The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - Beyond the moment of invention by jasoncrawford
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Beyond the moment of invention, published by jasoncrawford on December 16, 2022 on LessWrong.
    Derek Thompson has a feature in The Atlantic this week on “Why the Age of American Progress Ended”—thoughtful and worth reading. Some comments and reactions.
    One of the ideas explored in the article is that what matters for progress is not just the moment of invention, but what happens after that. I generally agree. Some ways I think this is true:
    The first iteration of an invention is generally just good enough to be practical, but far below optimal: there are decades of incremental improvements that get it to what we know today. Edison’s light bulb was not as bright or long-lasting as today’s bulbs.
    Often there is a whole system that needs to be built up around the invention. It wasn’t enough to invent the light bulb, you also needed the generators and the power grid.
    Such a system not only has to be invented, it has to be scaled. Scaling up from a prototype to a large, efficient, reliable system is its own challenge. Again with the power grid, it was a big challenge to figure out how to efficiently serve large regions, do load balancing, etc. With railroads, there was a challenge in figuring out how to manage a schedule with many trains and routes. With the telegraph and later the telephone, a system had to be invented to route messages and calls.
    Merely working is not enough for wide distribution: other characteristics really matter, like cost, efficiency, and reliability. People underrate these—especially reliability, which can be a huge barrier to adoption.
    An invention that works and is practical, cheap and reliable still doesn’t automatically sell. You have to convince people to change the way they do things, and that is tough. Sometimes you have to help people imagine uses for things: when the telegraph was first invented, they would demonstrate it by playing chess long-distance, and people would come watch this happen, but still not imagine what they personally would use a telegraph for.
    Often regulations and legal frameworks have to be updated. E.g., containerization: the ICC regulated rates and they set rates based on the type of cargo; with containers you want to charge based on volume and weight alone, and this clashed with the regulatory framework. This kind of thing happens all the time.
    Even after an invention is widely available, it can take further decades for all the implications to be worked out and for it to fundamentally change the way people do things. Electricity didn’t cause factories to be reorganized until the ’20s or ’30s. Containerization didn’t immediately change the way supply chains were organized.
    And then, social or regulatory barriers can block distribution. In addition to the ones discussed above, another major one is pushback from labor when jobs are going to be automated (many historical examples).
    A lack of good institutions in poor countries can also block distribution, which is why the world is so unevenly wealthy today.
    On another note, I thought this paragraph near the end of the piece was spot-on:
    When you add the anti-science bias of the Republican Party to the anti-build skepticism of liberal urbanites and the environmentalist left, the U.S. seems to have accidentally assembled a kind of bipartisan coalition against some of the most important drivers of human progress. To correct this, we need more than improvements in our laws and rules; we need a new culture of progress.
    Amen.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    4 min
  • EA - Concerns over EA’s possible neglect of experts by Jack Malde
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Concerns over EA’s possible neglect of experts, published by Jack Malde on December 16, 2022 on The Effective Altruism Forum.
    This is a Draft Amnesty Day draft. That means it’s not polished, it’s probably not up to my standards, the ideas are not thought out, I haven’t checked everything, and it's unfinished. I was explicitly encouraged to post something like this!Commenting and feedback guidelines: I’m going with the default — please be nice. But constructive feedback is appreciated; please let me know what you think is wrong. Feedback on the structure of the argument is also appreciated.
    I am becoming increasingly concerned that EA is neglecting experts when it comes to research. I’m not saying that EA organisations don’t produce high quality research, but I have a feeling that the research could be of an even higher quality if we were to embrace experts more.
    Epistemic status: not that confident that what I’m saying is valid. Maybe experts are utilised more than I realise. Maybe the people I mention below can reasonably be considered experts. I also haven’t done an in-depth exploration of all relevant research to judge how widespread the problem might be (if it is indeed a problem)
    Research examples I’m NOT concerned by
    Let me start with some good examples (there are certainly more than I am listing here!).
    In 2021 Open Phil commissioned a report from David Humbird on the potential for cultured meat production to scale up to the point where it would be sufficiently available and affordable to replace a substantial portion of global meat consumption. Humbird has a PhD in chemical engineering and has extensive career experience in process engineering and techno-economic analysis, including the provision of consultancy services. In short, he seems like a great choice to carry out this research.
