, could that person have emerged in the process of doing it?

Dwarkesh Patel 01:42:53

I guess my question is slightly different. Even if such a person did not exist, suppose Stalin already exists, he did all the stuff, and it's 1934. It seems like Stalin's starting to go a little Great Terror-y soon. Another copy of Stalin is in the Politburo. Just out of a sense of self-preservation, they're like, “In a couple of years, I don't want to be writing my own confession and ending up in the Gulag.”

Is Stalin, being the power player that he was and knowing how to align factions against each other to his own advantage in the very end, if somebody like him was in the Politburo, what would they have done? Or were they already there and there was nothing they could do by this point?

Stephen Kotkin 01:43:39

This is a question for every single dictatorship. Why didn't somebody just kill Stalin? He was gonna kill them, kill them all. Why didn't they just kill him and save themselves?

During the 1920s, Stalin resigned six times, three times in writing and three times orally, between 1923 and 1928. Every time those guys around him beg him to stay. Not only do they fail to try to push him out, but when he himself volunteers to go out, they beg him to stay. Then he kills all of them within 10 years. Every one of them, just with a few exceptions, is dead.

Jeez, so what were those guys thinking? It's clear that Stalin was not Stalin yet. If they knew in the 20s that in the 30s he was gonna murder them all, maybe they would have acted the way you said. He becomes Stalin in this process. He's not Stalin yet. That's a really important argument that I make in the book.

Today you look at Putin. Putin is ruining Russia. Why doesn't somebody just assassinate him? Xi Jinping, he's hurting China. He's making China enemies everywhere around the world when China was, until recently, popular. China was 75% favorable globally. Now it's 25%, more or less, favorable globally. That's Xi Jinping's doing. How can the elites around him let him do that?

Around Putin, the people are falling out of windows. Instead of falling out of the windows themselves, why don't they push him out the window? Khamenei in Iran, he's brought ruin on the country. Why don't they take him out and try to save themselves and save the country, not just themselves? In other words, they can be patriotic as well as securing self-survival. The answer is it rarely happens. One, you have a collective action problem.

Dwarkesh Patel 01:45:34

Sorry, why doesn't this prevent… People are trying to kill the Tsar constantly. They're killing Russian ministers in the tsarist regime.

Stephen Kotkin 01:45:44

There are more assassination attempts on Hitler, some of which come very close, than by far on Stalin.

Dwarkesh Patel 01:45:49

Also, Mussollini, but never against Stalin. Why is that?

Stephen Kotkin 01:45:52

Stalin is the guy who is building and personifying the system. The people around Stalin can see that he is unusually good at dictatorship. He is just carrying this entire system on his back through thick and thin, killing enemies, liquidating the kulaks, collectivizing agriculture, building a military-industrial complex, defeating Hitler in war. I mean, how much better from the system's logic, not from humanity's logic, not from the point of view of the kind of values that you and I share, how much better are you going to do than Stalin?

So there is a way in which they're pygmies and he's Stalin. Of course, they know that they can't do what he does. If they tried to unseat him, they might save their lives but they might lose the revolution, the system and the radiant future, the overthrow of capitalism, the abundance, peace, paradise on earth. That's a big move to lose if you believe in that, if your life is about that and you're dedicated to that. But in addition, you have a collective action problem that's really important.

Let's suppose that I'm in a Stalinist regime with you. You're a functionary and I'm a functionary. Stalin's collectivizing agriculture, which means he's destroying productivity, and we're going to be poor. All the way through the Brezhnev period, we're gonna be importing wheat even though we have this gigantic agricultural belt, wheat belt. Some people knew in real time that this was self-harm. I come to you and I say, “This Stalin guy, he's wrecking everything. We got to take him down.” You agree with me. But you know what? You don't know. Maybe I've been sent by Stalin to test your loyalty. Maybe I'm provoking you to reveal your disloyalty. Maybe I'm not being sincere.

You agree with me, but instead of saying yes, let's do it, immediately you run to Stalin and you say that this Kotkin guy is talking behind your back about how we need to take you down. It’s self-preservation, you're going to preserve yourself because you don't trust. There's a lack of trust inside these dictatorships. If you knew that Stalin hadn't sent me for sure 100%, you would say, “You know, you're right. You got a point there. What can we do about this?”

But you know that Stalin is constantly doing these provocations, or you suspect he is. You know that he's got people, provocateurs who are around the system doing things like this. The secret police are listening in on your phone conversations. The driver of your car works for Stalin, doesn't work for you, and is reporting any overheard conversations in the car. And the maid in your apartment is also working for the secret police and reporting up the chain of command.

So the system that you're in enmeshes you in this distrust, in this surveillance and distrust. So what looks like, “Geez, let's just take him down and save our own life, let alone save the country.” Yes, it's logical, but that's not the kind of life that they led. We would think that based upon the kind of lives and the system that we live in.

Dwarkesh Patel 01:49:37

But there's a bunch of revolutionaries who try to kill the Tsar and sometimes succeed. They're just random people. They're not people in the regime, they're just random people.

Stephen Kotkin 01:49:45

Yes.

Dwarkesh Patel 01:49:46

Why doesn't a kulak, one of the hundred million enslaved people, go out and…

Stephen Kotkin 01:49:51

The Tsar has less security than Stalin does.

Dwarkesh Patel 01:49:53

Didn't you say that in 1928 he had like one bodyguard when he would go to his dacha?

Stephen Kotkin 01:49:58

The bodyguard stuff increases over time, but the regime is walled off from the people. Stalin doesn't go out in public. He's not one of these populist types in public who's bathing in the adulation of the crowd. He's in the office, he's at the dacha, he's at the party meeting, he's at the party congress. He's not putting himself at risk.

