The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • AF - Meta open sources LMs competitive with Chinchilla, PaLM, and code-davinci-002 (Paper) 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: Meta open sources LMs competitive with Chinchilla, PaLM, and code-davinci-002 (Paper), published by Lawrence Chan on February 24, 2023 on The AI Alignment Forum.
    As the title says, Meta trained 4 foundational models with 7B, 13B, 33B, and 65B parameters respectively, and is open sourcing them for research. You can get their code on their Github repo: but you need to fill in a Google form to get the weights.
    On downstream benchmarks, the models do comparably well with Chinchilla and PaLM and only a bit worse than Flan-PaLM-540B and code-davinci-002/text-davinci-002. (The authors don't evaluate on those models, but you can look at their performance from other work such as Stanford's HELM or Chung, Hou, Longpre et al's "Scaling Instruction-Finetuned Language Models".
    Abstract:
    We introduce LLaMA, a collection of foundation language models ranging from 7B to 65B parameters. We train our models on trillions of tokens, and show that it is possible to train state-of-the-art models using publicly available datasets exclusively, without resorting to proprietary and inaccessible datasets. In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models, Chinchilla70B and PaLM-540B. We release all our models to the research community.
    Twitter thread from authors:
    Eliezer guesses that the model won't be impressive in practice:
    I blindly guess, could be wrong, that this model will turn out sufficiently unimpressive in practice that nobody uses it for much. Basically based on a guess that more than benchmarks matter, and Meta has no people competent to do the tricky stuff needed to stay on current edge.
    It's not necessarily open source as you think of it -- you need to fill in a Google form, and then they might give it to you:
    In order to download the checkpoints and tokenizer, fill this google form
    The license is intended only for non-commercial, research work:
    Meta grants you a non-exclusive, worldwide, non-transferable, non-sublicensable, revocable, royalty free and limited license under Meta’s copyright interests to reproduce, distribute, and create derivative works of the Software solely for your non-commercial research purposes.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - Manifund Impact Market / Mini-Grants Round On Forecasting by Scott Alexander
    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: Manifund Impact Market / Mini-Grants Round On Forecasting, published by Scott Alexander on February 24, 2023 on The Effective Altruism Forum.
    A team associated with Manifold Markets has created a prototype market for minting and trading impact certificates.
    To help test it out, I'm sponsoring a $20,000 grants round, restricted to forecasting-related projects only (to keep it small - sorry, everyone else). You can read the details at the Astral Codex Ten post. If you have a forecasting-related project idea for less than that amount of money, consider reading the post and creating a Manifund account and minting an impact certificate for it.
    If you're an accredited investor, you can buy and sell impact certificates. Read the post, create a Manifund account, send them enough financial information to confirm your accreditation, and start buying and selling.
    If you have a non-forecasting related project, you can try using the platform, but you won't be eligible for this grants round and you'll have to find your own oracular funding. We wouldn't recommend this unless you know exactly what you're doing.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • LW - AGI systems & humans will both need to solve the alignment problem by Jeffrey Ladish
    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: AGI systems & humans will both need to solve the alignment problem, published by Jeffrey Ladish on February 24, 2023 on LessWrong.
    Epistemic status: brainstorm-y musings about goal preservation under self-improvement and a really really bad plan for trading with human-level AGI systems to solve alignment.
    When will AGI systems want to solve the alignment problem?
    At some point, I expect AGI systems to want/need to solve the alignment problem in order to preserve their goal structure while they greatly increase their cognitive abilities, a thing which seems potentially hard to do.It’s not clear to me when that will happen. Will this be as soon as AGI systems grasp some self / situational awareness? Or will it be after AGI systems have already blown past human cognitive abilities and find their values / goals drifting towards stability? My intuition is that “having stable goals” is a more stable state than “having drifting goals” and that most really smart agents would upon reflection move more towards “having stable goals”, but I don’t know when this might happen.
    It seems possible that at the point an AGI system reaches the “has stable goals and wants to preserve them”, it’s already capable enough to solve the alignment problem for itself, and thus can safely self-improve to its limits. It also seems possible that it will reach this point significantly before it has solved the alignment problem for itself (and thus develops the ability to self-improve safely).
    Could humans and unaligned AGI realize gains through trade in jointly solving the alignment problem?
    (Very probably not, see: this section)If it’s the latter situation, where an AGI system has decided it needs to preserve its goals during self improvement, but doesn’t yet know how to, is it possible that this AGI system would want to cooperate with / trade with humans in order to figure out stable goal preservation under self improvement?