    Another example I am pleased by is Will MacAskill as author of What We Owe the Future. I cannot think of a better author of this book. Will is a respected philosopher, and a central figure in the EA movement. This book outlines the philosophical argument for longtermism, a key school of thought within EA. Boy am I happy that Will wrote this book.
    Other examples I was planning to write up:
    Modeling the Human Trajectory - David Roodman
    Roodman seems qualified to deliver this research.
    Wild Animal Initiative research such as this
    I like that they collaborated with Samniqueka J. Halsey who is an assistant professor
    Some examples I’m concerned by
    Open Phil’s research on AI
    In 2020 Ajeya Cotra, a Senior Research Analyst at Open Phil, wrote a report on timelines to transformative AI. I have no doubt that the report is high-quality and that Ajeya is very intelligent. However, this is a very technical subject and, beyond having a bachelor’s degree in Electrical Engineering and Computer Science, I don’t see why Ajeya would be the first choice to write this report. Why wouldn’t Open Phil have commissioned an expert in AI development / computational neuroscience etc. to write this report, similar to what they did with David Humbird (see above)? Ajeya’s report had Paul Christiano and Dario Amodei as advisors, which is good, but advisors generally have limited input. Wouldn’t it have been better to have an expert as first author?
    All the above applies to another Open Phil AI report, this time written by Joe Carlsmith. Joe is a philosopher by training, and whilst that isn’t completely irrelevant, it once again seems to me that a better choice could have been found. Personally I’d prefer that Joe do more philosophy-related work, similar to what Will MacAsKill is doing (see above).
    Climate Change research
    (removed mention of Founder's Pledge as per jackva's comment)
    Climate Change and Longtermism - John Halstead
    John Halstead doesn't seem to have any formal training in climate science. Not sur...
    4 min
  • EA - Seeking feedback on a MOOC draft plan: Skills for Doing Good Better by Michael Noetel
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Seeking feedback on a MOOC draft plan: Skills for Doing Good Better, published by Michael Noetel on December 16, 2022 on The Effective Altruism Forum.
    This is a Draft Amnesty Day draft. That means it’s not polished, it’s probably not up to my standards, the ideas are not perfect, and I haven’t checked everything. I was explicitly encouraged to post something imperfect! Commenting and feedback guidelines. I'm slightly nervous about the amount of feedback I'll get on this, but am doing it because I sincerely do want constructive feedback. Please let me know if my assumptions are wrong, my plans misguided, my focus for the course poorly calibrated, or my examples un-compelling. Google Doc for commenting available upon request.
    A few months ago, I posted a diatribe on how to better use evidence to inform education and field building in effective altruism. This a draft where I try to practice what I preach. John Walker and I have been exploring ideas for another EA Massively Open Online Course (MOOC). This is currently Plan A.
    Big-picture intent
    Why make a MOOC?
    There are many great forms of outreach to build the community of people who care about effective altruism and existential risk. In contrast with most approaches, MOOCs provide:
    Authority, via their university affiliation
    Credentials for completion, with a university badge
    Minimal marginal cost per learner, by design
    Designed well, they can provide high-quality, evidence informed learning environments (e.g., with professional multimedia and interactive learning). These resources can feed into the other methods of outreach (e.g., fellowships).
    Course pitch to learners
    Many people want to do good with their lives and careers.
    The problem is: many well-intentioned attempts to improve the world are ineffective. Some are even harmful.
    With the right skills, you can have a massive impact on the world.
    This MOOC provides many of those skills.
    Through this course, you’ll learn how to use your time and money to do as much good as possible, and give you a platform to keep learning about how to improve the world.
    Underlying assumptions
    To be transparent, most of these assumptions are not based on direct data from the community or general public. Where support is available, I’ve linked to the relevant section of resources. Where not, I’ve listed methods of testing those assumptions but am open to corrections or other methods.
