The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - Efficiency and resource use scaling parity by Ege Erdil
    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: Efficiency and resource use scaling parity, published by Ege Erdil on August 21, 2023 on LessWrong.
    An interesting pattern I've now noticed across many different domains is that if we try to do an attribution of improvements in outcomes or performance in a domain to the two categories "we're using more resources now than in the past" and "we're making more efficient use of resources now than in the past", there is usually an even split in how much improvement can be attributed to each category.
    Some examples:
    In computer vision, Erdil and Besiroglu (2022) (my own paper) estimates that 40% of performance improvements in computer vision from 2012 to 2022 have been due to better algortihms, and 60% due to the scaling of compute and data.
    In computer chess, a similar pattern seems to hold: roughly half of the progress in chess engine performance from Deep Blue to 2015 has been from the scaling of compute, and half from better algorithms. Stockfish 8 running on consumer hardware in 1997 could achieve an Elo rating of ~ 3000, compared to ~ 2500 for contemporary hardware; and Stockfish 8 on 2015 hardware could go up to ~ 3400.
    In rapidly growing economies, accounting for growth in output per worker by dividing it into capital per worker (resource scaling) and TFP (efficiency scaling, roughly speaking) often gives an even split: see Bosworth and Collins (2008) for data on China and India specifically.
    More pessimistic estimates of the growth performance of China compared to official data put this split at 75% to 25% (see this post for details) but the two effects are still at least comparable.
    A toy model
    A speculative explanation is the following: if we imagine that performance in some domain is measured by a multiplicative index P which can be decomposed as the product of individual contributing factors F1,F2,.,Fn so that P∝∏ni=1Fi, in general we'll have
    gP=1PdPdt=n∑i=11FidFidt=n∑i=1gFi
    thanks to the product rule. Note that gX denotes the growth rate of the variable X.
    I now want to use a law of motion from Jones (1995) for Fi: we assume they evolve over time according to
    gFi=1FidFidt=θiF-βiiIλii
    where θi,βi,λi>0 are parameters and Ii is a measure of "investment input" into factor i. This general specification can capture diminishing returns on investment as we make progress or scale up resources thanks to β, and can capture returns to scale to spending more resources on investment at a given time thanks to λ.
    Substituting this into the growth expression for P gives
    gP=1PdPdt=n∑i=1θiF-βiiIλii
    Now, suppose we have a fixed budget I at any given time to allocate across all investments Ii, and our total budget I grows over time at a rate g. To maximize the rate of progress at a given time, the marginal returns to investment across all factors should be equal, i.e. we should have
    ∂∂Ii(1FidFidt)=∂∂Ij(1FjdFjdt)
    for all pairs i,j. Substitution gives
    θiλiF-βiiIλi-1i=θjλjF-βjjIλj-1j
    and upon simplification, we recover
    λigFiIi=λjgFiIj
    In an equilibrium where all quantities grow exponentially, the ratios Ii/Ij must therefore remain constant, i.e. all of the Ii must also grow at the aggregate rate of input growth g. Then, it's easy to see that the Jones law of motion implies gFi=gλi/βi for each factor i, from which we get the important conclusion
    gFi∝λiβi=ri
    that must hold in an exponential growth equilibrium. The parameter ri is often called the returns to investment, so this relation says that distinct factors account for growth in P proportional to their returns to investment parameter.
    How do we interpret the data in light of the toy model?
    If we simplify the setup and make it about two factors, one measuring resource use and the other measuring efficiency, then the fact that the two factors account for comparable fractions in overall progress should mean that their associated retu...
    7 min
  • AF - Causality and a Cost Semantics for Neural Networks by scottviteri
    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: Causality and a Cost Semantics for Neural Networks, published by scottviteri on August 21, 2023 on The AI Alignment Forum.
    Epistemic status: I time-boxed this idea to three days of effort. So any calculations are pretty sloppy, and I haven't looked into any related works. I probably could have done much better if I knew anything about circuit complexity. There are some TODOs and an unfinished last section -- if you are interested in this content and want to pick up where I have left off I'll gladly add you as a collaborator to this post.