Even so, it is paradoxical because when Hitler goes to make a speech, every year Hitler makes a speech in Munich. It's known when he's going to make the speech. It's announced in the paper. There are a couple of assassination attempts on Hitler, one of which takes place in the hall where he's going to do, where someone plants a bomb. It's a working-class guy. He plants a bomb there and the bomb goes off, it blows up. Hitler left the hall more quickly than anticipated, based on the schedule that people thought he would be there longer. It was quicker. He was out and the bomb exploded and he survived.

There are military officials who tried to kill Hitler, famously, in 1944. They plant a bomb under the table which also goes off. It almost gets him but doesn't get him, during a military briefing. There are attempts on Hitler's life both from the society and from inside the regime.

Stalin doesn't have this. In fact, the people inside the regime are killing themselves. When they see that Stalin is leading them down a blind alley of murder and ration tickets in Gulag. They kill themselves rather than kill Stalin. Again, there's something special about the mentality of these communists. There's also something about Stalin's success as well as the threat that he represents to these individuals. Still, it is mysterious because there were opportunities and people didn't take up the opportunities.

Very few… There was no serious assassination attempt on Stalin. The very few times when they accused somebody of doing an assassination… There were shots fired at a boat when Stalin was on holiday in the south. It was not because Stalin was in the boat. It was because the boat was not in the system as marked as allowed to use that waterway. They were just performing their duties as border guards shooting at the boat. It got dressed up as an assassination attempt and people were arrested and executed and it was publicized as such. They didn't know that Stalin was in the boat.

Dwarkesh Patel 01:52:46

You've written other books about the collapse of the Soviet Union. There's this last ditch effort in the Eastern Bloc where there's falling productivity, to borrow more money, invest more into finding this last ditch technological miracle that can cure all their problems.

How similar is that, in your opinion, to what's happening in China? The dissimilarity is that while Eastern Europe was struggling to export and they had a trade deficit, China, many people argue, is exporting too much. Do you see any similarity between where Eastern Europe was in 1989 versus where China is today? Or are you not as concerned about China right now?

Stephen Kotkin 01:53:25

It's a very difficult question to answer very briefly, but it's a really important question. There are tremendous differences, of course, civilizational differences let alone system differences. We wouldn't want to elide those differences.

What's similar is the Marxist-Leninist monopoly on power. People ask me, is China a Marxist-Leninist regime? I usually say it's a Leninist regime because that's undisputable. Do they believe in Marxism or not? People disagree on this. How many true believers are there? Is Xi Jinping a true believer? That's a difficult argument to win because the evidence is contradictory and because we don't know the inside of the system. It's still in power, it hasn't fallen. But it's clearly a Leninist system.

A Leninist system can't be half-pregnant. You can't be half-communist. You either have a communist monopoly or you don't. What happens in communist political reform, not economic reform where they allow some market behavior, but where you liberalize the system and you open it up politically… You say, DK, let's have debate inside the party. Let's allow pluralism inside the party's rule, meaning we're going to keep the party but we're going to allow different tendencies in the party. The problem with that is some wiseacre raises their hand and says, “I don't want the party. I want another party. I don't want the Communist Party. I want a Social Democratic party, or I want a right-wing party, or I want a centrist party.” And you say, “No, no, no, that's not the rules here. The rule is only debate within the party's monopoly, not that you can have another party.”

What happens is there's no way for them to open debate and then to put a lid on the debate. It becomes a Pandora's box. You can't be half-communist, half-monopoly. You either have the monopoly or you don't. Every time they liberalize politically, the system liquidates itself. Hungary in '56, Czechoslovakia in '68, and Gorbachev. Had Gorbachev not happened, China might have done its own Gorbachev. Had they not seen Gorbachev accidentally liquidate the party, they might have done political reform, opened up the party, and watched the thing unravel. Or they would have had a crackdown and put the lid back on much bigger than the Tiananmen episode in June 1989.

They're not going to do political reform because they know from studying the Gorbachev case, which everybody studies in party school where all the cadres have to go to be trained. They're not going to open up the system politically because that's suicide. They're not going to commit suicide like what happened with Gorbachev. This means their policy options, their menu of policy options, is limited. They can open up the economy, but if it gets too open and too liberalized, too many people with independent sources of power and wealth, too many Jack Ma’s, they lose control potentially.

They can open up the system economically, but then they have to somehow reimpose controls. But if they reimpose too many controls, the GDP goes down and they don't have the job creation. You have this constant back and forth of how much economic liberalization you can have before it becomes a threat, or how little economic liberalization you can have before it becomes a threat to your ability to create jobs and wealth.

That's the dilemma they're in. The Soviets in the '70s and '80s were looking at the system. They didn't want to change the system Gorbachev-style. They didn't want to liberalize it politically. They were willing to introduce some market economic liberalization, some market incentives. They tried that in '65, it actually didn't work. Then the Prague Spring happened in '68 and scared the bejesus out of them. Reform looked like the end of the system. They tried a little bit of economic liberalization. It didn't work. They didn't want to open up the system politically. So what's left? Technological fantasies.

Maybe technology can perfect planning. All the pathologies of the planned economy, all the inefficiencies of the planned economy can be overcome with computers. Maybe if we invest heavily in tech, we don't have to make the hard choices of deep and fundamental structural change which would end our party's monopoly. We can keep the party. We can keep the party's monopoly. We can even keep the state-owned economy, but we can just tweak it with the tech and supercharge it or even turbocharge it and make it work that way. Computers, tech, will save us from the hard choices of deep structural reform which will threaten our power.