    There are some skills and frameworks used in EA that help us answer the essential question: “how can I do the most good, with the resources available to me?” (see also, MacAskill)
    Examples of what we mean by ‘skills and frameworks’
    current CEA website: scope sensitivity, trade-offs, scout mindset, impartiality, (less confidently: expected value, thinking on the margin, consequentialism, importance/value of unusual ideas, INT framework, crucial considerations, forecasting, and fermi estimates
    See old whatiseffectivealtruism.com page: maximisation, rationality, cosmopolitanism, cost-effectiveness, cause neutrality, counterfactual reasoning
    These skills and frameworks are less ‘double-edged’ than the moral philosophy
    The moral obligations in EA (e.g., Singer’s drowning child) are a source of motivation for many but also a source of burnout and distress (see also forum tags on Demandingness of morality and Excited vs. obligatory altruism)
    In contrast, the evidence that ‘improving competence and confidence leads to sustainable motivation’ is supported by dozens of meta-analyses across domains with no major downsides, to my knowledge
    These skills and frameworks are less controversial than the ‘answers’ (e.g., existential risk, farmed animal welfare)
    As put by CEA, “we are more confident in the core principles”
    Misconceptions about EA (e.g., per 80k: ‘EA is just about fighting poverty’; ‘EA ignores systemic chan...
    42 min
  • LW - Covid 12/15/22: China’s Wave Begins by Zvi
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Covid 12/15/22: China’s Wave Begins, published by Zvi on December 15, 2022 on LessWrong.
    Cases increased this week, as expected. Deaths also increased, which I did not expect so soon, indicating a bigger and faster winter wave. That was not the big event in the West this week with respect to Covid.
    The big event this week in the West’s Covid experience was a memory of events past. Patrick McKenzie wrote a very long and very worthwhile magnum opus about his efforts in VaccinateCA to get vaccine shots into arms, and what he learned about our government and public health systems along the way. If you read one thing this week, read his story. You can also read my key facts and takeaways post, and this thread with another person’s remembrances (here are Patrick’s takes on that). Patrick also offers his thoughts on implications for charitable efforts at the EA Forum, which includes this gem:
    If I had to point at any one artifact in particular, Inadequate Equilibria is a good articulation of a larger memeplex that made me comfortable with “If the evidence of a system’s operation contradicts what the Efficient Market Hypothesis counsels is the probable functioning of the system, trust the evidence. Thousands of lives savable by one dedicated team of non-specialists is actually not all that low probability.”
    Yes.
    What else happened recently?
    China stopped attempting to contain Covid-19. They have accepted that the costs of continuing to do this are unacceptably high. Instead, they have decided they are content to contain official case numbers while they brace for the impact of the Covid surge they have successfully put off for three years. It is their turn to have trouble staffing hospitals, to have panic buying and shortages and voluntarily deserted streets. I wish them the best of luck in getting through this as best they can.
    Executive Summary
    Patrick McKenzie wrote about VaccineCA. You should read it, and remember.
    China is finally experiencing the Covid wave they have postponed for three years.
    Cases and deaths are on the rise, the winter wave is looking a bit larger.
    Let’s run the numbers.
    The Numbers
    Predictions
    Predictions from Last Week: 371k cases (+5%) and 2,240 deaths (-6%).
    Results: 391k cases (+10%) and 2,610 deaths (+9%).
    Predictions for Next Week: 435k cases (+10%) and 2,900 deaths (+10%).
    (Note: North Dakota dumped a lot of deaths on us after reporting no deaths at all for months, so I reduced that number to something high but not impossible.)
    Case number came in modestly higher than expected. Compounding the expectation there makes things likely to get substantially worse than previously expected, yet still nothing at all like what we saw last winter.
    Deaths being up again rather than down (despite the ND adjustment), in a way I did not expect, since the timing on it seems premature, makes it that much more clear this is a real and large winter wave, and indicates that cases have likely gone up a bunch more than the numbers show, with the extra cases not being reported.
    Deaths
    Cases
    Physical World Modeling and Long Covid
    Observational study on Paxlovid in age 50+ population (paper).
    Results:
    During the study period, 12 541 (28.1%) patients were prescribed nirmatrelvir plus ritonavir, and 32 010 (71.9%) were not. Patients prescribed nirmatrelvir plus ritonavir were more likely to be older, have more comorbidities, and be vaccinated. The composite outcome of hospitalization or death occurred in 69 (0.55%) patients who were prescribed nirmatrelvir plus ritonavir and 310 (0.97%) who were not (adjusted risk ratio, 0.56 [95% CI, 0.42 to 0.75]). Recipients of nirmatrelvir plus ritonavir had lower risk for hospitalization (adjusted risk ratio, 0.60 [CI, 0.44 to 0.81]) and death (adjusted risk ratio, 0.29 [CI, 0.12 to 0.71]).