    Here is a "tech tree" for neural networks. I conjecture (based on admittedly few experiments) that the simplest implementation of any node in this tree includes an implementation of its parents, given that we are writing programs starting from the primitives +, , and relu. An especially surprising relationship (to me) is that "if statements" are best implemented downstream of division.
    Introduction
    While discussing with my friend Anthony Corso, an intriguing idea arose. Maybe we can define whether program p1 "causes" p2 in the following way: Given a neural network that mimics p1, how easy is it to learn a neural network which mimics the behavior of p2? This proposition is intriguing because it frames causality as a question about two arbitrary programs, and reduces it to a problem of program complexity.
    Suppose that p1 and p2 are written in a programming language P, and let P(ops) represent P extended with ops as primitive operations. We define a complexity function C:P(ops)R, which takes a program in the extended language and returns a real number representative of the program's complexity for some fixed notion of complexity. Let's define the degree to which p1 "causes" p2 as the minimum complexity achievable by a program p from P(p1) such that p is extensionally equal (equal for all inputs) to p2. If P2 is the set of all p in P(obs+p1) that are extensionally equal to p2, then causes(p1,p2)=minp∈P2C(p). We can also use this definition in the approximate case, considering the minimum complexity achievable by programs p such that E(p(x)-p2(x))2<ε with respect to some L1-integrable probability measure.
    We can define a particular complexity function C that represents the cost of executing a program. We can estimate this quantity by looking at the program's Abstract Syntax Tree (AST) in relation to some cost model of the primitive operations in the language. For this exploration, we have chosen the lambda calculus as the language. Lambda calculus is a minimalist Lisp-like language with just a single type, which in our case we will think of as floating point numbers. The notation is simple: lambda abstraction is represented as λ x. x, and function application as (f g), which is not the same as f(g) in most other languages.
    How I Would Like People to Engage with this Work
    By writing Ops in your favorite programming language
    By circumventing my proposed tech tree, by reaching a child without reaching a parent and using fewer (or equal) number of operations
    By training some neural networks between these programs, and seeing how difficult it is to learn one program after pre-training on another
    Cost Semantics
    Definition
    We define the cost of operations and expressions in the following manner:
    Ops op=1,for any operation op in opsOps c=0,for any floating-point constant cOps x=0,for any variable xOps (λx.e)=Ops eOps (f g)=Ops f+Ops g
    For operations of higher arity, we have({Ops }({op }x1.xn))=({Ops }{op})+∑i({Ops }xi)
    The selected operations for a neural network are ops = {+, , relu}.
    Basic Operations and Warm-Up
    Let's take a few examples to demonstrate this cost calculus:
    To derive subtraction, we first create negation neg.
    (Ops neg) = (Ops (λ x. ( -1 x))) = (Ops ( -1 x))= (Ops ) + (Ops -1) + (Ops x) = 1 + 0 + 0 = 1
    The cost of subtraction (-) ...
    17 min
  • EA - An Elephant in the Community Building room by Kaleem
    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 Elephant in the Community Building room, published by Kaleem on August 21, 2023 on The Effective Altruism Forum.
    These are my own views, and not those of my employer, EVOps, or of CEA who I have contracted for in the past and am currently contracting for now. This was meant to be a strategy fortnight contribution, but it's now a super delayed/unofficial, and underwritten strategy fortnight contribution.
    Before you read this:
    This is pretty emotionally raw, so please 1) don't update too much on it if you think I'm just being dramatic 2) I might come back and endorse or delete this at some point. I've put off writing this for a long time, because I know that some of the conclusions or implications might be hurtful or cause me to become even more unpopular than I already feel I am - as a result, I've left it really brief, but I'm willing to make it more through if I get the sense that people think it'd be valuable.
