The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - Reflective journal entries using GPT-4 and Obsidian that demand less willpower. by Solenoid Entity
    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: Reflective journal entries using GPT-4 and Obsidian that demand less willpower., published by Solenoid Entity on April 15, 2023 on LessWrong.
    Something I've wanted to do for years, but can't make myself do consistently because I'm bad at forming new habits and seem to have a limited reservoir of willpower that is generally depleted by the end of the day:
    Write a short journal entry at the end of every day, summarizing the day and reflecting on it.
    Put this in Obsidian, and identify key concepts and make them back-links.
    Enjoy as, over several years of consistent writing, the map of my life unfolds through the graph in Obsidian. Maybe this generates helpful insights, shows me patterns. Mostly it's just useful as a resource.
    It's a lot of effort to be detailed and reflective, and it's a lot of effort to add the back-links.
    The solution:
    Rather than having to create a detailed, reflective entry from whole cloth, I can just write some short, abbreviated, incoherent mumbled paragraph, like a surly teenager at family dinner, and then ask ChatGPT to ask more questions to flesh out the story.
    This works shockingly well.
    It's surprisingly easy to write detailed answers to specific QUESTIONS about your day. It's like a good listener, it can draw you out with its insightful questions.
    Then you can ask it to mark all the key words, names, etc, with double square brackets.
    The end result isn't perfect but it's pretty good, and much, much better than I would ever do all by myself. It takes less than half the time and virtually no willpower.
    Here's what I prompted ChatGPT today (That diary entry is roughly the level of writing I'm capable of at the end of the day.)
    Ask me 10 questions about the following diary entry, aiming to 'flesh it out' with extra details. If there are inconsistencies or missing information/context, you can ask for clarification or further information. Your aim is to ask questions that help make this a more complete and coherent account of the day. Some of the questions should relate to the theme of 'gratitude', and help provide more context and identify things to be grateful about in the text. Here's the text:Woke up early for no reason, lay in bed for 20 minutes thinking, then got up, said goodbye to Xena, headed off at around 7am for the train up to Waitara. Ended up missing the stupid thing because I was waiting on the wrong platform like an idiot. Still got to coffee in time with Adele, which was a relief since I've been an unreliable friend recently.
    Good coffee with her, turns out the guy who owns the cafe knows her as a regular and she doesn't even have to order, he just brings her french toast as soon as he sees her sit down. Anyway we talked about her new startup, then I left and caught up with my friend who's building a paperclip-optimising machine, which sounds stressful for him tbh, but his new girlfriend is nice, they seem like a good match. Got the late train over to Emu Plains and finally got to see my parents. Hopefully cook something nice with Mum tomorrow before she goes to line-dancing club.
    Response:
    My continuation:
    1. She's my friend from a long time ago, but she's been super busy with her job as a warrior princess and doesn't have much time for friends. I was there the night before to have dinner and drinks because she finally got a week off work, ended up having more drinks than planned. Her couch was not very comfortable which is probably why I woke up early. 2. Really annoyed with myself, it was such an avoidable mistake. Honestly, it was pretty positive because I was only 10 minutes late, and Adele was running late anyway. I was able to find a good alternative train route using the app, so it worked out ok. 3. We try to have a phone call every couple of weeks, but it's hard because her life is so unpredictable what with the record d...
    7 min
  • EA - AI Safety Europe Retreat 2023 Retrospective by Magdalena Wache
    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: AI Safety Europe Retreat 2023 Retrospective, published by Magdalena Wache on April 14, 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 - Floors and Ceilings, Frameworks and Feelings: SoGive's Impact Analysis Toolkit for Evaluating StrongMinds by ishaan
    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: Floors and Ceilings, Frameworks and Feelings: SoGive's Impact Analysis Toolkit for Evaluating StrongMinds, published by ishaan on April 15, 2023 on The Effective Altruism Forum.
    Primary author: Ishaan Guptasarma, Principal Analyst at SoGive
    We recently announced that we will be performing an independent assessment of StrongMinds. This is the first in a series of posts which will culminate in that assessment.
    Executive Summary
    In order to conduct our review of StrongMinds, we needed to make decisions about how to measure the effectiveness of psychotherapy. The approaches that we considered were mainly:
    An approach used by Happier Lives Institute (HLI), which measures the cumulative effect size of therapy over time (SD-Years) by postulating an initial effect which exponentially decays over time, and integrating under the curve.
    An approach used by some academics, which reports remission rates and number needed to treat (NNT) associated with psychotherapy, and of relapse rates at various time points as reported in longitudinal follow-ups.
    We decided that the SD-Years approach used by HLI best captures what we’re trying to capture, because remission, relapse, and NNT use cut-offs which are arbitrary and poorly standardised.
    The drawback of this method is that it's based on effect sizes, which can become inflated when ceiling and floor effects artificially reduce the standard deviation. Such artefacts are rarely ever accounted for in meta-analyses and literature that we have encountered.
    For each step in our methodology, we've created spreadsheet tools which others can use to quickly check and replicate our work and do further analysis. These tools can do:
    Meta-analysis, for calculating standardised mean differences and aggregating effect sizes across multiple studies to estimate the impact of a therapeutic intervention.
