State Space

Zhane Yamin: I am fascinated by the phase shift that happens when a scientific field stops being something that we study and starts becoming something that we build. In the late 19th century, parts of chemistry made that leap, and we got plastics, rubber, and synthetic fibers, and parts of physics made it in the 20th century, and we got transistors, computers, and wireless communication. My guest today is working at the forefront of the next leap. Thanks to AI, gene editing, and automation, and more tools, we’re learning to engineer living systems, to engineer biology the way we once learned to engineer materials and machines.

That’s a huge opportunity, and it’s also a major front of strategic competition. The National Security Commission on Emerging Biotechnology was charged with finding and recommending how America could keep that lead, and it has been remarkably effective. Twenty-five of its recommendations, in full or in part, have made it into law, and more are seeing major action from the federal government. So in this episode, I wanted to answer two questions.

What is this biotech moment? And how does it relate to science acceleration? And how is the commission meeting it so effectively?

There’s no better person, in my opinion, to ask than Caitlin Frazer, the executive director of the National Security Commission on Emerging Biotechnology. She leads the staff that supports the commissioners, and she spent much of her career on the inside of the legislative machine. Caitlin, welcome to the show.

Caitlin Frazer: Thank you so much for having me.

Zhane Yamin: Of course. So just to start off with, I was wondering if we could just go in your definition of biotechnology. Because I think for some people it’s vague, for some people they might have different definitions. So when your commission says emerging biotechnology, what are you guys actually talking about?

Caitlin Frazer: What do we mean? Fair question. It’s actually something that we spent a number of weeks when we first started this project making sure that we made the right scoping and definitional decisions because this project was huge at the outset, and we knew that to get it right, we needed to be very clear with our audience and with ourselves about what we meant. So biotechnology, of course, is a millennia’s old technology.

Humans have been fermenting beer and wine and cheese and bread for millennia. We’ve been using biotech for food and for medicine for a long time. But emerging biotechnology or modern biotechnology is really about the newer ability to engineer living systems and start to treat biology as a field of design and engineering. What has unlocked that over the last two decades or so is that researchers have figured out how to read, write, and edit DNA.

And in doing so, we can now program, the living code in the way that we might program a computer. When you apply AI and automation, on top of that, you get an accelerant effect. And so we’re just seeing this enormous potential of biotechnology, which humans have worked with and lived alongside forever, explode into — now we think about biotech as a general purpose technology because it has applications across economic sectors. And we think that that is even poised to grow even more, if we get the policy foundation right.

Zhane Yamin: So one thing I wanted to dig into a little bit too is why specifically a national security commission is a good fit for this, right? ‘Cause you know, you think about maybe health or science commissions, but why national security? What does that framing sort of mean for you guys?

Caitlin Frazer: Yeah, absolutely. So the Congress created our commission in an act of law about four years, four and a half years ago now. And we’re actually the third in a series of national security technology commissions. And so there was a history to Congress reaching for bipartisan outside experts to help them think through the convergence of national security concerns with emerging technology issues.

So the first of our tech commissions was on cybersecurity, then there was one on AI, and now emerging biotech. And I could imagine that there’ll be appetite for more coming. We’ve seen proposals for national security commissions on robotics and fusion and all kinds of other quantum and other technology disciplines. But I think what Congress was really asking for in creating this commission was for our group of outside experts, our commissioners, and our staff to think about the way that US leadership in biotechnology — which has historically been quite strong in many ways, modern or emerging biotechnology is an American invention — to think about that with respect to our national security, our supply chain resilience, our economic prosperity.

And so one of the other scoping decisions that we made very early was to define national security. Right, so we didn’t have to just define emerging biotechnology, but we had to think about what national security meant to us. Because you could define it fairly narrowly and say, Okay, what are the conventional military applications of biotechnology?

Like, how should we be thinking about these tools to advance military medicine or trauma care or combat trauma care? How should we be thinking about the ability to biomanufacture reshaping military logistical supply chains. And we did that.

There’s a whole chapter of our action plan that is explicitly about biotechnology for conventional defense capabilities. But our commissioners wanted to think much more broadly and to define national security to include, like I said, supply chain resilience, food security and agriculture security, energy resilience and energy supply, biotech for critical minerals, and of course, the advancement and reliability of our medical ecosystem as well. So we bit off quite a bit when we defined national security that way, but I think it was an important decision because what it enabled us to do was to share with the Congress all of these myriad applications of biotech that are real and available today, but also where they’re going in the direction of travel over the next five, ten, fifteen years.

Zhane Yamin: So I’m interested to hear what your perspective is on that national security framing for sort of like the normal person. Your report also cites some amount of companies relying on other international competitors for parts of their manufacturing processes. How does this broader national security framing relate to how people themselves should feel about or might feel it in their lives?

Caitlin Frazer: Yeah, absolutely. Well, candidly, when we started, one of the first things that my team and I did was to just immerse ourselves into the existing emerging biotechnology, synthetic biology community. We wanted to listen before we started talking. And we spent about a year doing that.

So a lot of conferences, a lot of stakeholder calls. And very honestly, the reaction to us showing up and saying, ‘Hey, I’m Caitlin, I’m with the National Security Commission on Emerging Biotech, we’re a government entity, we’re a legislative branch advisory group,’ was mixed. And so about half of the people that we would engage with would say, ‘Cool, I want to learn more. I’m interested in things that the federal government can do to accelerate biotechnology R&D, or maybe streamline product regulations, or maybe, you know, fund some of the work that we’re doing.’ But about half the people that we worked with in that first year were resistant to that framing.

And in many cases, the life sciences industry and the biotechnology industry has not thought of itself historically as part of the national security industrial base of the United States. And so we would meet with people and they would say, ‘Well, I’m never gonna work with the Department of Defense or Department of War. I’m never gonna work with the intelligence community. I don’t know why I should be a part of this project.’ And so it was kind of a two-way conversation, us learning about the shape and scope of the biotechnology ecosystem here in the United States, but also beginning to invite people to the national security conversation in a way that they really had not been before. And fast forward three and a half years, and I think you now see quite a bit more understanding of both perspectives from each side, which has been, I don’t think we’re singularly responsible for that, but I do think we helped bring those communities together.

But your question was really more about like how this framing applies to the American people. And I think that it’s an interesting question, but it’s almost like even a step too far, which is to say that the vast majority of Americans probably don’t have a deep expertise or appreciation for either biotechnology or national security considerations. And so our job as advisors to Congress and to policymakers is to help with bioliteracy.

We call it bioliteracy because we are kind of building off the analogy of computer literacy or IT literacy, which is that don’t need to know how to code or be a deep technological expert to use a computer well and safely. Most of us do. I’ve got like, four of them sitting in front of me here.

And so we want something similar for biotech and for national security, which is for people to be conversational in these topics and be able to be informed and safe consumers of biotechnology products, without necessarily feeling like they need to have an advanced technical degree to be part of the story. So yeah, I’ll leave it there. I think that is the message that we’ve tried to share with policymakers to then turn and share to their constituents about why it matters to think about biotechnology as a national security imperative.

Zhane Yamin: So I’m interested just in, do you see this idea — this analogy between computer literacy and bioliteracy — Do you see a future in which people are even somewhat as bioliterate as they are computer literacy? Or do you see this as more a policymaker thing where people who need to interact with this field are bioliterate in the sense where they need to? Or do you see it being something broader? Society as a broader entity could be more bioliterate fifteen or twenty years or whenever this stuff gets time to diffuse.

Caitlin Frazer: Yeah, absolutely. We hope for both. It’s probably more focused on policymaker bioliteracy in the near term, because the decisions that our policymakers are making right now and over the next, 18 to 24 months are, in our view, what’s going to decide the trajectory of American biotechnology leadership for the future. So that’s the most urgent bioliteracy priority.