    This post is not meant as a disparagement of any of my fellow African or Asian or Latin-American EAs. This is less about you, and more about how much the world sucks, and how hard the state of the world makes it for us to fully participate in, and contribute to, EA the way we'd like to. I think I'm hoping to read a bunch of comments proving me wrong or at least making me reconsider how I feel about this. That being said, I don't like letting feelings get in the way of truth seeking and doing what's right. So here it goes.
    Summary:
    I think community builders and those funding/steering community building efforts should be more explicit and open about what their theory of change for global community building is (especially in light of the reduced amount of funding available), as there could be significant tradeoffs in impact between different strategies.
    Introduction
    I think there are two broad conceptualisations of what/how EA functions in the world, and each has a corresponding community building strategy. If you think there are more than these two, or that these are wrong or could be improved, please let me know. From my experience, I think that all community building initiatives fall into one of two strategies/worldviews, each with a different theory of change. These are:
    Global EA
    EA can be for anybody in the world - The goal of EA community building is to spread the ideas of EA as far and wide as possible. By showing people that regardless of your context, you can make a difference which is possibly hundreds of times better than you would have done otherwise, we'll be increasing the chances of motivated and talented people getting involved in high-impact work, and generally increasing the counterfactual positive impact of humanity on the wellbeing of living and future beings. I have a sense that following this strategy currently leads to having a more transparent/non-secretive/less insidious optic for the movement.
    Efforts which fall into this bucket would be things like:
    funding city and national groups in countries which aren't major power-centers in the US, UK, EU, or China
    funding university groups which aren't in the top 100/200 in the world for subjects which have a track record of being well-represented amongst global decision-makers.
    Allocating community resources to increasing blindly-racial or geographic diversity and inclusion in the community (rather than specific viewpoints or underrepresented moral beliefs etc).
    Narrow EA
    Power and influence follow a heavy-tailed distribution, and we need power and influence to make important changes. If there is a small group of people who are extremely influential or high-potential, then the goal of community building should be to seek out and try to convince them to use their resources to have an outsized positive influence on the wellbeing of current and future beings. I have a sense that the way that ...
    10 min
  • EA - XPT forecasts on (some) Direct Approach model inputs by Forecasting Research Institute
    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: XPT forecasts on (some) Direct Approach model inputs, published by Forecasting Research Institute on August 21, 2023 on The Effective Altruism Forum.
    This post was co-authored by the Forecasting Research Institute and Rose Hadshar. Thanks to Josh Rosenberg for managing this work, Zachary Jacobs and Molly Hickman for the underlying data analysis, Kayla Gamin for fact-checking and copy-editing, and the whole FRI XPT team for all their work on this project. Special thanks to staff at Epoch for their feedback and advice.
    Summary
    Superforecaster and expert forecasts from the Existential Risk Persuasion Tournament (XPT) differ substantially from Epoch's default Direct Approach model inputs on algorithmic progress and investment:
    InputEpoch (default)XPT superforecasterXPT expertNotesBaseline growth rate in algorithmic progress (OOM/year)0.21-0.650.09-0.20.15-0.23Current spending ($, millions)$60$35$60Yearly growth in spending (%)34%-91.4%6.40%-11%5.7%-19.5%
    Epoch: 80% confidence interval (CI)
    XPT: 90% CI, based on 2024-2030 forecasts
    Epoch: 2023 estimate
    XPT: 2024 median forecast
    Epoch: 80% CI
    XPT: 90% CI, based on 2024-2050 forecasts
    Note that there are no XPT forecasts relating to other inputs to the Direct Approach model, most notably the compute requirements parameters.
    Taking the Direct Approach model as given and using relevant XPT forecasts as inputs where possible leads to substantial differences in model output:
    OutputEpoch default inputsXPT superforecaster inputsXPT expert inputsMedian TAI arrival yearProbability of TAI by 2050Probability of TAI by 2070Probability of TAI by 2100
    2036
    2065
    2052
    70%
    38%
    49%
    76%
    53%
    65%
    80%
    66%
    74%
    Note that regeneration affects model outputs, so these results can't be replicated directly, and the TAI probabilities presented here differ slightly from those in Epoch's blog post. Figures given here are the average of 5 regenerations.