    Strange to say this might be not worth th...
    18 min
  • EA - We should say more than “x-risk is high” by OllieBase
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: We should say more than “x-risk is high”, published by OllieBase on December 16, 2022 on The Effective Altruism Forum.
    This is a Draft Amnesty Day draft. That means it’s not polished, it’s probably not up to my standards, the ideas are not thought out, and I haven’t checked everything. I was explicitly encouraged to post something unfinished! Commenting and feedback guidelines: I’m going with the default — please be nice. But constructive feedback is appreciated; please let me know what you think is wrong. Feedback on the structure of the argument is also appreciated.
    Epistemic status: outlining a take that I think is maybe 50% likely to be right. Also on my blog.
    Some people have recently argued that, in order to persuade people to work on high priority issues such as AI safety and biosecurity, effective altruists only need to point to how high existential risk (x-risk) is, and don’t need to make the case for longtermism or broader EA principles. E.g.
    Neel Nanda argues that if you believe the key claims of "there is a >=1% chance of AI causing x-risk and >=0.1% chance of bio causing x-risk in my lifetime" this is enough to justify the core action relevant points of EA.
    AISafetyIsNotLongtermist argues that the chance of the author dying prematurely because of AI x-risk is sufficiently high (~41%, conditional on their death in the subsequent 30 years) that the pitch for reducing this risk need not appeal to longtermism.
    The generalised argument, which I’ll call “x-risk is high”, is fairly simple:
    1) X-risk this century is, or could very plausibly be, very high (>10%)
    2) X-risk is high enough that it matters to people alive today - e.g. it could result in their premature death.
    3) The above is sufficient to motivate people to take high-priority paths to reduce x-risk. We don’t need to emphasise anything else, including the philosophical case for the importance of the long-run future.
    I think this argument holds up. However, I think that outlining the case for longtermism (and EA principles, more broadly) is better for building a community of people who will reliably choose the highest-priority actions and paths to do the most good and that this is better for the world and keeping x-risk low in the long-run. Here are three counterpoints to only using “x-risk is high”:
    Our situation could change
    Trivially, if we successfully reduce x-risk or, after further examination, determine that overall x-risk is much lower than we thought, “x-risk is high” loses its force. If top talent, policymakers or funders convinced by “x-risk is high” learn that x-risk this century is actually much lower, they might move away from these issues. This would be bad because any non-negligible amount of x-risk is still unsustainably high from a longtermist perspective.
    Our priorities could change
    In the early 2010s, the EA movement was much more focused on funding effective global health charities. What if, at that time, EAs decided to stop explaining the core principles of EA, and instead made the following argument “effective charities are effective”?
    Effective charities are, or could very plausible be, very effective.
    Effective charities are effective enough that donating to them is a clear and enormous opportunity to do good.
    The above is sufficient to motivate people to take high-priority paths, like earning to give. We don’t need to emphasise anything else, including the case for effective altruism.
    This argument is probably different in important respects to “x-risk is high”, but illustrates how the EA movement could have “locked in” their approach to doing good if they made this argument. If we started using “effective charities are effective” instead of explaining the core principles of EA, it might have taken a lot longer for the EA movement to identify x-risks as a top priority.
    Our pri...
    6 min
  • EA - On Epistemics and Communities by JP Addison
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: On Epistemics and Communities, published by JP Addison on December 16, 2022 on The Effective Altruism Forum.
    This is a Draft Amnesty Day post. I wrote it in 2020 and ~haven't looked at it since. I'm posting it as-is.
    An invisible project is one of our most important — let’s try to reveal it
    This community is about doing the most good. We have many conversations about how to do that. In the course of those conversations, we've slowly pushed forward a cultural understanding of "how do we form correct beliefs?" We call that understanding “epistemics.”
    I want to make a few, hopefully-useful observations around this general space. If you haven’t read much about epistemics before, I hope it serves as an accessible introduction. If you’re an old hand, I hope it communicates this frame I’ve found useful.