    Linear regressions and meta-regressions, to calculate the rate at which therapeutic effects decay over time.
    Conversion from remission rates and NNTs into effect sizes, and relapse rates into decay rates, and vice versa.
    Conversion of scores between different depression questionnaires.
    Calculation of "standard deviations of improvement" for a single patient, for building intuitions.
    About SoGive: SoGive does EA research and supports major donors. If you are a major donor seeking support with your donations, we’d be keen to work with you. Feel free to contact Sanjay on [email protected].
    0 Introduction
    How should the EA community reason about interventions relating to subjective well being? We typically conceptualise the impact of donating to global health anti-malaria charities in terms of figures such as "£5,000 per child's life saved". While evaluating such estimates is difficult, the fundamental question arguably has a "yes or no" answer: was a child's death averted, or not?
    Measuring impact on subjective well being, which is continuous rather than discrete and is typically measured by self-report, requires a different framework. This post explains the dominant frameworks for thinking about this, explores some of the complications that they introduce, and introduces spreadsheet tools for deploying these frameworks. These tools and analytical considerations that will lay the groundwork for subsequent work.
    We recommend this post to anyone interested in doing analysis on mental health. It may also be useful to anyone doing impact evaluations on continuous phenomena which manifest as unimodal distributions, especially those which might be approximated as normal distributions.
    1 The SD-year framework, and why we prefer it to the alternative
    Academic studies usually measure the impact of a mental health intervention by using questionnaires to ask how people feel before and after the intervention, and then comparing their scores to a control group which did ...
    37 min
  • LW - Concave Utility Question by Scott Garrabrant
    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: Concave Utility Question, published by Scott Garrabrant on April 15, 2023 on LessWrong.
    This post will just be a concrete math question. I am interested in this question because I have recently come tor reject the independence axiom of VNM, and am thus playing with some weaker versions.
    Let Ω be a finite set of deterministic outcomes. Let L be the space of all lotteries over these outcomes, and let ⪰ be a relation on L. We write A∼B if A ⪰ B and B ⪰ A. We write A≻B if A⪰B but not A∼B.
    Here are some axioms we can assume about ⪰:
    A1. For all A,B∈L, either A⪰B or B⪰A (or both).
    A2. For all A,B,C∈L, if A⪰B, and B⪰C, then A⪰C.
    A3. For all A,B,C∈L, if A⪰B, and B⪰C, then there exists a p∈[0,1] such that B∼pA+(1−p)C.
    A4. For all A,B∈L, and p∈[0,1] if A⪰B, then pA+(1−p)B⪰B.
    A5. For all A,B∈L, and p∈[0,1], if p>0 and B⪰pA+(1−p)B, then B⪰A.
    Here is one bonus axiom:
    B1. For all A,B,C∈L, and p∈[0,1], A⪰B if and only if pA+(1−p)C⪰pB+(1−p)C.
    (Note that B1 is stronger than both A4 and A5)
    Finally, here are some conclusions of successively increasing strength:
    C1. There exists a function u:L[0,1] such that A⪰B if and only if u(A)≥u(B).
    C2. Further, we require u is quasi-concave.
    C3. Further, we require u is continuous.
    C4. Further, we require u is concave.
    C5. Further, we require u is linear.
    The standard VNM utility theorem can be thought of as saying A1, A2, A3, and B1 together imply C5.
    Here is the main question I am curious about:
    Q1: Do A1, A2, A3, A4, and A5 together imply C4? [ANSWER: NO]
    (If no, how can we salvage C4, by adding or changing some axioms?)
    Here are some sub-questions that would constitute significant partial progress, and that I think are interesting in their own right:
    Q2: Do A1, A2, A3, and A4 together imply C3? [ANSWER: NO]
    Q3: Do C3 and A5 together imply C4? [ANSWER: NO]
    (Feel free to give answers that are only partial progress, and use this space to think out loud or discuss anything else related to weaker versions of VNM.)
    EDIT:
    AlexMennen actually resolved the question in the negative as stated, but my curiosity is not resolved, since his argument is violating continuity, and I really care about concavity. My updated main question is now:
    Q4: Do A1, A2, A3, A4, and A5 together imply that there exists a concave function u:L[0,1] such that A⪰B if and only if u(A)≥u(B)? [ANSWER: NO]
    (i.e. We do not require u to be continuous.)
    This modification also implies interest in the subquestion:
    Q5: Do A1, A2, A3, and A4 together imply C2?
    EDIT 2:
    Here is another bonus axiom:
    B2. For all A,B∈L, if A≻B, then there exists some C∈L such that A≻C≻B.
    (Really, we don't need to assume C is already in L. We just need it to be possible to add a C, and extend our preferences in a way that satisfies the other axioms, and A3 will imply that such a lottery was already in L. We might want to replace this with a cleaner axiom later.)
    Q6: Do A1, A2, A3, A5, and B2 together imply C4? [ANSWER: NO]
    EDIT 3:
    We now have negative answers to everything other than Q5, which I still think is pretty interesting. We could also weaken Q5 to include other axioms, like A5 and B2. Weakening the conclusion doesn't help, since it is easy to get C2 from C1 and A4.