    Epoch is drawing on recent research which was not available at the time the XPT forecasters made their forecasts (the XPT closed in October 2022).
    Most of the difference in outputs comes down to differences in forecasts on baseline growth rate in algorithmic progress and yearly growth in spending, where XPT forecasts differ radically from the Epoch default inputs (which extrapolate historical trends).
    XPT forecasters' all-things-considered transformative artificial intelligence (TAI) timelines are much longer than those which the Direct Approach model outputs using XPT inputs:
    Source of 2070 forecastXPT superforecasterXPT expertDirect Approach model53%65%XPT postmortem survey question on probability of TAI by 20703.75%16%
    If you buy the assumptions of the Direct Approach model, and XPT forecasts on relevant inputs, this pushes timelines out by two to three decades compared with the default Epoch inputs.
    However, it still implies TAI by 2070.
    It seems very likely that XPT forecasters would not buy the assumptions of the Direct Approach model: their explicitly stated probabilities on TAI by 2070 are <20%.
    Introduction
    This post:
    Compares Direct Approach inputs with XPT forecasts on algorithmic progress and investment, and shows how the differences in forecasts impact the outputs of the Direct Approach model.
    Discusses why Epoch's inputs and XPT forecasts differ.
    Notes that XPT forecasters' all-things-considered TAI timelines are longer than those which the Direct Approach model outputs using XPT inputs.
    Includes an appendix on the arguments given by Epoch and in the XPT for their respective forecasts.
    Background on the Direct Approach model
    In May 2023, researchers at Epoch released an interactive Direct Approach model, which models the probability that TAI arrives in a given year. The model relies on:
    An estimate of the compute required for TAI, based on extrapolating neural scaling laws.
    Various inputs rel...
    34 min
  • EA - "Dimensions of Pain" workshop: Summary and updated conclusions by Rethink Priorities
    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: "Dimensions of Pain" workshop: Summary and updated conclusions, published by Rethink Priorities on August 21, 2023 on The Effective Altruism Forum.
    Executive Summary
    Background: The workshop's goal was to leverage expertise in pain to identify strategies for testing whether severity or duration looms larger in the overall badness of negatively valenced experiences. The discussion was focused on how to compare welfare threats to farmed animals.
    No gold standard behavioral measures: Although attendees did not express confidence in any single paradigm, several felt that triangulating results across several paradigms would increase clarity about whether nonhuman animals are more averse to severe pains or long-lasting pains.
    Consistent results across different methodologies only strengthens a conclusion if they have uncorrelated or opposing biases. Fortunately, while classical conditioning approaches are probably biased towards severity mattering more, operant conditioning approaches are probably biased towards duration mattering more. Unfortunately, the biases might be too large to produce convergent results.
    Behavioral experiments may lack external validity: Attendees believed that a realistic experiment would not involve pains of the magnitude that characterize the worst problems farmed animals endure. Thus, instead of prioritizing external validity, we recommend whatever study designs create the largest differences in severity.
    Studies of laboratory animals and (especially) humans seem more likely to generate large differences in severity than studies of farmed animals.
    No gold standard biomarkers: Biomarkers could elide the biases that behavioral and self-report data inevitably introduce. However, attendees argued that there are no currently known biomarkers that could serve as an aggregate measure of pain experience over the course of a lifetime.
    Priors should favor prioritizing duration: Attendees had competing ideas about how to prioritize between severity and duration in the absence of compelling empirical evidence. In cases where long-lasting harms are at least thousands of times longer than more severe harms and are of at least moderate severity, we favor a presumption that long-lasting pains cause more disutility overall.
    Nevertheless, due to empirical and moral uncertainty, we would recommend putting some credence (~20%) in the most severe harms causing farmed animals at least as much disutility as the longest-lasting harms they experience.