But we would like to invite everyone to that conversation. But if we kind of tease out the computer literacy and analogy a little bit, what vote fundamentally changed that? It was computers as consumer products, right, that everybody could buy and use.

I’m old enough to remember when I moved from a flip phone to a smartphone for the first time and it was like, I have to learn this whole, you know, new technology now. That makes me sound ancient. That’s true.

And so it’s biotech — with the exception of medicines and some agricultural biotech in farming, growing communities — is not particularly well appreciated as a consumer product yet. Because it’s often an input into a finished product that people don’t see or use in the same way that they use a computer or a smartphone. And that’s kind of the gap that we see today is that most consumers don’t yet appreciate how biotech is already showing up in their products and in our supply chains, and definitely don’t yet appreciate how it could be adopted as a means of making things in a new way, which is kind of the premise of your intro.

Zhane Yamin: So to get into the actual findings and recommendations of the commission, right? So you guys have this big report. I was wondering if you could just walk through some of the major headline findings and recommendations, and then we can tease out some different threads from there.

Caitlin Frazer: Yeah, for sure. So we made our action plan to Congress public in April of twenty five, and it really amounted to a two-track strategy to retain and regain American leadership in emerging biotech. On the one hand, how do we run faster? We have so much innovation and research and development here in the United States, but historically we have struggled, whether it’s for regulatory issues or lack of manufacturing infrastructure, any number of other reasons, we have struggled to translate those lab scale innovations into full-scale commercial production biotechnology products.

So a scale-up challenge. How do we run faster here at home and scale our innovations? But also, second part of the strategy is how do we push back against what we call the brute force economic tactics that the Chinese government has used to create an unfair non-market advantage for their companies and their ecosystem. And taken together that two-track strategy was 49 total recommendations to the Congress, which as a former congressional staffer, I wince every time we go into a congressional office and say, You gotta do 49 things.

But we do have to do 49 things because really the United States government has lacked for a whole of government national strategy to secure and endure American leadership in this space. So all of that historic leadership that the United States has enjoyed has been, for the most part, because of work done outside of the federal government. And we observed pockets of extraordinary work within the federal government, but there has not been a strategy to knit all of that together, nor has there been a senior leader at the White House empowered to oversee and drive the execution of that strategy over time. So that’s one of our major observations, a sort of national strategy and national leadership for a whole of government effort here.

But again, we made 49 recommendations. They’re sort of organized into six pillars, one of which is that strategy thing. The others have to do with that scale up challenge. So, how do we mobilize private capital to support scaling innovations from lab to market?

How do we streamline the regulation of biotechnology products so that it doesn’t take 10 years to move a product through federal regulatory pathways? So scale up. Third is around how we can mobilize biotechnology for conventional defense applications, as we were talking about earlier. Fourth is around out-innovating our strategic competitors, so kind of looking further upstream into the R&D pipeline and saying: What are the grand research challenges that will unlock, leap-ahead capabilities in biotechnology?

And how can we make sure that we set up the government systems to safely, securely, and responsibly innovate in biotechnology for the future. Fifth is around the people that we will need to power all of that. So talent pathways, not just into government, but also into careers in the private sector. And sixth is how do we mobilize our partners and allies to take advantage of our mutual capabilities.

Zhane Yamin: One thing we talk about a lot on the podcast is the things that are bottlenecking science that aren’t necessarily the science itself or the fact that the problems are hard, which they often are, but these things that we trip on, like tripping over our own shoelaces. And so I’m wondering from the commission’s perspective, what are those bottlenecks or those shoelaces that aren’t necessarily the scale of the technical problems themselves, but the things that we kind of do to ourselves in a sense of how we’ve set up these systems? I’m interested to hear if those kind of resonate with other sort of things that people have brought up as bottlenecks to the scientific process.

Caitlin Frazer: Yeah, there are a lot of them, regrettably, but I’ll just focus on one, which is around product regulation. So of course, there are regulations around how research is done safely and securely, but here we’re talking specifically about the risk assessment for the end product before it you know goes to consumers. And as you might imagine in biotech there’s a range of products. There’s medicines, there’s agricultural products, there’s now engineered microbes for release into the environment, biotechnology animals, biotech plants.

And so across all these different product categories, you involve three regulatory agencies in some cases. So medical products actually, for all the discussion in Washington these days about how we streamline and reform the FDA to move medicines through the regulatory process safely but faster, they actually have the easiest time of it because they only work with one agency, just the FDA. Other types of biotechnology products can refer to two or three federal agencies for regulatory risk assessment: FDA, USDA, and EPA. And in some cases, we found that there were no clear regulatory pathways for novel products of biotechnology, meaning nobody had yet figured out how to navigate going through those multiple agencies.

And the agencies themselves were not doing a particularly good job of coordinating behind the scenes. There’s a 40-year-old document called the Coordinated Framework that is supposed to govern the relationship between those three agencies, and it works okay. And we were able to determine that in a lot of cases these regulators were spending time on products where the science was familiar and not able to carve new regulatory pathways for novel science and novel products. And so we wrote extensively about regulatory reform, not just in our April twenty-five action plan, but also subsequently late last year, we put out a series of papers, with deep dives on each of those product categories that I described, because we think that there are sensible things that the agencies can do without new congressional action, and a handful of things that Congress will need to do to amend existing statutes or provide oversight, but taken together will streamline the product regulatory system in the United States and really focus on creating durable pathways for moving novel products of biotechnology into the market.

Zhane Yamin: Got it. And the second thing I wanted to ask about that we also talk a lot about at the coalition and on the podcast is this idea of AI ready data. And I know that that is something that you guys talk about a lot in the report and want to bring to the forefront.

So I wanted to ask, what is the problem that the commission sees with biological data today? And how is it becoming more of a moment with emerging biotechnology? What do we need to do with biological data for us to meet the moment?

Caitlin Frazer: Yeah. Again, we have to do a lot of things, but some of them will be easy, I think, or easier. So our thesis is basically that we need to start treating biological data as a strategic national resource. And that means getting our data house in order, within the US government.

That idea we call the web of biological data, which is essentially to again knit together into a web. It’s got a, you know, pretty obvious name. It would knit together all of the existing sources of biological data across the interagency and make it easier for researchers to access it in a safe and secure way so that we know who’s accessing the data and we’re able to secure data that is sensitive or maybe personally identifiable or even national security sensitive. But in doing so, make it easier for researchers to not have to go, onesie twosie to multiple different agencies to access the data they need to power their research.

We also need to create new data and make sure that that data is AI ready. So we have two recommendations around the expansion of programmable cloud lab infrastructure. Cloud labs, of course, are physical laboratories but driven by AI and highly automated, so that they can perform very precise experiments and derive highly replicable data out of it. Really exciting announcements on Cloud Labs out of the administration just over the last couple of weeks.

So that’s moving forward nicely. And the other recommendation in the data generation space is America’s Living Library Act, which is essentially a new initiative led by the Department of Interior to sample and sequence the flora and fauna of America’s public lands. So much of, especially our drug development is derived at some level from an understanding of the natural world. I can give you a million examples of that.

The one I like the best is that the PCR tests that all of us took, you know, the rapid PCR tests that we would take for COVID, I think I would take three a week to be able to come to work, that’s actually derived from a sample taken by researchers at Yellowstone National Park. There’s myriad examples like that. So data generation is programmable cloud labs and America’s Living Library. And then how do we make sure that all that good data, whether it’s existing data that we’ve better organized or newly generated data, is ready for use in AI models?

And that comes down to standards. So we have also a bill called the AI Ready Data Standards Act. We want to name things plainly. But that would empower NIST to work with industry and academia to create national standards for what constitutes AI ready biological data. So we’re all working from the same playbook and any time we generate new data, it is embedded in that is that it is AI ready.