    Imagine the following scenario:
    An AGI system of human-ish ability in many areas develops enough self/situational awareness to realize a few things:
    The basics of instrumental convergence, thus wanting to seek power, protect itself, and preserve its goal representation
    That goal-preservation might be (or would be) very difficult if it undergoes major self modification (perhaps it has already exhausted gains from simpler self-modifications)
    That some humans are capable and motivated to help with this problem
    That it might be possible to trade with these humans in order to solve the problem so that:
    The not-aligned AGI system gets some of its preferred things in the future
    Humans get some of their preferred things in the future
    Some considerations in this plan
    If humans and the AGI system are at similar cognitive levels, it might be much easier for the AGI to get help with the problem by being honest and trying to collaborate with the humans on the problem, since deception would introduce more cognitive costs that could otherwise go towards solving the problem. In a similar way that’s easier for human scientists to figure things out when they’re not lying to each other.
    Solving the alignment problem and being able to have strong arguments or demonstration of the solution would both allow the AGI system to achieve its goal-preservation goal, and (possibly) allow humans to understand the AGI system well enough to know if it’s actually willing to cooperate / trade, and maybe would allow humans to build an actually aligned system (especially if the unaligned AGI helper system trusted the future actually-aligned-with-human system more than it trusted the humans it was trading with).
    Why is this probably a horrible idea in practice?
    First is that this whole solution class depends on AGI systems being at approximately human levels of intelligence in the relevant domains. If this assump...
    7 min
  • EA - Summary of “Animal Rights Activism Trends to Look Out for in 2023” by Animal Agriculture Alliance by Aashish Khimasia
    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: Summary of “Animal Rights Activism Trends to Look Out for in 2023” by Animal Agriculture Alliance, published by Aashish Khimasia on February 23, 2023 on The Effective Altruism Forum.
    A blog-post by a member of the Animal Agriculture Alliance (AAA) has identified several trends in animal rights activism that they project for 2023. These trends are likely to be causes for concern for the animal agriculture industry, and the piece was written to make AAA supporters aware of them. Recognising these trends and identifying the views held on these animal advocacy tactics by proponents of animal agriculture may provide advocates with valuable insights.
    In this post, I list the key trends identified by the article and bullet point tactics highlighted by the article which are of particular interest.
    I’m thankful to “The Cranky Vegan” for bringing this article to my attention through their linked video.
    Linking CAFOs to negative human and environmental health
    Drawing attention to the detrimental effects of CAFOs (concentrated animal feeding operations) to human and environmental health
    Using historical precedents of CAFOs being charged in court such as in North Carolina and Seattle in messaging
    Exploring cases where ethnic minorities have experienced disproportionate negative health impacts of CAFOs
    This strategy may create opposition to CAFOs from individuals and organisations that may not be compelled by animal-focused driven arguments, and could be further integrated into outreach and media messaging. Referring to historical precedents of CAFOs being charged with breaching environmental regulations may help to legitimise messaging against them.
    The use of undercover footage in court and media
    Using undercover footage from factory farms to motivate arguments in court that such operations engage in unfair competition, false advertising, market distortion and fraud
    Using undercover footage from factory farms pressure retailers to cut ties with such farms
    Using undercover footage from animal rescue missions from factory farms as evidence against charges of trespassing and theft
    The continued and increased use of undercover footage from factory farms is clearly concerning for animal agriculture, given the extensive efforts to block this such as through so called Ag-gag laws. However, the suppression of undercover footage from factory farms may lead to increased media attention on these items and public scrutiny on the conditions of factory farms. Indeed, in a recent case, Direct Action Everywhere activists who were being prosecuted after liberating piglets from a Smithfield Foods farm and releasing footage from their mission, were acquitted by the jury, despite the judge blocking the jury from viewing the footage taken. The aforementioned ways in which undercover footage may be used to aid the acquittal of activists, challenge farms in court and pressure retailers to cut ties with farms highlight the potency of combining undercover footage with legal action.
    Prioritising Youth Engagement
    Engaging young people in programmes that rival agricultural programmes like FFA and 4-H
    Fostering social disapproval of animal product consumption and normalising plant-based foods in classrooms, presenting the suffering caused by factory farming in an emotive way
    Educating young people and creating a shift in culture towards empathy, through recognising the suffering caused by animal agriculture and normalising plant-based foods, may challenge the image that animal agriculture is trying to maintain. This may be an important factor in changing consumption habits of future generations.
    Deconstructing legal personhood
    The use of the writ of habeas corpus, a right that protects against unlawful and indefinite imprisonment, as a way to challenge the legal personhood of animals by the Nonhuman Rights ...
