The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - A review of Principia Qualia by jessicata
    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: A review of Principia Qualia, published by jessicata on July 12, 2023 on LessWrong.
    Principia Qualia, by Michael Johnson, is a long document describing an approach to theorizing about and studying consciousness. I have heard enough positive things about the document to consider it worth reviewing.
    I will split the paper (and the review) into 4 parts:
    Motivating bottom-up physicalist consciousness theories
    Reviewing Integrated Information Theory and variants
    Specifying the problem of consciousness and valence
    Symmetry
    I already disagree with part 1, but bottom-up is still a plausible approach. Part 2 makes sense to pay attention to conditional on the conclusions of part 1, and is basically a good explanation of IIT and its variants. Part 3 describes desiderata for a bottom-up theory of qualia and valence, which is overall correct. Part 4 motivates and proposes symmetry as a theory of valence, which has me typing things like "NO PLEASE STOP" in my notes document.
    Motivating bottom-up physicalist consciousness theories
    The document is primarily concerned with theories of consciousness for the purpose of defining valence: the goodness or badness of an experience, in terms of pleasure and pain. The document motivates this in terms of morality: we would like to determine things about the consciousness of different agents whether they are human, non-human animal, or AIs. The study of valence is complicated by, among other things, findings of different brain circuitry for "wanting", "liking", and "learning". Neural correlates of pleasure tend to be concentrated in a small set of brain regions and hard to activate, while neural correlates of pain tend to be more distributed and easier to activate.
    The methods of affective neuroscience are limited in studying valence and consciousness more generally. Johnson writes, "In studying consciousness we've had to rely on either crude behavioral proxies, or subjective reports of what we're experiencing." The findings of affective neuroscience do not necessarily tell us much about metaphysically interesting forms of valence and consciousness (related to Chalmers' distinction between easy problems of consciousness and the hard problem of consciousness).
    Johnson defines top-down and bottom-up theories, and motivates bottom-up theories:
    There are two basic classes of consciousness theories: top-down (aka 'higher-order' or 'cognitive' theories) and bottom-up. Top-down theories are constructed around the phenomenology of consciousness (i.e., how consciousness feels) as well as the high-level dynamics of how the brain implements what we experience...However, if we're looking for a solid foundation for any sort of crisp quantification of qualia, top-down theories will almost certainly not get us there, since we have no reason to expect that our high-level internal phenomenology has any crisp, intuitive correspondence with the underlying physics and organizational principles which give rise to it...Instead, if we're after a theory of valence/qualia/phenomenology as rigorous as a physical theory, it seems necessary to take the same bottom-up style of approach as physics does when trying to explain something like charge, spin, or electromagnetism.
    If our high-level phenomenology has no "crisp, intuitive correspondence with the underlying physics and organizational principles which give rise to it", it's unclear why a bottom-up theory would succeed either,
    To explain my take on this, I will first explain how my motivation for studying consciousness differs from Johnson's. Johnson is primarily concerned with identifying conscious agents experiencing pleasure and pain, for the purpose of optimizing the universe for more pleasure. I am primarily concerned with consciousness as a philosophical epistemic and ontological problem, which is foundational f...
    19 min
  • LW - Towards Developmental Interpretability by Jesse Hoogland
    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: Towards Developmental Interpretability, published by Jesse Hoogland on July 12, 2023 on LessWrong.
    Developmental interpretability is a research agenda that has grown out of a meeting of the Singular Learning Theory (SLT) and AI alignment communities. To mark the completion of the first SLT & AI alignment summit we have prepared this document as an outline of the key ideas.
    As the name suggests, developmental interpretability (or "devinterp") is inspired by recent progress in the field of mechanistic interpretability, specifically work on phase transitions in neural networks and their relation to internal structure. Our two main motivating examples are the work by Olsson et al. on In-context Learning and Induction Heads and the work by Elhage et al. on Toy Models of Superposition.
    Mechanistic interpretability emphasizes features and circuits as the fundamental units of analysis and usually aims at understanding a fully trained neural network. In contrast, developmental interpretability:
    is organized around phases and phase transitions as defined mathematically in SLT, and
    aims at an incremental understanding of the development of internal structure in neural networks, one phase transition at a time.