We know how that worked. It didn't work. Now you're looking at China today. You have a Communist Party monopoly. You can't be half-pregnant. You can't open up politically. How are you going to reintroduce dynamism? How are you going to get the GDP growth? How are you going to get job creation? How are you going to get societal buy-in? If you go too far in that direction, that could threaten the regime. If you don't go far enough… So we got tech.

The tech can make our dictatorship function better. Not just our economy, our productivity, our job creation, managing through the demographic crisis, not just the economic and social benefits from tech that tech could deliver…Tech could maybe make our communist dictatorship immune from challenges because surveillance is all-encompassing, because of our ability to spot things before they happen. So you can see where they would be so seduced, so tempted to believe that tech is the solution.

Here's the problem with that argument. First, it didn't work the last time. That doesn't mean it won't work this time, but the track record, even though it's a small number of cases, is not good. But the other problem is political legitimacy. You can't get political legitimacy. You might have thought that, "Oh, geez, if the GDP grows, that'll give us legitimacy." Then the GDP stops growing and you no longer have the economic benefits to claim that that's why you're in power.

Dwarkesh Patel 02:00:23

But do you need it? Stalin didn't have strong growth in the '20s and '30s, and it seems like you just double down on repression. If you double down on the NKVD… The Tsar actually had 2% growth up until 1917.

Stephen Kotkin 02:00:39

They're dead, and their system is gone. It's fundamentally a deficit of political legitimacy.

We talk about Iran today and how Russia and China didn't even help them with any military support or economic support while they're under tremendous strain from Israel rolling back Iranian power. So they kind of got betrayed by their strategic partners. The strategic partnership among the authoritarian regimes is a fake. That's true. It is a fake. They're out for themselves. They're opportunistic, and they will help the others to the extent that they feel it's helping for themselves. The day that they feel it's not helping themselves, forget it.

There's a deeper problem there. What the Iranian regime needs is political legitimacy. That's what it doesn't have. It's not just a failure economically. It's not just a failure in security. In its foreign policy terms. It's hated by its own people. It's got maybe 20% support in the population. A lot of people are indifferent but a majority of the people despise this regime and want to see it go. They're patriots for Iran, but they detest the mullah's clerical regime.

Neither Russia nor China can give political legitimacy to Iran. They can give dual-use technology. They can give them missiles. They can give them anti-missile defense. They can never give them political legitimacy. That's the vulnerability, which is why Iran is on the precipice now. Because the regime is illegitimate and the regime knows that it's illegitimate in the eyes of the people.

Dwarkesh Patel 02:02:24

I guess we'd like to think that's what matters. But historically it just seems like when authoritarians crack down really hard, it kind of just works. Right?

Stephen Kotkin 02:02:33

That's why they have gigantic repressive apparatuses. People talk about a bargain. The Chinese made a bargain or the Russians made a bargain. The people gave up their freedom and the regime provides a higher standard of living. So there's this bargain. There is no bargain. Because if the regime fails to raise the standard of living, the people can't sue them in court. They can't say, "You didn't live up to your end of the bargain. We gave away our freedom, but you didn't deliver on your part of the bargain, so the deal is over. You're out of power now because you didn't live up to your bargain."

Instead, they repress. They bring out the batons. They bring out the water cannon. They bring out the disappearances where people are arrested in cause. They're not even indicted, they just disappear them. You have this huge repressive apparatus, and it seems to work, especially when you have this moment where you fail to live up to some of the promises that you made. The challenge for them is that somebody has to do the repression. The repressive apparatus is not a machine. It's not AI. It's not something which is mechanical. It's people. It's people who grow up in neighborhoods, come from villages, and went to schools with other people.

This brings us back to the tsarist regime where we started. The tsarist secret police wasn't big enough to keep the lid on, so they had to use the military. The military was a peasant army. They were peasants in uniform. The working class, including women, was striking for bread in the capital and marching in the capital for bread in 1917. The military was told to shoot them. These are elite military units. Shoot these workers. They're peasants, and you think, “Okay, peasants, they'll shoot workers. No love lost.” The workers were peasants yesterday. Some of them were still peasants who went back to the villages where these soldiers were from, during downtime at the factory, during harvest time. They were the same people. The army decided not to shoot.

There came a point where the regime called out the repression, and the repressors had agency and didn't repress. That's what happens in these cases. You never know when it's going to happen. It's very hard to predict beforehand. But there comes a moment where the people who are supposed to do the shooting decide not to shoot. The people who are supposed to do the arresting decide not to arrest. The people who are doing the surveillance decide to stop. You have this huge repressive apparatus, and it works until the moment the people in it decide not to do it anymore.

That's where the political legitimacy variable is ultimately decisive. Because those people who were killing under communism and collectivizing those villages and dekulakizing and killing the people in their native village that they had grown up with… If they hadn't done that, the regime couldn't have done this. The regime couldn't have collectivized. Stalin wasn't out there shooting people. He was signing the decree, which then got communicated through the system to the point of the activist who enacted it or not, as the case might be.

When they don't enact, your power evaporates. It's like a bank run. You look at the bank, and you think, "Wow, that's pretty solid. They got these neoclassical columns. It's all made out of stone. Looks really impressive. This bank is really solid." One day, it gets in people's heads that the bank might not have your money. The bank might not be good in reality for your paper holdings. You rush to the bank to try to get your money out before that money is gone or doesn't exist. Everybody gets that idea at the same time, not just one person, and you get a bank run. This really solid institution with these neoclassical stone columns turns out to evaporate, evanescent.

You can have a political bank run in the repressive apparatus. They cease thinking that they should kill people like themselves on behalf of a system that they are no longer loyal to, no longer adhering to. That can happen in the forces of order, as we call them, in the repressive apparatus. It can happen inside the elite. Because when the leader gives the order, it's got to go through the whole chain of command. The leader doesn't give the order to the soldier. It goes to the one boss, the subordinate to that boss, the subordinate… Anywhere along the chain of command, there can be disloyalty and revolt. There can be what we call political defection.