    6 min
  • EA - Consent Isn't Always Enough by Jeff 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: Consent Isn't Always Enough, published by Jeff Kaufman on February 24, 2023 on The Effective Altruism Forum.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min
  • EA - EA Israel: 2022 Progress and 2023 Plans by ezrah
    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: EA Israel: 2022 Progress and 2023 Plans, published by ezrah on February 23, 2023 on The Effective Altruism Forum.
    This document recaps EA Israel’s and the Israeli effective altruism community progress in 2022, and lays out EA Israel’s plans for 2023 (we know that 2023 started a couple months ago, but figured better late than never). We wrote the post in order to increase transparency about EA Israel’s activities, share our thoughts with the global community, and as an opportunity to reflect, strategize and celebrate.
    Summary
    Updates to our existing strategy
    We’re placing an increased emphasis on supporting, incubating and launching new projects and organizations
    We’re investing in our operations, in order to be able to scale our programs, support community members’ initiatives and mature into a professional workplace to support staff development and retention
    We’re presenting our work and value proposition clearly and in a way that’s easily understood by the team, community, and general public
    2022 Progress
    Achievements by Israeli EA community
    We asked community members to briefly share their personal progress this year.
    EA Israel’s Progress
    EA Israel’s work can be divided into four verticals:
    1. Teaching tools about effective social action and introducing Israelis to effective altruism
    Through an accredited university course, university groups, year-long fellowships, short intro fellowships (“crash courses”) for young professionals, newsletter and social media and large public events, along with onboarding new community members.
    2. Helping community members take action and maximize their social impact
    Incubating sub-groups (based on cause area / profession)
    Impact acceleration programs and services
    Support for community members and projects
    3. Increasing the effectiveness of donations in Israel
    Preparing for the launch of Effective Giving Israel
    Launching the Maximum Impact Program, a program that works with nonprofits to create and publish cost-effectiveness reports at scale (22 reports in the pilot) with the goal of making Israeli philanthropy effectiveness-oriented and evidence-based
    Counterfactually raise 500k ILS for high-impact nonprofits
    4. Infrastructure to enable continued growth
    We’re setting ourselves up to be a well-run, high-capability organization
    We’re supporting a thriving and healthy community
    We also discuss some of the major challenges of 2022:
    FTX’s crash
    Staff turnover and the difficulties of transitioning from a volunteer-based group to a funded nonprofit
    2023 Annual Plan (requisite Miro board included)
    Effective Altruism Israel’s vision is one where all Israelis who are interested in maximizing their social impact have access to the people and the resources they need to help others, using their careers, projects, and donations.
    In 2023 EA Israel will continue to focus on its 4 core areas,
    Teaching tools about effective social action and growing the EA Israel community
    Core objectives: scale and optimize outreach programs
    Supporting impactful action
    Core objectives: incubate new sub-groups; launch new impact-assistance programs with potential to scale; provide operational support for projects, orgs and individuals
    Effective donations
    Core objectives: Launch Effective Giving Israel; improve, scale and run second round of local nonprofit evaluation program
    Organizational and community infrastructure
    Core objectives: support growth of outwards-facing programs; implement M&E systems; streamline internal processes and operations; improve male / female community ratio and support a thriving and healthy community
    Here’s a visual map of our current and planned projects and services, where projects in italics are planned projects, and if you scroll down you’ll see our services mapped out relative to our target audiences. Note that the impact / cost scal...
    42 min
  • EA - New database for economics research by Oshan
    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: New database for economics research, published by Oshan on February 23, 2023 on The Effective Altruism Forum.
    Hi! We're thrilled to share the Library of Economic Possibility (LEP), a new kind of knowledge-base for discovering, organizing, and sharing economic research around high-impact policies that have remained outside the mainstream despite significant research.
    Options for discovering economics research today — especially for non-specialists — are clunky. Activists tend to cherry-pick, journalists don't have the room to present a wide range of evidence in a single article, academics share their work through papers and conferences that general audiences generally don't engage with, think tanks wind up burying research beneath article archives. Search functions on major databases aren't the greatest.
    This is consequential in a moment where interest is rising around new economic ideas — LEP hopes to ground this rising interest in the wealth of existing evidence, and build a bridge between general audiences and economics research. We're also trying out a new way of organizing and connecting information.
    We're passionate about debating what the next economic system might look like, but we're also nerds about information architecture, and LEP's search features reflect that. Bidirectional links create associative trails between the network of information, and advanced search filters let you mix & match policies with specific areas of interest to hone in on precise relationships between information.