    The hope is that an understanding of phase transitions, integrated over the course of training, will provide a new way of looking at the computational and logical structure of the final trained network. We term this developmental interpretability because of the parallel with developmental biology, which aims to understand the final state of a different class of complex self-assembling systems (living organisms) by analyzing the key steps in development from an embryonic state.
    In the rest of this post, we explain why we focus on phase transitions, the relevance of SLT, and how we see developmental interpretability contributing to AI alignment.
    Thank you to @DanielFilan, @bilalchughtai, @Liam Carroll for reviewing early drafts of this document.
    Why phase transitions?
    First of all, they exist: there is a growing understanding that there are many kinds of phase transitions in deep learning. For developmental interpretability, the most important kind of phase transitions are those that occur during training. Some of the examples we are most excited about:
    Olsson, et al., "In-context Learning and Induction Heads", Transformer Circuits Thread, 2022.
    Elhage, et al., "Toy Models of Superposition", Transformer Circuits Thread, 2022.
    McGrath, et al., "Acquisition of chess knowledge in AlphaZero", PNAS, 2022.
    Michaud, et al., "The Quantization Model of Neural Scaling", 2023.
    Simon, et al., "On the Stepwise Nature of Self-Supervised Learning" ICML 2023.
    The literature on other kinds of phase transitions, such as those appearing as the scale of the model is increased, is even broader. Neel Nanda has conjectured that "phase changes are everywhere."
    Second, they are easy to find: from the point of view of statistical physics, two of the hallmarks of a (second-order) phase transition are the divergence of macroscopically observable quantities and the emergence of large-scale order. Divergences make phase transitions easy to spot, and the emergence of large-scale order (e.g., circuits) is what makes them interesting. There are several natural observables in SLT (the learning coefficient or real log canonical threshold, and singular fluctuation) which can be used to detect phase transitions, but we don't yet know how to invent finer observables of this kind, nor do we understand the mathematical nature of the emergent order.
    Third, they are good candidates for universality: every mouse is unique, but its internal organs fit together in the same way and have the same function - that's why biology is even possible as a field of science. Similarly, as an emerging field of science, interpretabi...
    15 min
  • EA - An Overview of the AI Safety Funding Situation by Stephen McAleese
    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: An Overview of the AI Safety Funding Situation, published by Stephen McAleese on July 12, 2023 on The Effective Altruism Forum.
    Introduction
    AI safety is a field concerned with preventing negative outcomes from AI systems and ensuring that AI is beneficial to humanity. The field does research on problems such as the AI alignment problem which is the problem of designing AI systems that follow user intentions and behave in a desirable and beneficial way.
    Understanding and solving AI safety problems may involve reading past research, producing new research in the form of papers or posts, running experiments with ML models, and so on. Producing research typically involves many different inputs such as research staff, compute, equipment, and office space.
    These inputs all require funding and therefore funding is a crucial input for enabling or accelerating AI safety research. Securing funding is usually a prerequisite for starting or continuing AI safety research in industry, in an academic setting, or independently.
    There are many barriers that could prevent people from working on AI safety. Funding is one of them. Even if someone is working on AI safety, a lack of funding may prevent them from continuing to work on it.
    It's not clear how hard AI safety problems like AI alignment are. But in any case, humanity is more likely to solve them if there are hundreds or thousands of brilliant minds working on them rather than one guy. I would like there to be a large and thriving community of people working on AI safety and I think funding is an important prerequisite for enabling that.
    The goal of this post is to give the reader a better understanding of funding opportunities in AI safety so that hopefully funding will be less of a barrier if they want to work on AI safety. The post starts with a high-level overview of the AI safety funding situation followed by a more in-depth description of various funding opportunities.
    Past work
    To get an overview of AI safety spending, we first need to find out how much is spent on it per year. We can use past work as a prior and then use grant data to find a more accurate estimate.
    Changes in funding in the AI safety field (2017) by the Center for Effective Altruism estimated the change in AI safety funding between 2014 - 2017. In 2017, the post estimated that total AI safety spending was about $9 million.
    How are resources in effective altruism allocated across issues? (2020) by 80,000 Hours estimated the amount of money spent by EA on AI safety in 2019. Using data from the Open Philanthropy grants database, the post says that EA spent about $40 million on AI safety globally in 2019.