Zhane Yamin: So one thing I think is really interesting as we’re talking about this is that a lot of the stuff that you’re saying, these bills that are being introduced, there’s a lot of motion around it. And even as somebody that’s not super following the biotech policy or the biotech sphere, I think people might have come across this, the Web of Biological Data Act that was also introduced. All of that stuff is very prescient. And so I kinda wanna transition as we’re coming to the last few minutes of the interview here of why do you think this Commission has been so successful in your words and what has made it successful, particularly?

Because I think when a lot of people think about policy, they’re like, yeah, like that’s a good idea. Now there’s 50 hundred other things you have to do to make that idea a thing, right? And so I am just very curious, what do you think has made this commission so successful as opposed to just a report that goes on a shelf somewhere?

Caitlin Frazer: Yeah. I mean it started there. Like we said out loud in our first few days on the job, we are not writing a report that’s gonna sit on someone’s shelf. We’re writing an action plan.

And that action plan is going to have bipartisan recommendations that Congress can seize upon. And then we’re gonna do the work after the action plan alongside members of Congress and their staff and then committee staff, to write the bills and then build the coalition that is going to advance these pieces of legislation. Congress was kind enough to give us more than 18 months post report to do that work, and we’ve leveraged every minute of that. So part of it was that we built a nice coalition up ahead of our launch, and so there were a lot of people that saw themselves in the action plan and wanted to be a part of moving it across the finish line.

We have an extraordinary subset of our team who engages directly with Congress. We now have more than 400 relationships with member offices and committee staff. There’s a woman named Katie Hamilton on my team is just like absolutely indefatigable and she deserves a lot of credit for that, but others on the team have been instrumental as well. And I think it also helps that Congress asked for this.

Right? We are a legislative branch advisory group. We have.gov at the end of our email addresses. We are funded by congressional appropriations, by taxpayer dollars.

And so we’re able to go back into Congress and say, this project was a bipartisan one, funded by Congress for Congress. It does not have donor bias behind it. We are here to work alongside you to make these things a reality.

That has helped a lot. We were really focused on making sure that we were giving our audience what they needed, not just writing the report that we wanted to write. And I think a lot of that credit goes, of course, to our commissioners for setting that vision early, but our team, our staff, my staff have been laser focused on that and really spent a lot of time engaging with Congress to make these things a reality.

Zhane Yamin: How do you think the idea of this commission being by structure, by its nature, temporary, how do you think that has sort of impacted the policy process and how you guys are thinking about how you’re gonna do this?

Caitlin Frazer: Yeah. So you’re right. All of these tech commissions are temporary in nature. They get a few years to do the analysis and research, make your ideas public and then depending on the commission you get a little time after.

We had a lot of time, which we again, like I said, took ample advantage of. But the temporary nature has helped us in a couple of ways. First is we were able to go to the biotechnology industry, especially the startups that work on a really accelerated timeline and say, guys, this is not some government project that is gonna take ten years to mature. Yes, in startup land, four or five years is a long time, but for government it’s a pretty short period of time.

And so that helped, I think, for people to understand that we felt the sense of urgency that they felt as well. It has also meant that when we told members of Congress as part of our action plan that we assessed that they had three years to act. And we meant that. And it wasn’t just because, we had our own timeline, but we assessed that the decisions made in the April 2025 to April 2028 timeframe, were truly going to determine the trajectory of American leadership in this space for the future.

And so then to be able to take that message and say — this is what our analysis tells us, here are the recommendations, you have a year to work with us — has I think also injected a sense of urgency among our policymakers that has yielded a lot of the results that we’re seeing in terms of legislation. And in your intro you talked about, more than half of our recommendations have already been enacted into law in full or in part. The statistic I’m really excited about these days is that more than 80% of our 49 recommendations are moving forward legislatively on a bipartisan basis.

Forty-one of those 49 recommendations have legislative progress. So that plus the fact that the House Biotechnology Caucus now has a membership of over eighty bipartisan members, didn’t even exist a year and a half ago. Those to me are the statistics that demonstrate that it’s not just us feeling a sense of urgency, it’s also our policymakers really meeting the moment.

Zhane Yamin: It’s an optimistic take counterintuitive to what a lot of people think about government, right? I think it’s very nice to think Okay, people are moving and we’re doing a thing and it’s good and it’s working.

Caitlin Frazer: Yeah, I don’t know how else to operate except I’m actually naturally a pessimist. And so like it is helpful to be able to put yourself in that optimistic sort of headspace. But I believe that, you know, I’ve always worked across the aisle in previous jobs, embedded in the commission’s ethos is that bipartisanship. Of course our chair is Senator Todd Young, our vice chair is Dr. Michelle Rozo, who was a senior leader in the Biden administration.

And then you could go through all the commissioners and see that we’re pretty evenly split across party and ideology. And so the result of that was that all the recommendations coming out of our group were tested for their bipartisan nature and also tested for their implementability, because we had four sitting members of Congress as part of our panel, and so they were able to say, Hey, I don’t know if that’s gonna work. This is too jurisdictionally complicated or too expensive or not expensive enough. And so we are able to have those internal debates to really stress test or red team the implementability of these ideas, which I think helps a lot.

But yeah, I mean Congress still gets a lot done, certainly in biotech. And I think we should be optimistic about that, and I think we should lean into the momentum. And certainly as we approach the end of the commission’s tenure, I’m excited about what 2027 and beyond is gonna look like with respect to action on all these ideas.

Zhane Yamin: I think that’s a great place to end. Caitlin, thank you for joining me to the show.

Caitlin Frazer: Thank you so much. This was fun conversation.

Zhane Yamin: People are always going on about data these days, especially with AI. Get the data, create the data, protect the data, manufacture the data, don’t steal the data, you get the idea. People especially talk about data sets when they’re trying to train AI models on them because their predictive ability improves with more data.

So everyone agrees, yes, we need more data, but more data is pretty much where the agreement ends and the trouble starts. My guest today has thought harder than most about why that happens.

Peter Koo is an associate professor at Cold Spring Harbor Laboratory, where he runs a lab at the Simon Center for Quantitative Biology, building machine learning methods to decode the regulatory genome. I would say that’s a pretty relevant background for this conversation, but he also wrote one of the winning pieces on Hesero’s MetaScience Essay Competition, a piece called “Before the Next Protein Data Bank,” where he provides an argument about how a field decided what to build before it spends a fortune building the wrong thing.

Excited to dive into it right from Peter’s lab in Cold Spring Harbor Laboratory. Peter, welcome to the show.

Peter Koo: Thanks for having me.

Zhane Yamin: It’s great to have you. So, softball, I guess, to start off with, depending on how you see it, but maybe also a very big question: why do we need data sets, and why are they helpful for machine learning or AI or these models that we like to build for science?

Peter Koo: Yeah, in the sciences, and in machine learning generally, we’re trying to understand the world. We’re trying to build a predictive understanding of the world. And in some fields, like in theoretical physics, we can write down equations that describe how dynamical processes will evolve over time, and that in itself is a predictive model of the world.

But we can’t build those kinds of models for everything. In other areas we need data to build statistical models, to learn patterns from the data in order to extract the rules of how the world works. So it’s a combination of the data itself, which encodes information about some aspect of the world that we’re trying to understand, and the statistical model that’s going to extract that information out from these models so that we can end up using it to predict future events or phenomena.

Zhane Yamin: So in your work specifically, what does that look like? What is that data? What are you trying to build? What are you trying to predict?

Peter Koo: So we are focused in the life sciences, and specifically we cover a lot of different scales of biology. The majority of my lab, we want to build models that understand how the DNA code functions. Within every cell in our body, they share the same copy of DNA, yet every cell in our body functions so differently.

So we want to understand what parts of the DNA encode and control different kinds of cell types all throughout our body. And just to step back a second, the genome is about three billion bases, meaning As, Cs, Gs, and Ts, and only 1% of those code for proteins. The vast majority, it’s not really known exactly what it does.