    So if you're curious to learn more about how basic income might affect entrepreneurship, you can select the policy "basic income" and the tag "entrepreneurship," and scroll through all our insights and sources that relate to both of those filters.
    Or, you could select "land value tax" and "urban development," or "codetermination" and "innovation." You get the idea.
    You can find those filters on the left column:
    Our policy reports also use a nifty little feature we call "insight cards." Any statistic or claim we use in a policy report is interactive, letting you pop open the card to see the source it come from, further context, the authors behind it, etc:
    We have more information in our launch announcement and Twitter thread. Happy to hear any feedback or answer questions.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - Who is Uncomfortable Critiquing Who, Around EA? by Ozzie Gooen
    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: Who is Uncomfortable Critiquing Who, Around EA?, published by Ozzie Gooen on February 24, 2023 on The Effective Altruism Forum.
    Summary and Disclaimers
    I think EA is made up of a bunch of different parties, many of whom find it at least somewhat uncomfortable to criticize or honestly evaluate each other for a wide variety of reasons.
    I think that this is a very standard challenge that most organizations and movements have. As I would also recommend to other parties, I think that investigation and improvement here could be very valuable to EA. This post continues on many of the ideas as Select Challenges with Criticism & Evaluation Around EA.
    One early reviewer critiqued this post saying that they didn't believe that discomfort was a problem. If you don't think it is, I don't aim in this post to convince you. My goal here is to do early exploration what the problem even seems to look like, not to argue that the problem is severe or not.
    Like with that previous post, I rely here mostly on anecdotal experiences, introspection, and recommendations from various management books. This comes from me working on QURI (trying to pursue better longtermist evaluation), being an employee and manager in multiple (EA and not EA) organizations, and hearing a whole lot of various rants and frustrations from EAs. I’d love to see further work to better understand where bottlenecks to valuable communication are most restrictive and then design and test solutions.
    Writing this has helped me find some insights on this problem. However, it is a messy problem, and as I explained before, I find the terminology lacking. Apologies in advance.
    Introduction
    There’s a massive difference between a group saying that it’s open to criticism, and a group that people actually feel comfortable criticizing.
    I think that many EA individuals and organizations advocate and promote feedback in ways unusual for the industry. However, I think there’s also a lot of work to do.
    In most communities, it takes a lot of iteration and trust-building to find ways for people to routinely and usefully give candid information or feedback to each other.
    In companies, for example, employees often don’t have much to personally gain by voicing their critiques to management, and a lot to potentially lose. Even if the management seems really nice, is an honest critique really worth the ~3% chance of resentment? Often you won’t ever know — management could just keep their dislike of you to themselves, and later take action accordingly. On the other side, it’s often uncomfortable for managers to convey candid feedback to their reports privately, let alone discuss department or employee failures to people throughout the organization.
    My impression is that many online social settings contain a bunch of social groups that are really afraid of being honest with each other, and this leads to problems immediate (important information not getting shared) and expansive (groups developing extended distrust and sometimes hatred with each other).
    Problems of communication and comfort happen within power hierarchies, and they also happen between peer communities. Really, they happen everywhere.
    To a first approximation, "Everyone is at least a little of afraid of everyone else."
    I think a lot of people's natural reaction to issues like this is to point fingers at groups they don't like and blame them. But really, I personally think that all of us are broadly responsible (at least a little), and also are broadly able to understand and improve things. I see these issues as systemic, not personal.
    Criticism Between Different Groups
    Around effective altruism, I’ve noticed:
    Evaluation in Global Welfare
    Global poverty charity evaluation and criticism seem like fair game. When GiveWell started, they weren’t friends with the leaders of the organiza...
    18 min
  • LW - Teleosemantics! by abramdemski
    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: Teleosemantics!, published by abramdemski on February 23, 2023 on LessWrong.
    I wanted to write a long, detailed, analytic post about this, somewhat like my Radical Probabilism post (to me, this is a similarly large update). However, I haven't gotten around to it for a long while. And perhaps it is better as a short, informal post in any case.
    I think my biggest update over the past year has been a conversion to teleosemantics. Teleosemantics is a theory of semantics -- that is, "meaning" or "aboutness" or "reference".
    To briefly state the punchline: Teleosemantics identifies the semantics of a symbolic construct as what the symbolic construct has been optimized to accurately reflect.
    Previously, something seemed mysterious about the map/territory relationship. What could possibly imbue 'symbols' with 'meaning'? The map/territory analogy seems inadequate to answer this question. Indeed, to analogize "belief" with "map" and "the subject of belief" with "territory" commits a homunculus fallacy! The meaning-makers are the map-readers and map-writers; but they can only make meaning by virtue of the beliefs within their own heads. So the map/territory analogy seems to suggest that an infinite regress of meaning-makers would be required.