    In The Precipice (2020), Toby Ord estimated that between $10 million and $50 million was spent on reducing AI risk in 2020.
    2021 AI Alignment Literature Review and Charity Comparison is an in-depth review of AI safety organizations and grantmakers and has a lot of relevant information.
    Overview of global AI safety funding
    One way to estimate total global spending on AI safety is to aggregate the total donations of major AI safety funds such as Open Philanthropy (Open Phil).
    It's important to note that the definition of 'AI safety' I'm using is AI safety research that is focused on reducing risks from advanced AI (AGI) such as existential risks which is the type of AI safety research I think is more neglected and important than other research in the long term. Therefore my analysis will be focused on EA funds and top AI labs and I don't intend to measure investment on near-term AI safety concerns such as effects on the labor market, fairness, privacy, ethics, disinformation, etc.
    The results of this analysis are shown in the following bar chart which was created in Google Sheets (link) and is based on data from analyzing grant databases from Open Philanthro...
    26 min
  • AF - What does the launch of x.ai mean for AI Safety? by Chris Leong
    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: What does the launch of x.ai mean for AI Safety?, published by Chris Leong on July 12, 2023 on The AI Alignment Forum.
    x.ai has now launched. It seems worthwhile to discuss both what it means for AI safety and whether people interested in AI safety should consider applying for the company.
    Some thoughts:
    It's notable that Dan Hendrycks is listed as an advisor (the only advisor listed).
    The team is also listed on the page.I haven't taken the time to do so, but it might be informative for someone to Google the individuals listed to discover whether they lie in terms of their interest being between capabilities and safety.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min
  • AF - Towards Developmental Interpretability by Jesse Hoogland
    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: Towards Developmental Interpretability, published by Jesse Hoogland on July 12, 2023 on The AI Alignment Forum.
    Developmental interpretability is a research agenda that has grown out of a meeting of the Singular Learning Theory (SLT) and AI alignment communities. To mark the completion of the first SLT & AI alignment summit we have prepared this document as an outline of the key ideas.
    As the name suggests, developmental interpretability (or "devinterp") is inspired by recent progress in the field of mechanistic interpretability, specifically work on phase transitions in neural networks and their relation to internal structure. Our two main motivating examples are the work by Olsson et al. on In-context Learning and Induction Heads and the work by Elhage et al. on Toy Models of Superposition.
    Mechanistic interpretability emphasizes features and circuits as the fundamental units of analysis and usually aims at understanding a fully trained neural network. In contrast, developmental interpretability:
    is organized around phases and phase transitions as defined mathematically in SLT, and
    aims at an incremental understanding of the development of internal structure in neural networks, one phase transition at a time.
    The hope is that an understanding of phase transitions, integrated over the course of training, will provide a new way of looking at the computational and logical structure of the final trained network. We term this developmental interpretability because of the parallel with developmental biology, which aims to understand the final state of a different class of complex self-assembling systems (living organisms) by analyzing the key steps in development from an embryonic state.
    In the rest of this post, we explain why we focus on phase transitions, the relevance of SLT, and how we see developmental interpretability contributing to AI alignment.
    Thank you to @DanielFilan, @bilalchughtai, @Liam Carroll for reviewing early drafts of this document.
    Why phase transitions?
    First of all, they exist: there is a growing understanding that there are many kinds of phase transitions in deep learning. For developmental interpretability, the most important kind of phase transitions are those that occur during training. Some of the examples we are most excited about:
    Olsson, et al., "In-context Learning and Induction Heads", Transformer Circuits Thread, 2022.
    Elhage, et al., "Toy Models of Superposition", Transformer Circuits Thread, 2022.
    McGrath, et al., "Acquisition of chess knowledge in AlphaZero", PNAS, 2022.
    Michaud, et al., "The Quantization Model of Neural Scaling", 2023.
    Simon, et al., "On the Stepwise Nature of Self-Supervised Learning" ICML 2023.
    The literature on other kinds of phase transitions, such as those appearing as the scale of the model is increased, is even broader. Neel Nanda has conjectured that "phase changes are everywhere."