    Background
    The Dimensions of Pain workshop was held April 27-28, 2023 at University of British Columbia. Attendees included animal welfare scientists (viz., Dan Weary, Thomas Ede, Leonie Jacobs, Ben Lecorps, Cynthia Schuck, Wladimir Alonso, and Michelle Lavery), pain scientists (Jeff Mogil, Gregory Corder, Fiona Moultrie, Brent Vogt), and philosophers (Bob Fischer, Murat Aydede, Walter Veit). William McAuliffe and Adam Shriver, the authors of this report, guided the discussion.
    Funders who want to cost-effectively improve animal welfare have to decide whether attenuating brief, severe pains (e.g., live-shackle slaughter) or chronic, milder pains (e.g., lameness) reduces more suffering overall. Farmers also face similar tradeoffs when deciding between multiple methods for achieving the same goal (e.g., single-stage versus multi-stage stunning). Our original report exploring the considerations that would favor prioritizing one dimension over another, The Relative Importance of the Severity and Duration of Pain, identified barriers to designing experiments that would provide clear-cut empirical evidence. The goal of the workshop was to ascertain whether an interdisciplinary group of experts could overcome these issues.
    No gold standard behavioral measures
    We spent one portion of the workshop reviewing some of the confounds th...
    25 min
  • LW - Ruining an expected-log-money maximizer by philh
    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: Ruining an expected-log-money maximizer, published by philh on August 21, 2023 on LessWrong.
    Suppose you have a game where you can bet any amount of money. You have a 60% chance of doubling your stake and a 40% chance of losing it.
    Consider agents Linda and Logan, and assume they both have £11. Linda has a utility function that's linear in money (and has no other terms), ULinda(m)=m. She'll bet all her money on this game. If she wins, she'll bet it again. And again, until eventually she loses and has no more money.
    Logan has a utility function that's logarithmic in money, ULogan(m)=ln(m). He'll bet 20% of his bankroll every time, and his wealth will grow exponentially.
    Some people take this as a reason to be Logan, not Linda. Why have a utility function that causes you to make bets that leave you eventually destitute, instead of a utility function that causes you to make bets that leave you rich?
    In defense of Linda
    I make three replies to this. Firstly, the utility function is not up for grabs! You should be very suspicious any time someone suggests changing how much you value something.
    "Because if Linda had Logan's utility function, she'd be richer. She'd be doing better according to her current utility function." My second reply is that this is confused. Before the game begins, pick a time t. Ask Linda which distribution over wealth-at-time-t she'd prefer: the one she gets from playing her strategy, or Logan's strategy? She'll answer, hers: it has an expected wealth of £1.2t. Logan's only has an expected wealth of £1.04t.
    And, at some future time, after she's gone bankrupt, ask Linda if she thinks any of her past decisions were mistakes, given what she knew at the time. She'll say no: she took the bet that maximized her expected wealth at every step, and one of them went against her, but that's life. Just think of how much money she'd have right now if it hadn't! (And nor had the next one, or the one after..) It was worth the risk.
    You might ask "but what happens after the game finishes? With probability 1, Linda has no money, and Logan has infinite". But there is no after! Logan's never going to stop. You could consider various limits as t∞, but limits aren't always well-behaved2. And if you impose some stopping behavior on the game - a fixed or probabilistic round limit - then you'll find that Linda's strategy just uncontroversially gives her better payoffs (according to Linda) after the game than Logan's, when her probability of being bankrupt is only extremely close to 1.
    Or, "but at some point Logan is going to be richer than Linda ever was! With probability 1, Logan will surpass Linda according to Linda's values." Yes, but you're comparing Logan's wealth at some point in time to Linda's wealth at some earlier point in time. And when Logan's wealth does surpass the amount she had when she lost it all, she can console herself with the knowledge that if she hadn't lost it all, she'd be raking it in right now. She's okay with that.