What motivates, what triggers, political defection? The lack of legitimacy, political illegitimacy. People are not going to die for something that they no longer believe in.

So that's a really big problem that the communist regime can't solve with tech. You say, “Well, the tech could produce power and prosperity and China could relegitimize its rule just like it did with economic growth. The economic growth did it for 30-40 years and that's how the regime legitimated itself. The tech will do that for the next 30-40 years.” The answer is, that hasn't happened yet. Maybe it can do that and maybe it can't, but it's never permanent.

What's permanent is power rooted in the people. What's permanent is their real citizens. They have real freedoms. They have the right to vote. Now, they can't get what they want. They go to the polls and they see bad candidate, worse candidate, even worse candidate. But they can punish what they don't like. They can exercise that agency. They can be citizens. They can realize their citizenship, just like we do as consumers in the marketplace, consumers of podcasts. That's where you get legitimacy from, where the system enables people, opportunity at home, opportunity for people who otherwise don't have opportunity. That's legitimacy. That's priceless. China doesn't have that. Russia doesn't have that.

Stalin had that legitimacy for a time based upon the idea that he was building a new world and overcoming the horrors of capitalism and imperialism and world war. Khrushchev gave a second wind to that, even as he was revealing more of the horrors. Then that just ran out, and they unwittingly destroyed the system trying to give it a third wind under Gorbachev.

So they have no way forward. They're stuck. They can't do structural reform and maintain their power, but without structural reform, without a legitimate system, they also can't maintain their power forever. It's really interesting. In the short run, we're all dead because there could be a World War III. But in the long run, we're all good because our system is better. So we have to elongate the short run. No world war between the US and China. Get to the long run, get to the competition, get to the cold war instead of hot war, where we're not having a hot war.

That's the beauty of cold war. It's not hot war. You can compete, you can have tensions, you can have rivalries, but you don't have hot war. So in the short run, potentially, we're all dead, because a world war with great powers… It was 55 million in World War II, the low estimate of deaths. That was a multiple of World War I, and World War III would be a multiple of World War II. If we can avoid that, in the long run, we're good.

It's the opposite of what Keynes said about how in the long run, we're all dead. It's in the short run that we're potentially dead. But I like the long run. So the tech and China thing might work, and it might not work, but it's not permanent even if it does work.

Dwarkesh Patel 02:11:28

All right, great note to close on. Thank you so much for coming on the podcast. It was a real pleasure to talk to you.

Stephen Kotkin 02:11:34

My pleasure. I apologize for not being succinct in my answers, but if you've read some of my books, not all of them, some of them go on at length. A few of them are short, though. I have to master the answers to questions on podcasts in order to be able to get through your whole magnificent list. Maybe next time we'll do better.

Dwarkesh Patel 02:11:55

The long drawdown is why I want you on the podcast. A lot of these issues are complicated, so I appreciate you doing it.

I’ve had a lot of discussions on my podcast where we haggle out timelines to AGI. Some guests think it’s 20 years away - others 2 years.

Here’s an audio version of where my thoughts stand as of June 2025. If you want to read the original post, you can check it out here.

I just interviewed historian Stephen Koktin yesterday. The full episode should be out next Thursday on my podcast.

I spend a long time researching, and it's impossible to explore more than a small fraction of my curiosities during the interview. This was especially true for Kotkin. These 1000-page-each volumes basically compile all of late 19th/early 20th century history. Naturally, reading them inspired lots of thoughts that I wasn't able to exhaust during the interview itself. So I thought I'd just release my notes publicly.

You can buy volume 1 here and volume 2 here.

One of Kotkin’s interesting takes from this book:

Even the package of attributes that we call modernity was a result not of some inherent sociological process, a move out of tradition, but of a vicious geopolitical competition in which a state had to match the other great powers in modern steel production, modern militaries, and a modern, mass-based political system, or be crushed and potentially colonized.

Part 1 starts by laying out the geopolitical context the tsarist government found itself in during the late 19th century. Bismarck had unified Germany and had put it at the leading edge in key modern industries like steel and chemicals. Britain controlled the seas, kickstarted the industrial revolution, and had a global empire. Japan was going through the Meiji reformation, whose implications would become clear to the world when they beat Russia in a war in 1905.

Meanwhile, what would happen to you if you failed to modernize was made amply clear by the example of Qing dynasty China. What had been the foremost power for almost all of recorded history now couldn’t even prevent a tiny island all the way across the world from forcing opium into its markets.

Russia, like every other country, wanted to avoid such a fate. Its long land borders rendered it permeable to invasion from almost all of Eurasia. Imperial Russia had expanded 50,000 square kilometers per year for 450+ years (to put that in perspective, that's adding an area the size of Slovakia or Costa Rica every single year, for four and a half centuries). This is the classic continental power trap that Sarah Paine talks about. You have to conquer more territory in order to defend the territory you already have. What this meant is that by the end of the 19th century, Imperial Russia had dominion over dozens of different nations, including Georgia, from which Stalin hailed.

China, of course, was also a very big country, but that didn't help much when the more industrialized British came knocking. Russia knew it was at least a generation behind. By 1900, Russia had the world's fourth or fifth largest industrial power. However, the vast majority of the population remained rural peasants living in conditions that had changed little since emancipation.