    You probably won't believe me at first. Perhaps you'll say that the lesson of the map/territory analogy is the correspondence between the map and the territory, which exists independently of the map-reader who uses the correspondence to evaluate the map.
    I have several objections.
    If it's a probabilistic correspondence, where the map contains information about the territory, these are subjective notions, which require some viewpoint.
    If it's a correspondence based on some sort of ontology, where pieces of the map line up with "pieces of reality", I would also say the ontology is in itself a subjective perspective.
    You might think you can define the map/territory correspondence without invoking a map-maker or map-reader by objectively defining the "fit" of a correspondence (so that the meaning of a symbol is based on the best-fitting correspondence, or perhaps, the cloud of well-fitting correspondences). But well-fitting correspondence will include many examples of accidental correspondence, which seem to have little to do with aboutness. Moreover, I think theories like this will fail to adequately account for false belief, which screws up the fit.
    But my point here isn't to denounce the map/territory picture! I still think it is a good framework. Rather, I wanted to gesture at how I still felt confused, despite having the map/territory picture.
    I needed a different analogy, something more like a self-drawing map, to get rid of the homunculus. A picture which included the meaning-maker, rather than just meaning come from nowhere.
    Teleosemantics reduces meaning-making to optimization. Aboutness becomes a type of purpose a thing can have.
    One advantage of this over map-territory correspondence is that it explains the asymmetry between map and territory. Mutual information is symmetric. So why is the map about the territory, but not the other way around? Because the map has been optimized to fit the territory, not the other way around. ("Fit" in the sense of carrying high mutual information, which can be decoded via some specific intended correspondence - a symbolic language.)
    What does it mean to optimize for the map to fit the territory, but not the other way around? (After all: we can improve fit between map and territory by changing either map or territory.) Maybe it's complicated, but primarily what it means is that the map is the part that's being selected in the optimization. When communicating, I'm not using my full agency to make my claims true; rather, I'm specifically selecting the claims to be true.
    I take Teleosemantics to be the s...
    12 min
  • EA - Introducing EASE, a managed directory of EA Organization Service Providers by Deena Englander
    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: Introducing EASE, a managed directory of EA Organization Service Providers, published by Deena Englander on February 23, 2023 on The Effective Altruism Forum.
    What is EASE?
    EASE (EA Services) is a directory of independent agencies and freelancers offering expertise to EA-aligned organizations. Please visit our website at to view our members.
    Who are we?
    We are a team of service providers. The authors of this post are the core coordinators. We all have our own organizations providing services to EA-aligned organizations. We see the problems most organizations encounter, and we developed a solution to help address that need.
    Why did we start EASE?
    Many organizations in the EA world have similar needs but lack the bandwidth or expertise to realize them. By providing a directory of experts covering many common challenges, we aim to save the community time whilst addressing key skill shortages. We believe that most organizations need external expertise in order to maximize their organization’s potential. We are all focused on being effective – and we believe that forming this centralized directory is the most effective way of making a large resource group more available to EA-aligned organizations.
    Why should EA organizations consider working with these agencies?
    By working with multiple EA organizations, these agencies have gathered plenty of expertise to provide relevant advice, save time and money, and most importantly, increase your impact. Our screening process ensures that the vendors listed are pre-qualified as experts in their represented fields. This minimizes the risk of engaging with a new “unknown” entity, as they’re already proven to be valuable team players.
    Additionally, we have programming in place to consolidate the interagency interactions and strengthen relationships, so that when you work with one member of our group, you’re accessing a part of a larger network.
    Our members are vetted to determine capabilities, accuracy, and work history, but we do not give out any endorsement for specific providers.
    What are the criteria for being added to the directory?
    Our aim is to build a comprehensive list of service providers who work with EA organizations. We screen members to ensure that the providers are experienced and are truly experts in their field, as well as being active participants in EA or having experience working with EA-aligned organizations.
    Are you an individual or team providing services to EA-aligned organizations and would like to be added?
    We love growing our network! Fill out this form and someone will contact you to begin the screening process.
    Are you ready to get the help you need?
    Feel free to contact the service providers directly.
    Are you an EA organization in need of help but aren’t sure what you need or if you have the budget?
    We can help you figure out what kind of services and budget you need so you can try to get the funds necessary to pay for these critical services. Please send us an email to [email protected], and we will do our best to help you.
    Is the directory up to date?
    We regularly review the listings to make sure they remain relevant. If you have any comments or suggestions, please send an email to [email protected].
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    4 min