    I suppose one thing you could do here is pretend you can fit infinite rounds of the game into a finite time. Then Linda has a choice to make: she can either maximize expected wealth at tn for all finite n, or she can maximize expected wealth at tω, the timestep immediately after all finite timesteps. We can wave our hands a lot and say that making her own bets would do the former and making Logan's bets would do the latter, though I don't endorse the way we're treating infinties here.
    Even then, I think what we're saying is that Linda is underspecified. Suppose she's offered a loan, "I'll give you £1 now and you give me £2 in a week". Will she accept? I can imagine a Linda who'd accept and a Linda who'd reject, both of whom would still be expected-money maximizers, just taking the expectation at different times and/or expanding "mone...
    14 min
  • LW - Chess as a case study in hidden capabilities in ChatGPT by AdamYedidia
    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: Chess as a case study in hidden capabilities in ChatGPT, published by AdamYedidia on August 21, 2023 on LessWrong.
    There are lots of funny videos of ChatGPT playing chess, and all of them have the same premise: ChatGPT doesn't know how to play chess, but it will cheerfully and confidently make lots of illegal moves, and humoring its blundering attempts to play a game it apparently doesn't understand is great content.
    What's less well-known is that ChatGPT actually can play chess when correctly prompted. It plays at around 1000 Elo, and can make consistently legal moves until about 20-30 moves in, when its performance tends to break down. That sounds not-so-impressive, until you consider that it's effectively playing blindfolded, having access to only the game's moves in algebraic notation, and not a visual of a chessboard. I myself have probably spent at least a thousand hours playing chess, and I think I could do slightly better than 1000 Elo for 30 moves when blindfolded, but not by much. ChatGPT's performance is roughly the level of blindfolded chess ability to expect from a decent club player. And 30 moves is more than enough to demonstrate beyond any reasonable doubt that ChatGPT has fully internalized the rules of chess and is not relying on memorization or other, shallower patterns.
    The "magic prompt" that I've been using is the following:
    1. e4
    and then in my later replies, providing the full current game score to ChatGPT as my message to it, e.g.:
    2. Nh3 fxe4
    3. Nf4 Nf6
    4. b4 e5
    5. b5
    This "magic prompt" isn't original to me - soon after GPT-4 came out, a friend of mine told me about it, having seen it as a comment on HackerNews. (Sorry, anonymous HackerNews commenter - I'd love to credit you further, and will if you find this post and message me.)
    The especially interesting thing about this is the sharp contrast between how ChatGPT-3.5 performs with and without the prompt. With the prompt, ChatGPT plays consistently legally and even passably well for the first 30 or so moves; without the prompt, ChatGPT is basically totally unable to play a fully legal game of chess.
    Here are a few example games of ChatGPT playing or attempting to play chess under various conditions.
    ChatGPT-3.5, with the magic prompt
    Playing against me
    Lichess study, ChatGPT conversation link
    I play white, ChatGPT plays black. In this game, I intentionally play a bizarre opening, in order to quickly prove that ChatGPT isn't relying on memorized opening or ideas in its play. This game isn't meant to show that ChatGPT can play well (since I'm playing atrociously here), only that it can play legally in a novel game. In my view, this game alone is more than enough evidence to put to bed the notion that ChatGPT "doesn't know" the rules of chess or that it's just regurgitating half-remembered ideas from its training set; it very clearly has an internal representation of the board, and fully understands the rules. In order to deliver checkmate on move 19 with 19...Qe8# (which it does deliberately, outputting the pound sign which indicates checkmate), ChatGPT needed to "see" the contributions of at least six different black pieces at once (the bishop on g4, the two pawns on g7 and h6, the king on f8, the queen on e8, and either the rook on h8 or the knight on f6).
    Playing against Lichess Stockfish Level 1
    Lichess game, ChatGPT conversation link
    Stockfish level 1 has an Elo of around 850. Stockfish is playing white and ChatGPT is playing black. In this game, ChatGPT quickly gains a dominating material advantage and checkmates Stockfish Level 1 on move 22.