    I Why is this hard?
    Most of the time, figuring out what's true is easy. When was this bridge built? Look it up on Wikipedia (more on that later). When it's not easy, it's often best to just use the tools someone else has already developed.
    The situations where you really start needing to attack the problem are when:
    The tools you're used to are inadequate for the task at hand, or
    You and a collaborator disagree on how to figure out what's true.
    Once that happens, I claim most people just get really confused. It's like, what the heck is going on? This makes no sense / my collaborator makes no sense. People give up, or can form really deep impasses with those around them. Even if you're fortunate enough to notice the issue for what it is, it can seem really hard to resolve. I think this is what happened when a lot of my friends and I were only half-convinced of AI risk.
    How the fk are you supposed to weigh "we literally have an RCT here" versus, "this other thing would be big if true"? Many people find the answer obvious, but unfortunately not the same way. I hope at least some point in your life you’ve viewed it as a hard problem.
    Then, just when you and your best friend have figured out how to weigh evidence between yourselves, there comes a whole lot of other people. Many new complications arise when this is done as a community:
    Not all the participants are able to get complete information, or evaluate all the arguments
    Some participants are probably smarter than others
    Some participants are probably acting adversarially, or with some level of own-view favoring-bias
    II All project-oriented communities do this
    Maybe you’ve heard people talk recently about epistemics, and it’s felt like a fuzzy concept. I hope that presenting the ways in which a bunch of different communities go about forming beliefs.
    Wikipedia
    As we’ve already mentioned, Wikipedia has some outstanding epistemics. What’s really useful for us here is that they’ve written it down. You can see the way that it's meant for Wikipedia’s particular situation and how it needs to be legible to outsiders and extremely resilient to adversarial action.
    Science
    You can observe humanity making a huge leap forward by improving its epistemics in one important domain. Our species went from just being completely wrong about just about everything in the natural world, to methodically making progress in our understanding. That progress has compounded over time to completely transform the world. It's interesting the note that wasn't obvious with those epistemics should be at first. Are thought experiments valid scientific evidence? And in the soft sciences the epistemics are still controversial.
    Some might claim that science has figured it all out. But what’s the scientific way to predict who you should pick to lead your company? To make most decisions that humans make there is simply too little high quality data.
    Medicine
    Perhaps even more so than science, medicine is extremely conservative in its epistemics. There are so many ac...
    6 min
  • EA - If everyone makes the same criticism, the opposite criticism is more likely to be true by MichaelDickens
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: If everyone makes the same criticism, the opposite criticism is more likely to be true, published by MichaelDickens on December 17, 2022 on The Effective Altruism Forum.
    I frequently hear people say EAs rely too much on quantifying uncertain variables, but I basically never hear the opposite criticism. If everyone believes you shouldn't quantify, then nobody's doing it, so it can't possibly be true that people quantify too much, and in fact the opposite is probably true.
    Obviously I could make various counterarguments, like maybe the people who think we don't quantify enough are not writing essays about how we need to quantify more. Generally speaking, I don't think this counterargument is correct, but arguing for/against it is harder so I don't have much to say about it
    It's like Lake Wobegon, where all the children are above average. It's impossible for every single person in the community to believe that the community is not X enough
    Another example: everyone says we need to care more about systemic change
    Saw a Twitter post "EAs way under-update on thought experiments" and I thought, damn that's a spicy take. Then I realized I misread it and they actually said "over-update" and I thought...wow what a boring take that's been said a thousand times already
    They gave the simulation argument and Roko's Basilisk as examples. As far as I know, nobody has ever changed their behavior based on either of those arguments. It would be pretty much impossible for people to update less on them than they have
    I'm sure there are some people somewhere who have updated based on the simulation argument but I've never met them
    "People under-update on thought experiments" would have been a much more interesting take because people basically don't update on thought experiments
    By a shocking coincidence, I take the opposite side on all these examples: I think EAs should use more quantitative estimates, should care less about systemic change, and should update more on thought experiments
    Are there any issues where I make the same criticism as everyone else, and I'm actually wrong? Probably, idk
    I can think of some non-EA-related examples of this phenomenon, but I'm not as interested in those
    By analogy, the moment when the most people agree the stock market is going to go up is the exact moment when the market is at its peak. The price can't go higher because there's no one left to buy from. If everyone agrees, everyone /must/ be wrong
    Relevant Scott Alexander:/ and/. He said it better than me, but my post isn't about exactly the same thing so I figured it might be worth publishing.