Sometimes it acts structurally to compact the genome into the nucleus, but there are large portions of it that control which genes to turn on and off. For instance, my heart doesn’t need a neurotransmitter, so I could just turn that off. Turn that gene off. The gene regulation program controls how much each protein should be expressed, and we want to understand how that code works.

Zhane Yamin: So then when we’re building these models, and maybe your work as an example, why is a model only as good as the data? What role does the quality of the data play into the quality of the model or the predictions?

Peter Koo: That’s a good question. The quality of the data is a very challenging question, because what we want to measure is at a very small molecular scale. If we could just see everything that’s going on, just visualize it, and see all the interactions taking place, and just take a movie of that, then we understand how the process works. But we can’t do that at this molecular scale.

All we have tools to probe different aspects of this process. Maybe we can see which proteins are binding to DNA, but we can’t look at all the proteins binding to DNA. We can only look at the specific protein we’re interested in and see where that protein is bound to DNA all throughout the genome. These all provide little noisy snapshots.

Also, even when we look at a protein binding to DNA, we’re not actually visualizing that. We’re indirectly measuring that, because we have to figure out biochemical processes, steps to interlock the protein with the DNA. Then we have to get rid of the cell walls and extract out the DNA bound to the protein, and we have to somehow sequence where that protein was bound.

So there are a lot of processes happening. Biochemical steps where you cut up the DNA, where you remove the protein from the DNA, and then you have to sequence it. And even then you have lots of processes happening. So what you end up getting at the end, the actual raw data measurements, is going to be a very indirect and very noisy measurement of the underlying biological process that we were truly wanting to capture.

We need models to be able to take in that raw data and try to go back and prune out all that noise and extract out the biological signals from this. Different data types have di fferent noise levels. They have different ways of indirectly monitoring this, and so the quality is of course different across the board.

Zhane Yamin: Is it more so the job of the model to take out the patterns in that data? Or how much responsibility do we have as researchers, or the creators of these data sets, to make sure that those patterns are easily as noticeable?

I guess what I’m saying is, in terms of cleaning the data, what does that mean?

Peter Koo: So cleaning the data means processing it and getting it ready for machine learning models to analyze it. If you give a machine learning model too much noisy data, maybe it could learn the underlying biology on its own, but the noise level is so large that it’s often hard for it to find those patterns. And maybe in this sort of framework you would need lots more data, because the noise level is so high that you need to be able to see so many different samples of noise that the shared underlying biological signal, you can identify that.

But by using human priors, humans figuring out how to clean up the noise of the data, it ends up making the data quality a lot higher, and so you don’t need as much data for a machine learning model than to learn the underlying patterns. But you still need lots of data. It’s just you don’t need as much data, and it’s more data efficient in that sense.

Zhane Yamin: Would it be fair to say then that there’s active trade-off between the amount of data we need and the quality of that data? Like if you need less data... Could you theoretically just brute force it with a bunch of data that maybe is very noisy? Obviously that’s suboptimal, but is that theoretically possible, or is there sort of a roof or a ceiling or something?

Peter Koo: I think when we think of data as high quality versus noisy, there are other axes that are important at play here. The task that we want these models, what patterns should they be learning from the data? If it’s a relatively simple set of rules or patterns that the models need to infer, like a protein binding to DNA might, if the binding is very, very strong for a specific subset of nucleotides, then learning that rule can be more straightforward, versus a protein that is very promiscuous and just binds loosely across the genome.

And so the underlying patterns aren’t as strong, so the complexity of the data can also matter. When the underlying patterns aren’t that complex and the underlying rules aren’t so complex, then you don’t need as much data to learn those rules. As the complexity of the underlying mechanisms grows, then of course you need more data.

Zhane Yamin: So one question I’ve been wondering is, are all data points created equal? And how does that factor into our conversation around convergence and the creation of data sets? I feel like before, at least in your essays or in previous writing, you mentioned this idea of mechanisms and how mechanisms inform predictive distributions. I was wondering if you could just talk more to that in general.

Peter Koo: Definitely, definitely. So maybe I’ll just walk it back a step for a quick second and talk about how we typically build AI models right now in biology. Biologists typically generate data on their own, and then we train a model on that data. We try to build a statistical predictive model based on that data, and it can do certain tasks.

And we never ask whether that was the right data to train the models on. I think as modelers we need to start asking what kind of data do we want our models to be trained on? That’s the right question. Not all data is created equal.

I like to think of data as living in a combinatoric space. For sequences this is all possible sequence space; for images this might be all possible image spaces. It’s a huge, vast space.

For DNA sequences, the space of even just very short DNA sequences of only 200 nucleotides long, that’s so tiny. Our genome is three billion nucleotides long. For 200 nucleotides, the space of all possible sequences is 4 to the L, 4 to the 200.

That’s already more combinations than the number of atoms in the universe. So that’s the scale that we’re talking about. It’s just practically speaking infinite. It is a finite number, but it’s practically infinite, where we can never sample all of the data, even for short DNA sequences.

So the space is vast, and when we sample data points, this is only a very sparse sampling of this vast space. Then we train a model on that sparse amount of samples, maybe even 30,000 data points. A million data points sounds huge, but even then, compared to the number of atoms in the universe, it’s very, very small. And you train a model on this small amount of data that covers a small, sparse sampling, and then we want these models to generalize everywhere.

It seems like an impossible task. And it is impossible if every single data point is independent. But in actuality, all of those data points aren’t independent, because they’re governed by the same underlying biophysical rules.

And so the biophysical rules again, we go back to complexity. If we’re only looking for a protein binding to a DNA sequence, that’s not very complex in terms of the rules. Learning the core binding sites of the protein with the DNA may help it unlock more regions of the sequence space by just learning that key mechanism. You can almost cover most of the regions of sequence space just from that main role, but then there are also other variations that might help the model generalize everywhere.

But of course, not every data set, not everything we’re modeling, is that simple. Sometimes we want to understand gene regulation, and there’s many complex processes happening where proteins are binding dynamically, they’re interacting with each other over distant positions. These are all temporal and dynamic, and we have to think of their spatiotemporal dynamics within the cell, within the tissues and organisms, and so forth. So this is much more complex.

I think the right thing to think about for data generation is that the data space is vast and impossible to sample everything, but there’s a latent mechanism space that governs all of the data space. For me it’s sequence space. And learning those mechanistic rules helps it to generalize, to unlock more regions of the data space.

And so the question we should be asking is, is it just more data? Well, more data could just be sampling the same subset of mechanisms over and over again, because you can sample that nearly infinitely, but you’re not going to uncover all the other mechanistic rules. So what we need these AI models to do is to learn those mechanisms, but also to have them guide us on all the unknown mechanisms that it doesn’t know yet.

In this way we can start sampling beyond just thinking about data as just more data, but thinking about data through the lens of informative data. What kind of data points, if you understood it, would unlock more data regions that you don’t need to measure now because you understand those rules so well?

Zhane Yamin: So then I guess just zooming out a little bit, are there any other fields or disciplines where you feel like this convergence step has worked really well? Are there things that we could learn from them or generalize? And if so, where does that generalization break down?

Peter Koo: Yeah. So there are several fields that have undergone some prototypical examples of convergence: astronomy, when those fields have to decide what the next satellite or telescope is going to be, or even particle physics, where you’re converging to decide what kind of particle accelerator is going to be the next one, a linear one or a circular one and so forth.

Now, there’s something interesting happening there, where people who ask very different questions all can come together and decide what technology they want to invest in, in these billion-dollar projects that span multiple countries and multiple subfields within their own field, where every scientist has a different question, and yet they still can figure out what experimental technology they want to invest in. I think that’s really neat that they can do that.

But while that shows an example of convergence, which is something we probably should study a little bit more, it’s also different from other areas where investigators are studying different questions and they can’t use the same exact experimental technique to study those questions.