    Playing against Lichess Stockfish Level 2
    Lichess game, ChatGPT conversation link
    Stockfish level 2 has an Elo of around 950. Stockfish is playing white and ChatGPT is playing black. In this game, ChatGPT starts a dangerous kingside attack and gai...
    10 min
  • LW - Steven Wolfram on AI Alignment by Bill Benzon
    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: Steven Wolfram on AI Alignment, published by Bill Benzon on August 21, 2023 on LessWrong.
    Joe Walker has a general conversation with Wolfram about his work and things and stuff, but there are some remarks about AI alignment at the very end:
    WALKER: Okay, interesting. So moving finally to AI, many people worry about unaligned artificial general intelligence, and I think it's a risk we should take seriously. But computational irreducibility must imply that a mathematical definition of alignment is impossible, right?
    WOLFRAM: Yes. There isn't a mathematical definition of what we want AIs to be like. The minimal thing we might say about AIs, about their alignment, is: let's have them be like people are. And then people immediately say, "No, we don't want them to be like people. People have all kinds of problems. We want them to be like people aspire to be.
    And at that point, you've fallen off the cliff. Because, what do people aspire to be? Well, different people aspire to be different and different cultures aspire in different ways. And I think the concept that there will be a perfect mathematical aspiration is just completely wrongheaded. It's just the wrong type of answer.
    The question of how we should be is a question that is a reflection back on us. There is no "this is the way we should be" imposed by mathematics.
    Humans have ethical beliefs that are a reflection of humanity. One of the things I realised recently is one of the things that's confusing about ethics is if you're used to doing science, you say, "Well, I'm going to separate a piece of the system," and I'm going to say, "I'm going to study this particular subsystem. I'm going to figure out exactly what happens in the subsystem. Everything else is irrelevant."
    But in ethics, you can never do that. So you imagine you're doing one of these trolley problem things. You got to decide whether you're going to kill the three giraffes or the eighteen llamas. And which one is it going to be?
    Well, then you realise to really answer that question to the best ability of humanity, you're looking at the tentacles of the religious beliefs of the tribe in Africa that deals with giraffes, and this kind of thing that was the consequence of the llama for its wool that went in this supply chain, and all this kind of thing.
    In other words, one of the problems with ethics is it doesn't have the separability that we've been used to in science. In other words, it necessarily pulls in everything, and we don't get to say, "There's this micro ethics for this particular thing; we can solve ethics for this thing without the broader picture of ethics outside."
    If you say, "I'm going to make this system of laws, and I'm going to make the system of constraints on AIs, and that means I know everything that's going to happen," well, no, you don't. There will always be an unexpected consequence. There will always be this thing that spurts out and isn't what you expected to have happen, because there's this irreducibility, this kind of inexorable computational process that you can't readily predict.
    The idea that we're going to have a prescriptive collection of principles for AIs, and we're going to be able to say, "This is enough, that's everything we need to constrain the AIs in the way we want," it's just not going to happen that way. It just can't happen that way.
    Something I've been thinking about recently is, so what the heck do we actually do? I was realising this. We have this connection to ChatGPT, for example, and I was thinking now it can write Wolfram Language code, I can actually run that code on my computer. And right there at the moment where I'm going to press the button that says, "Okay, LLM, whatever code you write, it's going to run on my computer," I'm like, "That's probably a bad idea," because, I don't know, it's going ...
    6 min
  • EA - "Being an EA" - Dissertation on EA by Joanna (Asia) Wiaterek
    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: "Being an EA" - Dissertation on EA, published by Joanna (Asia) Wiaterek on August 21, 2023 on The Effective Altruism Forum.
    As part of my undergraduate course, I have written a dissertation in Social Anthropology on "'Being an EA': How Effective Altruism is understood and lived out by members of its transnational community."
    A few points about it:
    I am not proposing conclusions about the entirety of the EA due to its sheer diversity, but merely work with a few specific ethnographic examples.