    Second, they are easy to find: from the point of view of statistical physics, two of the hallmarks of a (second-order) phase transition are the divergence of macroscopically observable quantities and the emergence of large-scale order. Divergences make phase transitions easy to spot, and the emergence of large-scale order (e.g., circuits) is what makes them interesting. There are several natural observables in SLT (the learning coefficient or real log canonical threshold, and singular fluctuation) which can be used to detect phase transitions, but we don't yet know how to invent finer observables of this kind, nor do we understand the mathematical nature of the emergent order.
    Third, they are good candidates for universality: every mouse is unique, but its internal organs fit together in the same way and have the same function - that's why biology is even possible as a field of science. Similarly, as an emerging field of science,...
    16 min
  • LW - A transcript of the TED talk by Eliezer Yudkowsky by Mikhail Samin
    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: A transcript of the TED talk by Eliezer Yudkowsky, published by Mikhail Samin on July 12, 2023 on LessWrong.
    The TED talk is available on YouTube and the TED website. Previously, a live recording was published behind the paywall on the conference website and later (likely accidentally) on a random TEDx YouTube channel but was later removed.
    The transcription is done with Whisper.
    You've heard that things are moving fast in artificial intelligence. How fast? So fast that I was suddenly told on Friday that I needed to be here.So, no slides, six minutes.
    Since 2001, I've been working on what we would now call the problem of aligning artificial general intelligence: how to shape the preferences and behavior of a powerful artificial mind such that it does not kill everyone.
    I more or less founded the field two decades ago when nobody else considered it rewarding enough to work on. I tried to get this very important project started early so we'd be in less of a drastic rush later.
    I consider myself to have failed.
    Nobody understands how modern AI systems do what they do. They are giant, inscrutable matrices of floating-point numbers that we nudge in the direction of better performance until they inexplicably start working.
    At some point, the companies rushing headlong to scale AI will cough out something that's smarter than humanity.
    Nobody knows how to calculate when that will happen. My wild guess is that it will happen after zero to two more breakthroughs the size of transformers.
    What happens if we build something smarter than us that we understand that poorly?
    Some people find it obvious that building something smarter than us that we don't understand might go badly. Others come in with a very wide range of hopeful thoughts about how it might possibly go well. Even if I had 20 minutes for this talk and months to prepare it, I would not be able to refute all the ways people find to imagine that things might go well.
    But I will say that there is no standard scientific consensus for how things will go well. There is no hope that has been widely persuasive and stood up to skeptical examination. There is nothing resembling a real engineering plan for us surviving that I could critique.This is not a good place in which to find ourselves.
    If I had more time, I'd try to tell you about the predictable reasons why the current paradigm will not work to build a superintelligence that likes you or is friends with you, or that just follows orders.
    Why, if you press thumbs up when humans think that things went right or thumbs down when another AI system thinks that they went wrong, you do not get a mind that wants nice things in a way that generalizes well outside the training distribution to where the AI is smarter than the trainers.
    You can search for Yudkowsky, List of Lethalities for more.
    But to worry, you do not need to believe me about exact predictions of exact disasters. You just need to expect that things are not going to work great on the first really serious, really critical try because an AI system smart enough to be truly dangerous was meaningfully different from AI systems stupider than that.
    My prediction is that this ends up with us facing down something smarter than us that does not want what we want, that does not want anything we recognize as valuable or meaningful.
    I cannot predict exactly how a conflict between humanity and a smarter AI would go for the same reason I can't predict exactly how you would lose a chess game to one of the current top AI chess programs, let's say, Stockfish.
    If I could predict exactly where Stockfish could move, I could play chess that well myself. I can't predict exactly how you'll lose to Stockfish, but I can predict who wins the game.
    I do not expect something actually smart to attack us with marching robot armies with glowing red...
    6 min
  • AF - Goal-Direction for Simulated Agents by Raymond D
    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: Goal-Direction for Simulated Agents, published by Raymond D on July 12, 2023 on The AI Alignment Forum.
    tldr: consistent LLM failure suggests possible avenue for alignment and control
    epistemic status: somewhat hastily written, speculative in places, but lots of graphs of actual model probabilities
    Today in 'surprisingly simple tasks that even the most powerful large language models can't do': writing out the alphabet but skipping over one of the letters.