In particle physics, you accelerate two different protons together and you see what spills out. People who are asking different questions can study the data and slice and dice that same exact data in different ways. Same with astronomy. But in molecular biology, for instance, everyone has their own techniques, because they have to develop these techniques because the techniques to answer their question don’t exist.

And so there isn’t this one resource that, if we just generate at scale, they’ll be able to analyze that data to answer their biological questions. And so this all brings up about challenges in trying to figure out that this convergence.

Zhane Yamin: Alright, I think we’ll end it there. Peter, thank you for joining me on the show.

Peter Koo: Thanks for having me.

Zhane Yamin: My guest today isn’t a scientist, though I think you could call him an engineer of sorts. Chris Fussell is president of business operations at Lila Sciences and vice chairman of McChrystal Group. Before that, he spent 15 years as an officer on Navy SEAL teams. And suffice it to say, Chris has got some knowledge in trying to get large siloed organizations to move like fast connected networks.

Luckily, or perhaps unluckily for us, the same thing could be said about science. That’s why I think it’s such an amazing opportunity to talk to him right now, because the question of whether America meets the AI for science moment might not be entirely a question of science. It might be a question of whether a sprawling enterprise, whether the agencies, labs, universities, and companies can act like they’re on the same team. And that’s the one problem Chris has a lot of experience solving. Chris, welcome to the show.

Chris Fussell: Thanks, Zhane. Yeah, and I think the scientists that I work with at Lila would definitely agree that I’m not a scientist. But I appreciate it. Yeah.

Zhane Yamin: No problem. So I guess I just wanted to start with an honest assessment because I think when we talk about scientific competitiveness, there are a lot of alarm bells that are rung, and they’re rung pretty frequently. And I don’t want to just ring them like everyone else. So I want to start off with one question: when you look at where the U.S. actually stands scientifically right now, how competitive are we?

Chris Fussell: The entire field is at a transition point, right? So there’s a lot of unknowns in that question, right? Go back two years, where are we competitively? We’ve got, you know, but a 60-year advantage, post-World War II, right? But now as AI starts to intersect with science, which is a totally unique space inside of the AI zip code, you know, AI and science is a very specific neighborhood inside of that much broader world. Where are we competitively in that neighborhood? I feel very good about it.

But it’s not one—it needs a special type of attention, both from the U.S. government and allies and from companies like Lila. And there are not a lot of companies in our space. I would say we are uniquely positioned globally. But where are we in that competition globally? I think we have an advantage and I think we are ahead of others, but I think it requires a certain level of attention and understanding from government authorities, private industry, et cetera. And if we don’t give it the proper attention, there’s a world where you get passed. I’m confident we won’t, and I’m confident we have the right pole position right now, but we can’t take our eye off of it.

Zhane Yamin: So one thing I kind of want to track with that is, honestly, this idea of watching, and trying to know what we need to watch. Because I think when we’re talking about scientific competitiveness, there’s always a lot of metrics that are thrown around, like China’s taking us in this, but we’ve got this in the U.S. And so I want to get your thoughts on what sort of things we should be watching and metrics we should be taking account of when we’re talking about AI for science and AI competitiveness globally.

Chris Fussell: Yeah, on the AI for science question, and again, caveating with your opening, I am not a scientist, but I spend my days with them. And so there’s, again, in this neighborhood inside of the AI discussion, once you’ve seen this, you can’t unsee it, which is: the winners of AI and science, I believe, will have to do what we are doing. And this gets to your point of measurability and how do we know how far along we are?

Lila’s thesis and mission and what we are building is an AI model that’s able to connect with physical labs. We’re building many tens of thousands of square feet of labs that will be controlled by AI models, with autonomous systems, robotics, humans very much involved in the labs, which allows incredible throughput, right? And so I like to think of it as a time machine for science. We’ll be able to reach years, decades into the future and pull knowledge into today. So what will be measurable there, rather than just the speed of your model or a lot of stuff that the average person just can’t really get their heads around?

We’re right on the cusp of being able to look at a lab and say—sort of like what happened in manufacturing, if you go back to, I don’t know, the post-industrial revolution all the way through the early 20th century—we changed the way we looked at output. Like, how much can this factory produce per square foot of X, and how much energy does it need? It wasn’t just a human question or a knowledge question or an artisan question. You could look at the size of a factory and say it’s going to be able to produce this much X.

We’re going to be able to do that with science. Like, this lab, with this configuration and the power of this model, can produce this much new scientific knowledge. And we’ll eventually be able to measure that, I believe, down to sort of the square foot. And that’s where the measurement question, I think, becomes really interesting and different than it’s ever been in science. We’ve seen that evolution in other industries, but we’re on the brink of doing that in the scientific space.

Zhane Yamin: So on that point, I want to backtrack into your background and the perspective it gives you. Because again, you’re saying you’re not a scientist, but I think that actually offers you a very unique perspective on this entire landscape. How does that inform your work with Lila Sciences and this idea of autonomous science and building out these labs of the future? What does that strategic perspective offer you in that sense?

Chris Fussell: Yeah, just quickly on my background. My first career, for about 15 years from the late nineties until 2012, I guess, I was an officer in the SEAL teams. And so for your listeners, the U.S. special operations community is now, I don’t know, twice as big as it was 25 years ago, but say 70,000 people. Each branch has its own component of special operations. The SEAL teams sit inside the Navy. And there’s a long history to that that goes back to 1980 when those communities were really created. For the first 20 years or so of their history—and this is the force that I joined in the late ’90s—they were built for siloed excellence. The SEALs had to be really good at X. Army units had to be really good at Y, et cetera. And there wasn’t a world where we’d have to integrate together.

Well, post-9/11, now 25 years ago, we were put into a little bit of a time machine ourselves, because we jumped onto a battlefield that was networked and digitally enabled, and our much less capable, much less well-funded adversaries were connected globally through a digital network, right? That’s obvious now, right? It’s at our fingertips every day. Twenty-five years ago that was new and interesting, and it broke the way that we had to operate. And we had to rebuild and say, okay, how do we create a truly networked set of units that weren’t built for that culturally, operationally—none of our systems integrated, from the human level to the technical level? But we have to be able to do that to defeat this adversary.

And so we went through this whole rebuilding process, and then I spent years post-military thinking about that, writing about that. I teach on it, and then obviously built a consulting firm based on it with retired General Stan McChrystal, who was sort of the father of this inside of special operations. So that’s a quick background on what I think is hyper-applicable here, because if you go back just a few minutes, what we were talking about is building these highly integrated lab systems.

AI doesn’t care if you’re a chemist or a life scientist, et cetera. Eventually it will just understand, like we see in large language models, except there are way more combinations in science. It’s just going to see atoms and molecules and the way that life works and the universe works. But you have to teach it through all those different verticals that we as humans have come to understand. And so the chemist needs to work with these labs and develop an understanding of the AI model of how chemistry works. Same with life sciences, same with material sciences.

All the different expertise that goes into building these types of labs, from AI specialists to software engineers that can connect the AI model down into the physical labs, the labs that need to be built in a very special way that don’t exist yet, the robotics that have to go in and operate in a way that doesn’t yet exist—all of these are very cutting-edge, interdisciplinary requirements. You don’t have to squint too hard to see the parallels with bringing together a global task force of highly siloed, highly competitive, uniquely special people. Right?

So all big complex problems come down to very strong-willed and thoughtful people living in their lane, and you have to connect them across if you want to truly function as a network. So there’s an evolution story there, and that’s how I ended up in this very fun, fortunate, interesting position with Lila.

Zhane Yamin: So do you think, in terms of people currently working on meeting the moment in AI for science, that what you see is a lot of people trying to brute force it—like, we need more budget or we need more force, we just need to try really hard—instead of trying to rework these things? Or do you see that a balance is necessary, in terms of brute forcing it, or maybe the problem is more funding, and then maybe half of our attention span should go to rewiring, I guess? What is that balance that you see, and what do you think it should be? What’s the difference between those two, I guess?