    Two questions I tackled:i) What does it mean to "be an EA" for individual members of the transnational EA community?ii) What are the characteristics and motives of the EAs' "altruism"?
    Key takeaways:i) I argue that EA can be productively analysed as a lifestyle movementii) I also argue that the kind of altruism among EAs resembles a spatio-temporally extended form of sharing
    I worked on it between August 2022 and May 2023, so it captures briefly some reflections on the SBF crisis and its impact on EAs' self-identification with the movement.
    It was my first article-length piece of work, so it is far from perfect, but I hope it can prompt productive discussion on the future of the movement and its community.
    I initially aimed to write it for: community organisers, academics interested in social movements, people intrigued by EA; but I hope that the personal stories described in it might be quite inspiring to other EAs, too.
    Because of the marking boycott, I cannot post the full version here yet, but if you would like to get access to the document, please fill out this Google Form.
    If you'd like to chat about it, feel free to contact me on [email protected] or Calendly:. I would like to say a big thank you again to those of you who contributed to this dissertation!
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    2 min
  • EA - My EA Journey by Eli Kaufman
    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 EA Journey, published by Eli Kaufman on August 20, 2023 on The Effective Altruism Forum.
    Summary:
    This is the story of my EA journey. Sharing it as I believe that my story could be relevant to others particularly those who arrive in EA mid-career or not having worked on a cause area before.
    Background:
    Around two years ago a podcast appeared on my spotify playlist. It was one of the 80,000 hours episodes. It piqued my curiosity and after listening to it I wanted to learn more. Listening to some other episodes I realized there are a lot of concepts I don't understand so that drove me to look for more resources online. Up until that point I have not come across EA. I had a broad interest in global health & development, philanthropy and doing good but no practical insight how to go about this. I started reading about EA online, then signed up for the virtual intro fellowship. Within a couple of months I read a few books (Doing Good Better, The Life You Can Save, The Precipice to name a few). I didn't expect that, but soon I started having realizations that would make an impact on my life.
    Career:
    I work in IT with background in operations and specialize in implementing solutions based on Salesforce platform for organizations. For years I have been working with organizations that I didn't feel particularly aligned with. I kind of accepted the reality that my work is not where my passion is and had other hobbies and interests which I was excited about. The realization that I could use my skills and expertise to do something impactful and meaningful was an important one. I figured out that sooner or later I will come across the right opportunity and in the meantime was actively networking. Here's a post I wrote at the time.
    Community:
    One of the main sources of inspiration for me was the people I met in the EA community and the stories they shared. Being part of such a community dedicated to doing good was something that resonated with me. I found out that where I live (Amsterdam) has a pretty active EA community (including an awesome co-working space) so had a chance to attend meetups, virtual and in-person events, attended a couple of EAGx conferences, signed up to a bunch of Slack spaces and interacted with people from around the world.
    Fast forward:
    I signed a giving pledge as I felt this is something that makes sense to me. I applied to a job with The END Fund and started there earlier this year, feeling excited about using my skills to help an highly effective organization in the field of Neglected Tropical Diseases.
    Main lessons:
    Initially it may seem that everyone in EA are so dedicated and it makes you feel you're not doing enough. Don't try to be a maximalist! Just do your bit.
    I found people in the EA community to be helpful, and willing to share their experiences, offer advice, point newcomers to a useful direction. (If you would like to chat feel free to reach out here)
    There are many resources out there such as career advice, coaching, specific professional and cause area groups, podcasts.
    Networking is perhaps the most important aspect if you're looking for a way to get more involved or make a career change. If you happen to live in a place with an active community - check out local events. If you don't - EA Anywhere is a good starting point. EAGx conferences are a great way to learn more and talk to people.
    Don't feel that only highly specialized experts can contribute to making the world better. Each of us has something to bring.
    Go ahead and take the first step - write a post, go to a conference. Who knows, perhaps it will put you on a life-changing journey?
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    4 min