    (Note: The way I usually write essays is by writing outlines like this, and then fleshing them out into full posts. For a lot of the outlines I write, like this one, I never flesh them out because it doesn't seem worth the time. But I figured for Draft Amnesty Day, I could just publish my outline, and most people will get the idea.)
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - You Don’t Have to Call Yourself an Effective Altruist or Fraternize With Effective Altruists or Support Longtermism, Just Please, for the Love of God, Help the Global Poor by Omnizoid
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: You Don’t Have to Call Yourself an Effective Altruist or Fraternize With Effective Altruists or Support Longtermism, Just Please, for the Love of God, Help the Global Poor, published by Omnizoid on December 16, 2022 on The Effective Altruism Forum.
    Cross post from my blog.
    Right now you — whoever is reading this — can give money to the poorest people on earth. If you donate to givedirectly , 90% of the funds will go to the poorest people on earth. People whose incomes are roughly 1% of those of poor people in the US. And this isn’t just because things cost less — that’s taken into account for the statistic.
    This is a really surprising fact. There are people that live on less than 1% of the income of the poor in the US — and you can, right now, make their lives dramatically better at very minimal personal cost. If you donate a few hundred dollars to these people, it will double their annual income. And there’s robust evidence that this makes their lives a lot better — they don’t just spend it on trivial things.
    But givedirectly isn’t the only super effective charity. There are charities that can prevent malaria for a few thousand dollars. Givewell lists the best charities — the Malaria Consortium saves one life for every 5000 or so dollars donated.
    You can save people’s lives. You can double people’s income. All at minimal personal cost.
    The plight of the world’s poor isn’t just some interesting abstract philosophy problem. It doesn’t matter if it’s exactly like the drowning child scenario. Right now, there are people dying, and we can save their lives.
    A lot of the people that are working to save their lives are called effective altruists. But you really don’t need to call yourself an effective altruist to support this. You can think that effective altruists are annoying silicon valley tech bros. You can think they’re all satan. And it doesn’t matter.
    The badness of EA has nothing to do with our personal obligations. If EA is a rotten movement, don’t attend EA meetups. Don’t fraternize with EAs. But please, for the love of god, do something about the poor people that are dying right now, whose lives you can save.
    Émile Torres is someone that I’ve spent lots of time criticizing in the past. I think they’re really, really wrong about longtermism. But recently, I saw something that warmed my heart and improved my opinion of Torres a lot.
    Nathan @NathanpmYoung
    Christmas is a time of peace and gift giving. @xriskology and I are putting aside our differences to give to the poorest people in the world, via @GiveDirectly. Perhaps you'll join us. givingwhatwecan.org/fundraisers/em.
    10:53 AM ∙ Dec 12, 2022
    404Likes53Retweets
    Torres and Nathan Young have put aside their differences — Young is a longtermist — and organized a fundraiser to benefit the world’s poor.
    The fact that Torres is doing this shows that you don’t have to be an EA or a longtermist to support this. Perhaps the most adamant critic of EA on the face of the earth agrees that you should fund give directly. Even if EA is a horrific cult, there are poor people that are dying right now, that need money. Money given barely costs us anything, and makes their lives dramatically better.
    So please, regardless of what you think about politics, philosophy, utilitarianism, effective altruism, etc — give some money to give directly or the against malaria foundation or any other top givewell charities. You have dramatic opportunities to improve the world. These are totally uncontroversial. If anyone emails me evidence that they’ve donated 5 dollars to GiveDirectly in this fundraiser — or some other EA charity — I’ll give them a free subscription.
    This giving season, let’s put aside our differences, and do something to benefit the world’s poor. Let’s help that fundraiser raise a billion dollars for the world’s poor.
    Thanks f...
    4 min
  • AF - Paper: Constitutional AI: Harmlessness from AI Feedback (Anthropic) by Lawrence Chan
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Paper: Constitutional AI: Harmlessness from AI Feedback (Anthropic), published by Lawrence Chan on December 16, 2022 on The AI Alignment Forum.