Chris Fussell: I mean, we’re a young company, so I’m always going to say more money, right? Especially now, more compute. You want all the talent that you can find, right? You need the best people in the world. So, yeah, some of those statements are obvious, but I do think one of the big bifurcations in this world right now is: give me more compute, more power, more and more, and I can do all of this in silico. Eventually the model will just understand chemistry. Well, I think that will get us to a certain point where we’re seeing strides there, and that’s all important. But eventually you have to be creating and finding new data that you’re verifying in the real world. If you think you can build that entire world in silico, I think that’s going to hit a wall, right? So we are building labs, which is hard, and that’s a lot of labor, it’s a lot of money, it’s a lot of space. But we’re doing that because we believe that the world, the universe, is the ultimate verifier. Science has been done in the world, not just in people’s brains, since we’ve had science for a reason. You have to do something in the real world to learn how does it actually work out here in this madness we call the universe. And you can get a lot done in silico. Obviously, even our system will bounce back and forth between those two, because you can move incredibly fast when you’re doing it just on compute. But there are going to be times, clearly, where you need to jump out into the physical world and say, let me check that. Let me make sure that works. And so I think that’s the big separation that is already happening, but it’s going to become more and more of the conversation as we move forward.

Zhane Yamin: So one thing I want to take a second to expand on for our audience—I think you talked about it a little bit before—when you’re talking about this work you’re doing at Lila, bringing this rewiring approach to how we approach science and meeting the moment for AI for science, what does that actually look like? Or what are the things that need to get rewired, or what are the nodes in the network, I guess, would be a question that I have.

Chris Fussell: You mean organizationally?

Zhane Yamin: Yeah, organizationally.

Chris Fussell: Yeah, I mean, I’ll compare this to the world I grew up in. This isn’t unlike a lot of governments, a lot of big companies, et cetera. The way we described this back then was: the whole U.S. national security apparatus had been built for generations to do specific things in purpose-built silos, especially inside of special operations. And the language I’ll describe here—there are versions of it that you can see in letters that Caesar would have written to his legions, or Lincoln wrote to his generals in the Civil War. Our language was find, fix, finish, exploit, and analyze. So in the counterterrorism world, there are groups or individuals behind these things, right? It’s not huge numbers.

You’re not looking for the German army. So you have to find where this group is, because they tend to want to hide, right? So there are certain agencies who specialize in finding things, right? To go too deep—that’s probably pretty obvious if you’ve watched any movies about spies. Like, I’m going to go find this person or this group. And then there are other ways that you fix them in time and space. And again, like President Lincoln, he has letters to his generals in the Civil War: I need you to send a reconnaissance force into the Shenandoah Valley and find out exactly where so-and-so’s army is right now. So they are fixing that army. They are in this part of the Shenandoah. And then they can do the math and say, well, they’re going to be here, here, here within a week. We have agencies that go fix things in time and space. It’s a little more technical now, but it’s the same idea. And then I was part of a finish force. Let’s go finish that problem. Send the army out, send the unit out.

And then for years we were trained to grab what we learned from that finish exercise—raw intelligence, insights, et cetera—bring it back and give it to units who specialized in exploitation. I need to download that hard drive, I need to look at those pictures, whatever it is, and I need to clean up my intelligence picture. Then even other groups will analyze that compared to the broad global understanding of what’s going on with this terrorist element. And then you feed it back into the cycle, right? So you find, fix, finish, exploit, analyze, repeat. Those were built in silos so that senior leadership, presidents, secretaries of X, whatever, could control the pace that it moved. And in the 1980s or ’90s, you didn’t have to do that very fast, right? That process I just described could take months, and that’s totally fine. We got into a world where that—

And we used to call the gaps between those blinks, because you were sort of closing your eyes for a second as you hand off to the next organization. And we realized we’re good at all of these, we’re losing on the blinks. Because when you’re fighting in a truly complex, fast-moving battlefield, nobody else is blinking, right? So we have to close the gap between those. And so eventually that became—we described it as a constant circle, which was networked, so you were never blinking. There was always information flowing between all of those different elements. That’s a long-winded way to say I think the exact same approach is necessary inside of science.

And if you think of the scientific method, where you observe—you see an apple fall out of a tree, you make a hypothesis about how gravity might work, you design experiments, you conduct those experiments, you write down all the math that you’ve learned, you refine your hypothesis, you feed it back in—that’s a similar sort of cycle. We put humans into a lot of those components right now, and I could map that over to Lila, but it’s probably obvious. Now, in today’s world, what we’re doing, when you can apply both autonomous labs that can run that nonstop, humans that can plug in where and when they want to, and an AI that can network across all of that and is in a constant state of learning—it doesn’t see the difference between what you’re doing over here and what I’m doing over here.

It can be everywhere at once if it’s allowed to. All of that accelerates, obviously, in a massively exponential version compared to what I was describing in the world I grew up in. But the same sort of theory applies. You have to look at yourself as building organizations that are in a continuous loop of learning and then networked on top of that, so that the cross-functionality is happening in the background without ever thinking about it.

Zhane Yamin: So I guess just an added question on that. I’m getting this idea that we have these fast loops, like on the battlefield. That’s what we want, right? And the analogy to science would be this loop of the autonomous lab or the scientific method and scientific process. There are no blinks in between them; we’re running them as fast as possible. Humans are plugging in where they need to be.

A question that I have is: I think a lot of people have an understanding of scientific institutions. So when you’re thinking about funding institutions or other sorts of scientific institutions that researchers may liaison with, it’s kind of cliché at this point, but maybe that they’re slow or they’re bureaucratic. So I wonder how much you’re thinking, at Lila, when you’re creating these platforms—you’re creating this super fast loop, but is the institution around it, or these bigger agencies?

I’m not sure how much you at Lila specifically might liaison with them, but how do you think of—to maybe make a poor analogy to the defense world—you have this fast-moving battlefield, but when you’re talking about the broader Department of Defense or something, or people not on the battlefield, are they moving at a slow speed, and how does that impact what you’re able to do on the battlefield?

Chris Fussell: It’s a great point, and not to sound cliché, but I think it is actually a proper use of the concept of a paradigm shift from Kuhn’s work, you know, forty years ago. Which is to say, there’s an entire ecosystem of thinking, of organizations, a bureaucratic process that’s going to have to change as a result of this shift that’s happening inside of science. And you named a few. There are many, many more than that, right? Because we built a very specific paradigm based on—one of the key components is how fast can a human learn something?

You know, let’s make it very simple. I’m a researcher, I have to get through school, then I’m going to get into a lab and I have to put stuff in beakers and da-da-da-da. It’s a very methodical process. So you build all these systems around it, and the system can predict what the throughput will look like, how much time I need to assure that that new finding is safe and releasable, and can I go through my testing phases, et cetera. Everything we know. I pass laws about it, I give funding to certain things on this end, whether it’s a university or an individual lab, et cetera.

All that works—sure, it’s problematic, but it’s better than any other system we’ve ever come up with, right? But when one of those things gets exponentially faster, everything else is going to have to change over time, right? The way we think about releasing drugs to the market in, pick a decade, fifty years from now will be archaic. It just will have to be, right? We’re not the ones that are going to change that, right? But we are, again, in the zip code we live in, a wonderful upward pressure that will start to influence all of those other systems.

And when there is—I’m, of course, making this up—at some point there will be a backlog of new drugs in a certain category, and we’re just going to realize the system doesn’t work anymore. We have to figure out how to move these things faster because there are real lives on the line right now. Or new discoveries in material science—you could just pick your category and say, wow, these systems now, we’re feeling this pressure. A lot of other things have to change around it. And I think we wouldn’t have to think too long to find historic analogies of having to massively increase the speed with which we allow certain things to happen as technology advances.