    This is GPT-4, and it seems to manage about half the time with the ninth letter. You can do the same thing with numbers - the general principle is 'write out a very common pattern, with a slight deviation.
    We can probe the phenomenon more carefully in text-davinci-003, which lets us easily track the probabilities:
    So for the sixth letter, it's almost getting it - it assigns 21.74% to the correct answer. If we plot the graph for how probable the correct answer (skip) is depending on which letter we ask it to omit, it looks like this:
    Some Background: How To Misquery Predictors
    Let's imagine you had a truly perfect oracle-like predictor, trained on 1000 agents that have been given the task of turning $1,000 into $10,000.
    5 of them make sound financial choices, carefully investing their money and exploiting a subtle but consistent error in the market to all make $10,000
    The other 995 blow all their money on lottery tickets, causing 950 to go broke and 45 to luck into $10,000
    If you ask your perfect predictor what action maximises the chance of making $10,000, it will indeed reply with the careful investment strategy. But if you simply ask it what you'd expect to see from an agent that makes $10,000, it will say 'lots of lottery tickets', because 90% of the agents that made $10,000 did so with lottery tickets.
    Janus gives the following definition of simulators:
    I use the generic term "simulator" to refer to models trained with predictive loss on a self-supervised dataset, invariant to architecture or data type (natural language, code, pixels, game states, etc). The outer objective of self-supervised learning is Bayes-optimal conditional inference over the prior of the training distribution, which I call the simulation objective, because a conditional model can be used to simulate rollouts which probabilistically obey its learned distribution by iteratively sampling from its posterior (predictions) and updating the condition (prompt)
    I claim that when you simulate an agent conditioned on some goal or target, it falls into exactly the trap above, because of the way it is queried. Moreover, I claim that it should be possible to elicit the model's latent true understanding of how to write the alphabet while skipping a letter, not by adding anything extra to the prompt or any interpretability but through a consideration of its causal structure. Instead of asking "supposing someone makes money, what did they do?" we must ask "what might someone do to most increase their chance of making money?". The challenge is how to do the equivalent of that for a language model writing the alphabet.
    Predictive Completion vs Goal-Directed Completion
    If you are the kind of person who likes to pause in the middle of essays to work things out for yourself, now would be the moment: how do you correct the distribution of 'what actions are agents likely to take given that they get this reward' to get the distribution of 'what actions make an agent likely to get this reward', in general, without needing to create a separate query for each action on which to condition?
    Formally, we're querying P(a|r), the probability of action given reward (I'll call this the 'predictive completion'), when we should be querying P(r|a) (I'll call this the 'goal-directed completion'). Per Bayes' Theorem, the latter distribution is the former multiplied by P(r)P(a), ...
    10 min
  • EA - An expert survey on social movements and protest by James Özden
    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: An expert survey on social movements and protest, published by James Özden on July 12, 2023 on The Effective Altruism Forum.
    Summary
    We're very excited to release some new research that Social Change Lab has been working on for a while. We teamed up with Apollo Academic Surveys to create an expert survey of academics who study social movements and protest. We surveyed 120 academics across Political Science, Sociology and other relevant disciplines, and picked many academics because we thought they had made significant contributions to the understanding of social movements and protest. We hope this survey provides some strategic insight that is useful for EAs working on a range of important issues.
    You can see the full results on the Apollo Academics Surveys website, but we wanted to share a few findings we found really interesting. We'll also be releasing our own report (with greater analysis, interpretation and limitations) soon. However, in this short post we'll cover some of the expert views on the following topics:
    Which are the most important strategic and organisational factors that lead to social movements succeeding
    The most common reasons social movements fail to achieve their goals
    The effectiveness of disruptive protest based on what kind of issue you're working on (e.g. high public support vs low public support)
    The effectiveness of disruptive protest within animal advocacy
    Some things we think are interesting from the full results but we excluded for brevity:
    The extent to which a social movement's success is related to factors within their control (e.g. tactics and strategy) vs factors outside their control (e.g. wider political context)
    To what degree polarisation is inevitable or necessarily a bad outcome
    What intermediate goals are important to focus on if you care about ultimately passing government policy.
    Additionally, you can click "See Additional Participant responses" to read what experts wrote to give additional context to their quantitative responses.