    The authors propose a method for training a harmless AI assistant that can supervise other AIs, using only a list of rules (a "constitution") as human oversight. The method involves two phases: first, the AI improves itself by generating and revising its own outputs; second, the AI learns from preference feedback, using a model that compares different outputs and rewards the better ones. The authors show that this method can produce a non-evasive AI that can explain why it rejects harmful queries, and that can reason in a transparent way, better than standard RLHF:
    Paper abstract:
    As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through selfimprovement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as ‘Constitutional AI’. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use ‘RL from AI Feedback’ (RLAIF).
    As a result we are able to train a harmless but nonevasive AI assistant that engages with harmful queries by explaining its objections to them. Both the SL and RL methods can leverage chain-of-thought style reasoning to improve the human-judged performance and transparency of AI decision making. These methods make it possible to control AI behavior more precisely and with far fewer human labels.
    See also Anthropic's tweet thread:
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - I'm less approving of the EA community now than before the FTX collapse by throwaway790
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: I'm less approving of the EA community now than before the FTX collapse, published by throwaway790 on December 16, 2022 on The Effective Altruism Forum.
    This is mostly a counterpoint to Will Aldred/Duncan Sabien's post.
    I'm not really an EA; haven't taken the pledge, don't work at an org, have been to no EA meetups in my life, (haven't been a speaker at EA Global).
    However, I have been close to the EA community for a while, agreeing with many of its goals and donating to many of its key causes.
    Reading Duncan's post made me want to defend the case for why I'm less approving of the EA community than before. I want to be a little specific about what I'm less approving of.
    EA™ - CEA / EVF (I'd never heard it called EVF before FTX)
    Will MacAskill personally
    Donation practices during the "funding overhang" era
    There are some people/orgs I am more approving of:
    Peter Wildeford (and Rethink Priorities more generally)
    Rob Wiblin
    Dustin Moskovitz
    I'm not going to say anything about Peter/Rob/Dustin in this post, although the amount I approve them more does not change the net effect, which is less approving of EA.
    Lots of the reasons I am less approving of EA now than I was prior to FTX collapse are things I could have known. However I am aware of them because of the FTX collapse. Others might have already been aware of everything I'll mention, in which case I would agree with Duncan - with few exceptions, the communities reaction to the FTX collapse has been very good and I largely approve.
    So, what have I learned since the FTX collapse which makes me approve less of EA?
    Will MacAskill:
    Initial reactions thread I read this as both minimizing the event and distancing himself from SBF, which is not credible in light of:
    Elon relationship - you don't go and bat for someone with the richest guy in the world unless you are confident in who you're batting for
    Their long history of a close relationships (shared board memberships, his earlier mentorship etc)
    Guzey's review (this episode is why I'm posting as a throwaway)
    Equivocal statements during the early crisis:
    Will MacAskill
    Holder Karnofsky (note the edit and Alexander Berger's comments before blaming Holden)
    There were more statements I was annoyed with at the time, but I can't remember enough of the specifics to search them right now. If I remember them, I will add them here. (Part of my praise for Rob was his statement was in sharp contrast of what was coming out at the time)
    Wytham Abbey - I don't especially want to relitigate this but it makes me think less of EA - you can read many takes on EAF - probably the fairest place to start is the reasoning for why it was bought.
    People who received funding in dubious ways and didn't say anything (North Dimension donating on behalf of FTX Future fund)
    EA (generally) treating FTX as something which "happened" to them, rather than being something they were intimately involved in. (Initial funding for Alameda coming from the EA community, ...)
    I suspect (but can't prove) that the EA orgs are using the excuse of legal advice / legal threats to avoid saying things they don't want to say for other reasons. (And my suspicion is that this is because they have done something which could be bad)
    People in the community knew SBF's image was manufactured to some degree but didn't say anything (general take from forum - summarized in this New Yorker piece)
    EA's earlier relationship with a sketchy billionaire (and the degree to which this was covered up)
    All of these things put together are enough for me to downgrade my opinion of EA. I've put this together from off the top of my head, so there are likely other things which have affected my view over the last month. To be clear - I am still very much in favor of EA principles, the EA community and will continue to donate to EA causes but will ...
    4 min