So, yeah, we are one of many in the broad discussion around AI that will end up influencing that. I do think Lila will have a specific voice in that, because we’re going to be doing amazing things in this autonomous lab space before many, many even are aware that it’s happening. So that’s exciting. I think on the government side, there are a lot of folks in that world that are thinking forward on this topic. But it’s going to take more legislators understanding what’s happening in this space. It’s going to take more government focus, government dollars, government influence to try to get ahead of that problem set as well.

Zhane Yamin: So then, not to lean too hard on the historical analogies, but I always wonder about this idea. I think it’s recent that people think about the Sputnik moment, right, when they think about technological competitiveness—and this idea that, instead of a gradual change that slowly shifts paradigms, something happens and everybody flips into a different mode of operation.

Not to make any predictions here, but what do you see? Do you see it being a moment where something happens, either on the international or the domestic stage, and that paradigm shift happens? Or do you think it’s more likely what you’re saying, where it’s a backlog of maybe drugs and at some point people are like, all right, we’ve just got to switch this up, and this has been mounting for a while? Maybe not any specific predictions, but what do you think about that sort of dichotomy?

Chris Fussell: Yeah, I mean, I was a consultant for years, right? So the answer is it’s a little bit of both, right? There will be a series of wow moments, and there will be a series of, you know, now we’ve got to change things to catch up with this. I think the wow moments will—again, it’s fascinating, exciting, sometimes mind-boggling to spend all day every day with brilliant PhDs. They have these moments regularly. But for the average duck, you’re like, I guess that’s really interesting, right? But that’s true in any deeply technical space.

There will be moments in the very near future, I’m quite confident, where the public will be able to hear the story and say, wow, that material used to weigh this much and now it weighs this much, or at least we’re on the path to that. That means my vehicle is going to be lighter. Or that means this offshore drill now has a lubricant, and the parts under salt water and high pressure conditions will last ten times longer, right? And so now we’re reducing all sorts of environmental impacts. There are going to be these wonderful, very translatable wins.

I think the material science place is the one that will shock and surprise the world the most, because we’re surrounded by materials all the time. And when those things start to change, I think that’s going to be one of the first door openers for the public to see, okay, science is really changing the world around me at this accelerated pace. Because most of us have never—those things happen slowly, right? I was kidding around with somebody today, like WD-40: we all have a bottle of it somewhere in our house. It’s the same thing my grandfather used, I think, right?

But all of those little micro changes—we usually don’t see any of those in our world. We’re going to see a lot of them in our lifetimes, and they’re going to have tangible impact that people will feel. And then, of course, the same thing in life sciences, but I think materials is where people are going to really start to connect with this change that’s happening in science.

Zhane Yamin: Wow, yeah, that’s great. I was going to ask about what’s on the table here, and I think you just covered it perfectly. I think for somebody listening at home, and what they can take home and be like, why does this matter, even those materials examples are a very clear example of it.

So now that we have what’s on the table, if you were talking to the policymaker or the policy analyst that’s listening to this podcast, what’s one thing that you would want them to take away from this? What’s one thing where you’re like, this is something that we need everybody to be on the same page about, that I see as a big problem, where people are often either butting heads or they’re not sure, they’re not informed? What is that one issue, and what would you say to that?

Chris Fussell: Well, I guess for the D.C. audience—and I do say we’re down there a lot—there are real strategic issues at play here. Probably stating the obvious based on our conversation so far. But part of the reason I was passionate about coming to Lila from the very start was, like I said at the beginning, once you see this, you can’t unsee it. I grew up in a world in the military where technology changes, material science changes, they have real impacts very quickly, right?

Pick your example from the special operations world: improvement to lasers, improvement to night vision devices, improvement to thermal imagery, all these things. Those happened relatively quickly when an advanced military is at war. I saw many of those cycles take place over a decade. If you turn up the speed on that—not just for military application, obviously, but broadly—there are significant strategic implications as you look at how you then interact with your adversaries. How do allies move up this together and stay connected as technology moves so quickly?

So I think for policy folks, one, be very aware that this is going to be a fast-moving shift with global implications for how we stay connected with our allies, and aware of the balance with adversaries. Because I do believe in geopolitics, national security, if you get too disproportionate, even unintentionally, you can create second-order problems you didn’t want to, right? Because you can force a reaction, like, now I’m really nervous, so I better do something before you’re too far down that path, right? So we want to avoid those.

But then thinking downstream, what are the implications of these sorts of advancements? Look at things like our industrial base. How many things have been happening the same way for generations? And how quickly can that shift? We talked about some examples a moment ago, but how will this impact agriculture? How will this impact the materials we produce every day with a very high energy cost? Look at supply chains globally.

What materials are we dependent upon from nations we wish we weren’t, right? Are there alternatives we’ll be able to create and produce at home or in allied nations? The answer is yes, right? So huge implications there. When there are fast-moving problems—I mean, COVID’s an obvious big example, but you can imagine others—how much faster can our response get to problems like that with this sort of time machine for science, so to speak?

Again, massive implication, and one that you want to make sure you’re keeping the right people informed on as you move up that speed chain, right? So it’s one of those arenas where you want to think, okay, what are the possible, likely outcomes over the next five to ten years? And then let’s assume we get there, what are the downstream effects that that’s going to create? And how do we start mapping that into how we’re thinking about policy, funding, et cetera, in the here and now?

Zhane Yamin: Good advice to heed. I think that’s a perfect ending note. Chris, thank you for joining me on the show.

Chris Fussell: Thanks, Zayn. This was great.

In a world where scientific discovery is often hampered by the limited number of exceptional scientists, Sam Rodriques, co-founder and CEO of Edison Scientific, sheds light on how AI can bridge this talent gap. With a background in physics and a history of pioneering work in AI for scientific research, Sam is at the forefront of a movement to augment human talent with intelligent systems.

Defining AI Scientists

Sam begins by addressing the common misconceptions surrounding AI scientists. He notes that while the term is frequently used, its meaning can often be unclear. For Sam, the purpose of an AI scientist is to **augment human scientific talent**. By providing access to exceptional scientific intelligence, AI scientists aim to accelerate the pace of research and discovery.

“Removing talent as the bottleneck in science is going to be one of the fastest ways to accelerate progress in science today.” — Sam Rodriques

The Role of AI Agents

Sam explains that the AI agents developed at Edison Scientific are designed to operate within the scientific process, facilitating everything from hypothesis generation to laboratory experimentation and data analysis. By collaborating with pharmaceutical companies, these agents are already making novel discoveries that can be validated in real-world settings.

Key Components of Scientific Progress

According to Sam, successful scientific inquiry requires three essential components: intelligence, knowledge, and experimentation. He emphasizes that while intelligence in AI is becoming increasingly commoditized, the focus will be on ensuring that models have the knowledge and experimental capabilities necessary for groundbreaking research.

Addressing the Knowledge Explosion

With the vast amount of scientific literature published today, it’s becoming impractical for any single individual to keep up. Sam argues that AI scientists are uniquely positioned to manage this overwhelming influx of information. Their ability to aggregate knowledge and facilitate new discoveries can lead to significant advancements in various fields.

“The proliferation of knowledge means that there’s now too much for any one person to hold in their head. AI can help manage and leverage that information effectively.” — Sam Rodriques

Infrastructure and Sovereignty in AI

As AI tools become more integral to scientific research, the infrastructure supporting these technologies is evolving. Sam highlights the importance of **sovereignty in AI**, where companies need to run models independently to protect their proprietary data. This shift is crucial for companies that want to leverage AI without relinquishing control over their most valuable asset—their data.

The Importance of Benchmarks

To evaluate the effectiveness of AI agents, Sam and his team have developed benchmarks like Labbench and Bixbench. These tools assess the performance of AI in scientific tasks, ensuring that as capabilities advance, the benchmarks do too.