    Results
    Important factors for social movement success
    Out of the factors we asked about, experts thought the most important tactical and strategic factor for a social movement's success is "the strategic use of nonviolent disruptive tactics". Overall, 69% of experts said this factor was either "very important" or "quite important".
    Figure 1: Answers to Question 3 from the survey: "How important do you think the following tactical and strategic factors are in contributing to a social movement's success?
    We were really struck by the contradiction between what the public (78% of people think that disruptive protests hinder the cause) and the media say about disruptive protests and what academics said. The experts who study social movements not only believe that strategic disruption can be an effective tactic, but that it is the most important tactical factor for a social movement's success (of the factors we asked about). This, and other relevant evidence, suggests we shouldn't necessarily take people's first reactions as the best indicator of effective protest. We include many more results on disruptive protest further down.
    Experts were also asked which factors of a social movement's governance are the most important in driving success. The factor they rated most highly was 'the ability to mobilise and scale quickly in response to external events'. The factor they rated least important, possibly a surprise to activists themselves, was 'decentralised decision-making'
    Figure 2: Answers to Question 4 from the survey: "How important do you think the following governance and organisational factors are in contributing to a social movement's success?"
    Common reasons social movements fail
    The most commonly cited reason for social movements to fail was internal conflict and movement infighting, closely fo...
    11 min
  • EA - Free Social Anxiety Treatment for 100 EAs (Sign-up here) by John Salter
    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: Free Social Anxiety Treatment for 100 EAs (Sign-up here), published by John Salter on July 12, 2023 on The Effective Altruism Forum.
    Arguably the most important factor in being happy and impactful is one's social connections. Anxiety makes them harder to form and maintain. Graded exposure is the gold-standard treatment. We're offering it for free to all EAs. Here's how it works:
    Arrange a chat with one of our head coaches. They'll help you figure out if this is right for you and, if so, pair you up with a suitable coach.
    Meet with your coach (can be via email/text initially). Explain your situation.
    Construct a hierarchy of fear (see below for an example)
    We'll help you work through the tasks, teaching you relevant skills along the way.
    By approaching rather than avoiding anxiety-provoking tasks, you break the loop that perpetuates social anxiety.
    Meta-analyses suggest you can get an effect size of ~1.0 with this approach. It takes two to three months of weekly sessions, each around 45 minutes long.Click here to sign up
    Questions You're Too Socially Anxious to Ask
    Why is it free?
    We're a charity, and our coaches are either donor-funded or volunteers
    What evidence is there that you can execute?
    We've seen 100+ EAs over the past 9 months or so with a wide range of different issues. Our start-to-end retention is ~70% (50% is considered good), and feedback has been unanimously positive.
    I don't want other EAs to find out I do this. How can you ensure anonymity?
    I'm the only EA who'll see that you've signed up. Our coaches aren't themselves effective altruists. Only your coach will have access to notes about your session. You're free to use a fake name and alias email if you'd like..
    What if I don't like my coach?
    Email me and I'll assign you a new one (john at overcome dot org dot uk)
    What if I'm done early or I need more sessions?
    If early, great! You won't be pressured to keep attending.
    If you need more, we'll consider extending it for free. Otherwise, it'd be $40 per session. Historically, 90% of people have found the free sessions adequate to address the problem they came to us with.
    I'm too anxious to talk via video. Are there any other options?
    Yes. We can also do email, texting, or voice-only.
    We'd expect you to progress from email/texting towards being comfortable with video calls as the program goes on.
    What if I sign up and someone who needs it more is thus denied a place?
    It's unlikely. We prioritise based on need & likelihood of success.
    Don't you usually focus on LMICs? Why are you offering services for EAs?
    It costs me nothing to use my coaches' spare capacity on services for EAs. So long as I have spare capacity, I'll offer it to EAs.
    Click here to sign up
    Any other questions, post them in the comments below!
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    3 min
  • LW - My Weirdest Experience by Bridgett Kay
    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: My Weirdest Experience, published by Bridgett Kay on July 11, 2023 on LessWrong.
    I wrote about this a long time ago in some comments here- I'm not sure where or when- but there seemed to be a lot of interest in it. Because of this, I thought I would share the post I've written about it here.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min