Russia had a Bismark like figure in Finance Minister Sergey Witte (1892-1903). He kickstarted Russia’s belated industrialization, but failed to prevent a war with Japan which Russia lost, and which was catastrophic for the Tsar’s image. In 1905, the tsarist government almost fell, and was only barely rescued by a brutal crackdown followed by concessions from Nicholas II to form a parliament-like Duma. This was a fucked institution from the very start. Nicholas has still retained more or less all the powers he had before and turned the Duma into a debating club. Kotkin thinks that history would have been much better if the government was just allowed to fall then instead.

Despite remaining an autarchy, Russia did have another great reformer in Peter Stolypin who pursued land reform and further industrialization. In terms of economic output, the effect seemed to have been pretty good. In the 20 years preceding the Bolshevik Revolution, median incomes in Russia had risen 50%. The problem was that this new government had the support of nobody. The liberals and constitutionalists in the Duma felt that it was too little, too late, and also politically illegitimate. Meanwhile, the aristocracy felt that all these reforms were against their own interests. The regime's last chance at survival floundered because there was no constituency which felt bought into the project.

Before we move on to the Bolshevik revolution, some questions about Kotkin’s take on modernization:

• Does he think that countries adopt leading-edge technologies faster than you might expect, given domestic opposition to displacement and automation? Because of this overwhelming geo-political pressure to modernize? If so, does he think that we are overrating the regulatory or political barriers to the adoption of AI? America, will we feel compelled to race on AI against China the same way that late 19th century Russia felt compelled to race against Germany on industrialization?

• Does he think this vicious geopolitical competition still incentivizes the adoption of modernity, given that wars of colonization and territorial conquest have subsided? Sure, this is obviously not true in some parts of the world, but there's no amount of fucking up that some European country could do which would actually get its territory physically dismembered by another country in Europe.

The defeat of Tsarism came not when Kolchak was routed, not when the February Revolution was raging, but much earlier! It was overthrown without hope of restoration once Russian literature adopted the convention that anyone who depicted a gendarme or policeman with any hint of sympathy was a lickspittle and a reactionary thug

Sozenitsn, Gulag Archipeligo

I remain quite confused about how repressive the tsarist regime actually was. The whole revolution was apparently motivated by how the autocracy and their secret police, Okhrana, behaved. But then you have all these people who are literally calling for the overthrow of the government just hanging around. People like Stalin and Lenin who are literally calling for the overthrow of the government are just spending time in exile, living on government stipends, robbing banks, writing articles for Pravda, and having affairs with farmers’ wives. Here is a passage from the Gulag Archipelago:

Let us examine, for instance, some generally known biographical facts about Lenin. In spring, 1887, his brother was executed for an attempt on the life of Alexander III. And what happened to him? In the autumn of that very year Vladimir Ulyanov was admitted to the Imperial University at Kazan, and what is more, to the Law Faculty! Surprising, isn’t it?...

Then a few years later this same young revolutionary was arrested for founding in the capital a “League of Struggle for the Liberation of the Working Class”—no less! He had repeatedly made “seditious” speeches to workers, had written political leaflets. Was he tortured, starved? No, they created for him conditions conducive to intellectual work…

But then, of course, he was condemned by a three-man tribunal and shot? No, he wasn’t even jailed, only banished. To Yakutya, then, for life? No, to a land of plenty, Minusinsk, and for three years…

He asked for an allowance from the state, and they paid him more than he needed. It would have been impossible to create better conditions than Lenin enjoyed in his one and only period of banishment…

Tsardom was always weak and irresolute in pursuit of its enemies. The most important special feature of persecution (if you can call it that) in Tsarist times was perhaps just this: that the revolutionary’s relatives never suffered in the least.

—

Kotkin on what Trotsky could have done in the position he found himself in in 1924, despite being censured by the Central Committee, and missing Lenin’s funeral, and generally finding himself getting sidelined:

In the name of the greater cause of safeguarding the revolution, he [Trotsky] could have violated party discipline by reading aloud on Red Square from Lenin’s purported dictation, using as his mantra Lenin’s summons to “remove Stalin” as general secretary, then flown from factory to factory to rally workers, just as in 1917—let them arrest him. Of course, to do all that, Trotsky needed to perceive Lenin’s death as a strategic opportunity, and he needed a persuasive story line about how the grand socialist dream could be revived, why all those harsh exchanges he had had with Lenin were incidental, and why he (Trotsky) was uniquely qualified to carry forward the sacred Leninist cause. A tall order, to put it mildly. But who could doubt that if Lenin had found that others were conspiring against him, he would have mounted a coup against his own party?

This is such a good question, and one I will ask Kotkin about.

Keep in mind that Trotsky is no naive pushover. He was the primary organizer and tactical leader of the armed insurrection that overthrew Russia's Provisional Government in October 1917. Trotsky directed the Red Guards to seize post offices, telegraph stations, railway stations, bridges, and state banks, effectively paralyzing the Provisional Government's ability to govern. And then of course came the successful assault on the Winter Palace.

Same goes for all the old Bolsheviks, who we know were sooner or later purged, exiled, and/or killed by Stalin. In May 1924, why did Khamanev only read Lenin’s last testament, which called for the removal of Stalin from the position of General Secretary, in front of a small closed session of the Central Committee instead of the full 1,300 Congress? Once Zinoviev broke with Stalin on socialism in one country in 1925, why didn’t he go back to his power base in Leningrad and start an insurrection? Bukharin was Lenin’s golden boy, and the fucking editor of Pravda for god’s sake. All of Russia could have woken up to Lenin’s testament printed on the front pages. What is going to do, kill you? By the time the show trial is being scheduled, that’s guaranteed anyways. How did these hardened revolutionaries who had overthrown a regime which had lasted for hundreds of years get cucked by Stalin?

I just remembered how OpenAI’s board tried to kick Sam out: all they said was, “He has not been consistently candid.” What is the lesson here? No half measures. If you come for the king, you best not miss.