“You can’t tell whether or not the thing that you’re doing is correct if you don’t have a benchmark.” — Sam Rodriques

Conclusion

Sam’s insights into the role of AI in science reveal a future where human and artificial intelligence work together to push the boundaries of research. As the landscape of scientific inquiry evolves, the integration of AI promises to unlock new possibilities for discovery and innovation.

For those interested in the intersection of technology and science, this conversation offers a compelling glimpse into what the future may hold.

Rafael Gomez-Bombarelli sits down with Zhane to discuss the intersection of AI and science, exploring how AI models are transforming research in chemistry, materials science, and beyond. He shares insights on the evolution of AI for science, the role of scalable architectures, and the future of autonomous laboratories.

The Evolution of AI in Science

Rafa’s career illustrates a pivotal moment in science when computational methods began merging with experimental research. He shared,

“Back around 2014, 2015, the simulation community realized that what they were doing was highly scalable.”

This scalability allowed researchers to harness vast amounts of simulation data, coinciding with the rise of deep learning techniques. As Rafa explained, this period marked a significant shift:

“We had the ability to make large data sets and deep learning architectures that performed well on these data sets.”

He emphasized how the emergence of specialized neural networks transformed the way scientists approach data and simulations. The integration of machine learning into scientific research has sparked what Rafa terms the first incarnation of AI for science.

The Bitter Lesson

Rafa introduced the concept of the “bitter lesson,” which posits that the most effective AI models are those that leverage large datasets rather than relying on human intuition or domain-specific knowledge. He noted,

“The models that ultimately succeed are the ones that can just take the most available data into the most expressive general models.”

This understanding has profound implications for how scientists approach problem-solving in their fields, suggesting a need to adapt to more scalable methods of inquiry.

A New Era for Scientific Discovery

So why is there such a surge of interest in AI for science right now? According to Rafa, the urgency of scientific advancement is a major driver. He articulated that

“Science is a very noble, important, and sort of unambiguously good for humanity.”

The need to innovate, especially in fields like medicine and technology, creates a fertile ground for the application of AI. As Rafa pointed out, science inherently generates new knowledge, making it a compelling domain for AI applications.

The Role of Lila Sciences

At Lila Sciences, Rafa is pushing the boundaries of what’s possible with AI in the lab. He describes a vision for creating programmable laboratories or “AI Science Factories” that integrate AI’s capabilities with experimental verification. He explained that,

“The only way to have scalable verification for AI-driven science is to give the whole wheel of science to AI.”

This revolutionary approach aims to automate not only data analysis but also the execution of experiments, allowing AI to verify its hypotheses in real-time.

Rafa’s work serves as a bridge between language-based AI models and the nuanced, complex world of scientific inquiry. His role involves translating scientific concepts into formats that AI can understand, ensuring that the rich data of the scientific domain can be effectively utilized by machine learning models.

In this exciting intersection of AI and science, we are witnessing a transformation that promises to accelerate discovery and innovation like never before. To explore more about these insights and the future of AI in scientific research, listen to the full episode.

Chapters

00:00 Introduction to Rafa Gomez-Bombarelli and his background in AI and science

01:24 The merging of experimental science and simulation in Rafa’s career

02:24 The rise of data science and deep learning in materials and chemistry

03:20 The impact of neural networks and large data sets in AI for science

04:20 The difference between physics-based simulation and machine learning models

05:43 Generative models for chemistry and materials design

06:37 How language models relate to scientific data and models

09:03 Transformers and the scaling of AI architectures

10:38 Why now is a pivotal moment for AI in science

Lila Sciences and the role of AI in scientific discovery

The concept of AI Science Factories and autonomous labs

Concrete examples of AI enabling scientific breakthroughs

The potential and limitations of AI in generating new knowledge

Open questions and the future role of human scientists

In this episode, Dr. Prachee Avasthi discusses the transformative potential of open science, challenging traditional publishing and funding models, and exploring how technology and policy can accelerate scientific progress.

Chapters

00:00 Introduction to open science and its importance

01:51 What is open science and how is it defined?

04:00 The role of open access and open data in research

06:04 Challenges of traditional publishing and peer review

07:57 Funding models and their influence on research direction

10:10 The impact of AI and new technologies on science dissemination

12:09 Strategies for implementing open science practices

14:02 The distinction and relationship between for-profit and nonprofit research

15:52 Policy initiatives and organizational efforts to promote open science

18:12 The cultural shift needed for open science adoption

20:08 The future of scientific collaboration and data sharing

22:06 The societal stakes and benefits of open science

23:54 Overcoming inertia and resistance in the research community

26:01 The role of grassroots efforts and individual action

28:06 The importance of infrastructure and resources for open science

30:07 The potential of AI and machine learning to transform research

32:04 How to navigate the evolving landscape of scientific publishing

33:49 The impact of open science on innovation and societal progress

36:07 Final thoughts: embracing change for a better scientific future

In this episode, we explore the Genesis mission’s focus, funding challenges, and the importance of clear communication in AI for science. Dan Turner-Evans of the Institute for Progress shares insights from his diverse background in physics, neuroscience, and policy, emphasizing the need for strategic focus and effective messaging to advance AI initiatives in government and research.

Chapters

00:00 Introduction to the Genesis mission and guest background

02:01 Dan’s diverse scientific background and policy experience

07:08 Overview of the Genesis mission and its recent expansion

11:10 The importance of focus and funding for the mission

15:11 Challenges of broad AI initiatives and strategic focus

19:52 Communication strategies for AI in science to the public and Congress

25:08 Platforms versus challenges in AI research and development

30:00 Balancing basic research with targeted AI applications

34:53 Concluding thoughts on focus, communication, and the future of AI in science

In this episode, Josh New, director of policy at SeedAI, shares insights on accelerating science through AI, policy challenges, and innovative models like autonomous labs and FROs. We explore how to unlock AI’s potential for scientific breakthroughs and the role of policy in fostering a more dynamic research ecosystem.

Chapters

00:00 Introduction and episode overview

01:00 Guest introduction: Josh New and SeedAI

02:05 Josh’s path from tech policy to AI for science

03:23 The rise of AI in scientific research

05:14 Changing conversations in AI policy

07:43 Rapid evolution of AI and policy challenges

08:35 The impact of AI saturation on scientific workflows

12:00 Unlocking AI’s potential versus steering its direction

14:02 Decentralized science and policy intervention

17:28 Reimagining research labs with AI and automation

20:55 Funding models and data infrastructure for AI-driven science

23:07 Lessons from the Human Genome Project and big science

29:45 The ASN model as connective tissue in AI and science

34:30 Closing thoughts and call to action

Episode Resources

SeedAI - https://seedai.org

AlphaFold - Protein structure prediction - https://deepmind.com/research/highlighted-research/alphafold

Human Genome Project - https://www.genome.gov/human-genome-project

ASN Coalition - Accelerate Science Now - https://acceleratesciencenow.com/

Introducing State Space

In this trailer, host Zhane Yamin introduces State Space — a new show exploring the ideas shaping science acceleration, AI for science, and meta-science. Drawing on the concept of a "state space" (the mathematical universe of every possible configuration a system can take), Zhane frames the show's core question: how do we find the right variables — the right basis — to cut through the noise and make sense of a research ecosystem that's evolving faster than ever under AI?

Each episode of State Space will feature a conversation with a leading voice in science acceleration, offering a new angle on how science moves faster in an AI-native era.

In this episode:

- What a "state space" is, and why it's a useful lens for thinking about complex systems — from a mass on a spring to a neural network to a game of chess

- Why the scientific research ecosystem is itself a kind of state space, full of decades of accumulated ideas and new AI-driven breakthroughs

- What to expect from the show: conversations at the intersection of science acceleration, AI for science, and meta-science

Host: Zayn Yaman, Science Policy Fellow at Seed AI

About the show: State Space is powered by the Accelerate Science Now Coalition, a Seed AI project.