There’s an important lesson in here about the instability of collective leadership. It supposedly worked in China between 1976 with the death of Mao and 2012 with the appointment of Xi as general secretary? But is it fair to call the period that Deng was paramount leader collective leadership? When the premier Li Peng wanted to reverse economic liberation after the Tiananmen Square protests, Deng came out of retirement and went on his Southern tour and declared that, "Whoever is not for reform must step down." Was that really collective leadership? Just because we like what he did doesn’t make it collective leadership.

Stalin’s strategy (which according to my previous guest Victor Shih is also Xi’s) was the align the main members of the party against the next person he wanted to purge. Rinse and repeat. Stalin aligns Khamanev and Zinoviev against Trotsky, then Bukharin against Khamanev and Zinoviev, and by the time it’s Bukharin’s turn, Stalin is basically god dictator already..

An obvious question at this point: at some point you should catch on right? If you’re Bukharin in 1930, and you know you have this independent credibility and power base, what, you think that’s just gonna be fine by Stalin for the next however many decades until you peacefully retire?

Anyways, back to our original question, why doesn’t collective leadership seem to work in practice? Why does it always devolve into one man rule? Will ask Kotkin.

Was Bukharin the Soviet Deng? Killed of course during the great terror, unlike Deng, who was merely rusticated and thus could eventually return.

“the well-off upper stratum of the peasantry and the middle peasant who strives to become well-off are now afraid to accumulate. The situation is created such that a peasant is afraid to mount a metal roof over his house so as not to be called a kulak; if he purchases machinery he does so in a way that the Communists do not see. Higher technology becomes conspiratorial.” Poor peasants, meanwhile, complained that Soviet power hindered their hiring by the better-off peasants. (Most peasants who hired labor themselves worked; they were not rentier landlords.) Party attitudes were holding down production on which the state’s well-being and industrialization hopes rested. Bukharin dismissed the fantasy of collective farms, because the peasants were just not joining them. “That we should in all ways propagandize among the peasants formation of collective farms is true, but it is not true when people maintain that there is a highway to the movement of the peasant mass toward the path of socialism,” he stated. Rather, the answer was to benefit from economic incentives. “It is necessary to say to the entire peasantry, to all its strata: ‘Enrich yourselves, accumulate, develop your farms,’” he told the party activists. “Only idiots can say that we should always have the poor; now we need to conduct policy in such a way that the poor would vanish.”

—

NEP’s dilemma was not merely that the rate of industrial growth seemed too low, making people wonder how long under the NEP it would take before the USSR became a truly industrial country. The dilemma was not merely the unmodernized technical level and small, divided plots of Soviet agriculture, which produced harvests insufficient to support the kind of grain exports necessary to finance imports of machines, including for agriculture. The dilemma was not even just the fact that the regime lacked control over the food supply or the countryside, rendering it hostage to the actions and decisions of the peasantry. All these were profound problems, but the core dilemma of the NEP was ideological: seven years into the NEP, socialism (non-capitalism) was not in sight. NEP amounted to grudgingly tolerated capitalism in a country that had had an avowedly anticapitalist or socialist revolution.

There is this morbid contrarian curiosity about whether collectivization was necessary for Russia’s industrialization. In many cases, I don't even think it's motivated by actual Marxism. There's something appealing about this revisionist mentality of "Are taboo bad things in history good actually?"

But the reality is that there's every reason to expect Russia to have industrialized more successfully without this brutal enslavement of a hundred million peasants. There's not only the example of how almost every other country industrialized, from America to Europe, but also the basic 101 story of how it would have happened is totally credible. More successful peasants would use their profits to mechanize or drive other efficiency gains. Their growing wealth would stimulate a growing consumer economy. Perhaps most importantly, a government not dedicated to the overthrow of wealthy democracies abroad would have been much more capable of attracting foreign direct investment from abroad to support industrialization.

Only in 1928 did Soviet Russia's industrial output recover to the level it was at in 1913 - the point at which it was already a generation or two behind Germany, Britain, and America. And just when things were stabilizing, Stalin launches the first five-year plan, which reduced grain production by over 32% and led to the halving of overall livestock.

The traditional story of what should happen with developing countries is that both the pareto frontier expands and they move from agriculture to industry, as a smaller number of people can handle the food demands of the country. Communist collectivization definitely had the move to industry along the Pareto frontier. But this was coupled to a catastrophic implosion of the Pareto frontier. While Russia was able to produce less grain overall after collectivization, the Soviet government had control over it. It could now export this grain to fuel its military-industrial goals.

Thus, the correct way to understand collectivization is not as a rapid, though costly, speed run to industrialization. Rather, it is as a policy which reduces economic output and even harms industrialization but lets the state control the full economy.

I'm curious if Kotkin thinks that a lesson from this period of history is that one should fall over themselves to support the lesser of two evils.

The Russian examples seem to support this. Arguably, liberals should have backed the authoritarian (but reformist) Stolypin, and the Provisional Government factions should have united more decisively to prevent the greater evil of the Bolsheviks.

Yet this lesson seems to directly contradict the German example, where conservatives did support what they perceived as the lesser evil, Hitler, to stop the Communists, leading to a catastrophic outcome.

Stalin has an inhuman appetite for administration and micromanagement over an empire that streches 11 time zones, from economy to foreign policy to even arrest lists and film edits. He went through dozens of often lengthy memos and documents, and this doesn’t even include his 70 hours of meetings a week. I found it funny how often he marks up memos and letters with the following two phrases:

Regarding something he thought must be changed: “Is it possible that we can allow X? Of course it is not.”

Regarding a supposed blocker to his will: “Is it possible to do X? Of course it is.”

• I remain quite confused about how repressive the tsarist regime actually was. The whole revolution was apparently motivated by how the autocracy and their secret police, Okhrana, behaved. But then you have all these people who are literally calling for the overthrow of the government just hanging around. Lenin’s brother had tried to kill the tsar Alexander III for god’s sake! People like Stalin and Lenin are just spending time in exile, living on government stipends, robbing banks, writing articles for Pravda, and having affairs with farmers’ wives.

• Given how much the war with Japan had weakened the regime, why did Nicholas think that fighting World War I would be good for strengthening his control and image?

• I feel like the revolution of 1905 is understudied. How close did the regime get to falling? Why didn’t Wite’s reforms for the decade prior do anything to placate people? How useful were those reforms in the first place? And Kotkin says it would have been better if the regime were allowed to fall then. What likely would have replaced it? Why not think there might have been another leftist revolution afterwards?

• Why is the tsarist regime sending people to Siberia instead of just killing them?

• Would the hoped for reforms from Wite and Stoyplin have actually caused modernization, even if they had actually been implemented?

• You say that by 1917 some kind of leftist revolution was inevitable, but it didn’t have to end with the Bolsheviks. Why was Russia radicalized in the left-wing direction by World War I, whereas Germany was radicalized in the right-wing direction? Not to mention that Germany had a much more vibrant social democratic tradition than Russia. This is related to the whole question of why communism happened first in peasant societies, which was the opposite of Marx's prediction.

• What distinguished the people like Churchill who immediately saw the folly of both socialism and fascism? What were they able to see that others missed?

• It’s interesting to me that other leaders of the revolution like Trotsky, Lenin, etc have written extensive manifestos. And Stalin is considered the ‘outstanding mediocrity’ because despite having many articles to his name, he isn’t a prominent intellectual. It seems that these other leaders thought that intellectualizing was all you needed, and day-to-day administrative ability was a peripheral concern.

• Do you think a lesson from this period of history is that one should fall over themselves to support the lesser of two evils?

  • The Russian examples seem to support this. Arguably, liberals should have backed the authoritarian (but reformist) Stolypin, and the Provisional Government factions should have united more decisively to prevent the greater evil of the Bolsheviks.

  • Yet this lesson seems to directly contradict the German example, where conservatives under Franz von Papen did support what they perceived as the lesser evil, Hitler, to stop the Communists, leading to a catastrophic outcome.

• Why did Marxist revolutions always happen in places Marx thought worst for them? Marx saw industrialization as a prerequisite for revolution; most actual Marxist revolutionaries came from agricultural soceities and ended up doing really terrible things to try and jump start industrialization. What did Marx and Engels not understand here?

• So Stalin ends up being this inhuman administrator and micromanager. But he’s happy spending over a decade plus before the Bolshevik revolution just hanging out in Siberia, impregnating farmer girls, and other small time shenanigans. Wouldn’t you think that someone who has that much aptitude for organization and leadership would get up to something more ambitious, even if revolutionary?

• There’s an interesting book called Stalin’s Library, which goes through Stalin’s wide and diverse readings of literature, science, history, and economics. And the book notes that in none of Stalin’s extensive markups does he even hint at doubting Marxism. How can this be? How can someone who has been exposed to so many different ideas not even come across something which makes him doubt his principles? Does this make us question the whole value of humanistic education?

  • What's the best way to think about what Marxism and Stalinism were? How did so many people who consider themselves part of the intelligentsia develop this almost religious attitude towards criticism of the party? Why were they so convinced that history was inevitably headed in this direction? Is Marxism a replacement religion, a secular apocalyptic cult, the psychological equivalent to a spiritual experience?

• How do we explain the surplus of sadism in Russia, which allowed Stalin to enforce collectivization and the great terror? The 25 thousanders, the tens of thousands of interrogators and torturers in the gulag system.

• What if the Politburo members had simply accepted Stalin's resignation? Would there be any face-saving way for him to still come back? Or would that have been it? Moses tried to pull a similar trick on every single mayor in New York City until Governor Rockefeller actually just accepted it, and nothing bad happened. His bluff was called.

• Did any old Bolsheviks ever express regret about the whole revolution once fruits of the revolution turned sour? Not just about letting Stalin get so much power, but the whole Marxist uprising thing in general?

• Suppose someone else had won the succession struggle after Lenin. What would Stalin have done if he had found himself as a Politburo member serving someone else’s dictatorship within a dictatorship in 1930? Would he have been able to instigate the kind of conspiracy that nobody dared assemble against Stalin?

• Say what you will about the Bolshevik revolutionaries, they certainly had balls. Why did so many of them fold to pressure? Why did Old Bolsheviks like Nikolai Bukharin and Grigory Zinoviev, who had faced down the Tsar, confess to fabricated charges of treason during Stalin's public Moscow Trials? Knowing execution was certain, why didn't they use that final platform to expose Stalin's tyranny? Similarly, during the Cultural Revolution, why didn't a figure like Premier Liu Shaoqi, once Mao's designated successor, use a final speech at a party plenum to launch a desperate, last-stand denunciation of Mao as he was being systematically purged and humiliated? In all these cases, there was an extensive period where they were basically a dead man walking, but still in their normal positions of power.

• Stalin is personally choreographing many of these show trials. So at some level he must have known that they were mistaken. Why does he do it nonetheless?

• Why doesn’t the (more than) decimitation of the Red Army instigate a military coup?

• Relatedly, in Tsarist Russia, ministers - and even tsars - are getting assassinated left and right. In your coda for the first volume, you talk about how easy it would have been for someone to assassinate Stalin. This is someone who a 100 million peasants he has enslaved have something against, and who dozens of top leaders have good reason to fear. Why does no one even try to kill him?

• So you say that Stalin was hoping for a war between the capitalist powers because he thought it would expand the reach of communism, just as the First World War had. Given how outrageously successful he was in using the Second World War to increase the global footprint of Communism, did Stalin wish for a Third World War?

• Stalin seems oddly good at foreign policy. I think in a lot of alternate worlds, either Chiang Kai-Shek completely wipes out the Chinese Communist Party instead of creating a united front with them (which gave the communists a lot of unearned prestige, given that they were supplying literally 1/40 of the troops as the nationalists), or alternatively, he is just killed and kidnapped by the Communists, which would likely force a Japan collaborationist government in China.

  • One big update from your biographies was that Stalin was very reluctant to promote socialist or communist parties abroad, including in Spain and China.

  • He seems much more successful and less ideological with foreign policy than domestic policy.

• If Bolsheviks don’t take power in Russia, does Communism just not become a historical force. Maybe Europe still goes social democratic, but you don’t have brutal leftist dictators take control in China, North Korea, parts of Africa and South America? Or do you think this kind of Marxist-Leninist political philosophy is enough of an attractor state that it would have arisen independently?

• If Stalin is supposed to be this dedicated Marxist, why is he running a more pragmatic, less ideological foreign policy? Half hearted support of Communists during Spanish Civil War, trying to get Mao to stop at the Yangtze River when chasing down Chaing Kai Shek

• Why doesn’t communism have a bigger impact on the longer run growth trajectory of Russia? Do you suspect that without communism, Russia might have been hyperbolic instead? Do you think without communism, due its large population (1.29x US in 1939) and resources, Russia might have ended up with a bigger economy than even America (which it never had during the real Cold War).

• • Is it possible that while less than optimal, communism actually was somewhat workable in the early 20th century because technology just required massive investment, didn’t require that much signal from consumer demand?

• Was Stalin right that communism wouldn’t survive in the long run if most of the economy (which was agriculture at that point) was run on a capitalist basis? Of course, it would have been a good thing if communism collapsed. But nonetheless, given his political philosophy, was he empirically correct about the right next action?

• As of 1924, did the communists really believe that you could just get socialism without coercion? How exactly was that supposed to happen? To the extent it was supposed to be from state companies just freely out-competing small peasants, as Bukharin hoped, isn't that just capitalism?

• You have an interesting take that “the package of attributes that we call modernity was a result not of some inherent sociological process, a move out of tradition, but of a vicious geopolitical competition in which a state had to match the other great powers”.

  • Do you think that countries adopt tech faster than we might expect because of this overwhelming geo-political pressure to modernize? If so, do you think that we are overrating the regulatory or political backlash to the adoption of AI? Will we feel compelled to race on AI against China the same way that late 19th century Russia felt compelled to race against Germany on industrialization?

• Do you think this vicious geopolitical competition still leads to the adoption of more advanced government/tech, given that wars of colonization and territorial conquest have subsided? There's no amount of fucking up that some European country could do today which would actually get its territory physically dismembered by another country in Europe.

• You note how the early 20th century had seen more technological change than any time before (and probably even since) - planes, tanks, cars, airplanes, radio, extension of railways and steamships, further mechanization, etc etc. Is there any way this pace of change is implicated in the chaos and extremism in Russia (and Europe overall)?

• At a high level, how do you think about the differences between the regime change in 1917 and 1991? Why doesn’t the latter produce any kind of grand ideology?

• Does China today share to any significant degree the weaknesses of the Eastern block before 1989? There is falling productivity in the last-ditch effort to make up for it by investing heavily in some technological miracles. But certainly they have the opposite of Polish disease, in that people are concerned they’re exporting too much, right?

• China has preserved the Leninist core of the Soviet system (with party control over state, cadres running everything, state subsidization of militarily relevant heavy industries and technologies) but not the Marxism. Does it disprove Stalin's claim that political monopoly cannot survive capitalism?

• Why does collective leadership always seem to devolve into one person dictatorship in practice? Why is it an unstable political equilibrium? Soviet Union under Stalin, China now under Xi, etc.

• Say you are Xi Jinping. What lessons from Stalin would you be heeding? What lessons should you be heeding but maybe are not?

• Xi Jinping makes a big deal about how the Soviet Union fell in part because it repudiated Stalin (ergo, China must not repudiate Mao). What do you think of this judgement?

• Every morning you wake up, you get to choose whether you will either go through the next many dozen papers in the Stalin archive and write the next few pages of the meetings he conducted on some given day, or whether you will go on some podcast and attract a million people through your personality and reservoir of scholarship. How do you make that choice?

Special thanks to Ege Erdil and Tanner Greer (see his here), but also many others, for excellent suggestions for questions.

I sometimes joke that on a viewer minute adjusted basis, I host the Sarah Paine podcast, where I sometimes also talk about AI.

I am delighted to announce that Sarah is coming back for a 6 lecture series in July. We will record the lectures in person in San Francisco. As a token of gratitude to my paid subscribers, below are the links to get free tickets one day in advance of anybody else.

Sarah has become super famous now (and rightfully so). But just in case you’re hearing about her for the first time, she is Professor of History and Strategy at the Naval War College. She is not only by far my most popular guest to date, but with the 3 additional lectures we recorded last year, she now constitues 4 out of my 5 most popular episodes.

Sarah is a true authority on the history of conflict and strategy, and has spent decades unearthing insights in archives all over the world, from China to the Soviet Union to Taiwan to Japan. Her lectures distill the lessons from her many decades of ground-breaking scholarship. I’m excited to experience her insights on 6 fresh new topics live with all of you in person.

If you can’t join us in San Francisco, please do not worry - all these lectures will be made available online within a few months.