    I would still really like some axioms that get us all the way to a concave function, but I doubt there will be any simple ones. Concavity feels like it really needs more structure that does not translate well to a preference relation.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    5 min
  • EA - "Risk Awareness Moments" (Rams): A concept for thinking about AI governance interventions by oeg
    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: "Risk Awareness Moments" (Rams): A concept for thinking about AI governance interventions, published by oeg on April 14, 2023 on The Effective Altruism Forum.
    In this post, I introduce the concept of Risk Awareness Moments (“Rams”): “A point in time, after which concern about extreme risks from AI is so high among the relevant audiences that extreme measures to reduce these risks become possible, though not inevitable.” This is a blog post, not a research report, meaning it was produced quickly and is not to Rethink Priorities’ typical standards of substantiveness and careful checking for accuracy.
    Summary
    I give several examples of what a Ram might look like for national elites and/or the general population of a major country. Causes could include failures of AI systems, or more social phenomena, such as new books being published about AI risk.
    I compare the Ram concept to similar concepts such as warning shots. I see two main benefits: (1) Rams let us remain agnostic about what types of evidence make people concerned, e.g., something that AI does, vs. social phenomena; (2) the concept lets us remain agnostic about the “trajectory” to people being concerned about the risk, e.g., whether there is a more discrete/continuous/lumpy change in opinion.
    For many audiences, and potential ways in which AI progress could play out, there will not necessarily be a Ram. For example, there might be a fast takeoff before the general public has a chance to significantly alter their beliefs about AI.
    We could do things to increase the likelihood of Rams, or to accelerate their occurrence. That said, there are complex considerations about whether actions to cause (earlier) Rams would be net positive.
    A Ram - even among influential audiences - is not sufficient for adequate risk-reduction measures to be put in place. For example, there could be bargaining failures between countries that make it impossible to get mutually beneficial AI safety agreements. Or people who are more aware of the risks from transformative AI might also be more aware of the benefits, and thus make an informed decision that the benefits are worth the risks by their lights.
    At the end, I give some historical examples of Rams for issues other than AI risk.
    From DALL-E 2
    Definition
    I define a Risk Awareness Moment (Ram) as “a point in time, after which concern about extreme risks from AI is so high among the relevant audiences that extreme measures to reduce these risks become possible, though not inevitable.”
    Extreme risks refers to risks at least at the level of global catastrophic risks (GCRs). I intend the term to capture accident, misuse, and structural risks.
    Note that people could be concerned about some extreme risks from AI without being concerned about other risks. For example, the general public might become worried about risks from non-robust narrow AI in nuclear weapons systems, without being worried about misaligned AGI. Concern about one risk would not necessarily make it possible to get measures that would be helpful for tackling other risks.
    Additionally, some audiences might have unreasonable threat models. One possible example of this would be an incorrect belief that accidents with Lethal Autonomous Weapons would themselves cause GCR-level damage. Similar to the bullet point above, this belief might be necessary for measures to tackle the specific (potentially overblown) threat model, without necessarily being helpful for measures to tackle other risks.
    Relevant audiences will differ according to the measure in question. For example, measures carried out by labs might require people in labs to be widely convinced. In contrast, government-led measures might require people in specific parts of the government - and maybe also the public - to be convinced.
    Extreme measures could include national-le...
    20 min
  • EA - Paths to reducing rodenticide use in the U.S. by Holly Elmore
    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: Paths to reducing rodenticide use in the U.S., published by Holly Elmore on April 15, 2023 on The Effective Altruism Forum.
    This is the third report in a sequence of reports on reducing the use of rodenticide poisons. It is not necessary to have read the previous reports to read this one, although this report will take for granted conclusions that were discussed and weighed in depth in the previous reports. Part 1 describes why rodenticides are crueler to rodents and more dangerous for human children, pets, and wildlife than most alternative methods for removing rodents from homes, businesses, and farms. Part 2 explains why both consumers and businesses have strong incentives to continue relying on rodenticides. This report describes and ranks interventions to reduce rodenticide use in the U.S. according to their expected impact, neglectedness, and tractability. We leave aside how the interventions we discuss might change attitudes toward pest populations in the long term due to a lack of relevant evidence.
    The report is grouped into sections by the class of intervention: Legislative interventions, Information campaigns, Technological disruption, and Funding research. We are pessimistic about the legislative interventions that severely restrict legal rodenticide use. For example, California's recent ban on second-generation anticoagulant rodenticides (SGARs) is riddled with exemptions and may increase the use of alternatives that are even crueler to rodents. However, there are less controversial reforms, such as sanitation reform, that would reduce some rodenticide use, especially at the local level.
    Our top recommended intervention is investing in improved rodent birth control. [part 1] of this sequence was enthusiastic about existing EPA-approved rodent birth control ContraPest, but in part 2 we reported additional findings that led us to conclude that ContraPest is too expensive and cumbersome to replace the role that rodenticides currently play. Birth control baits that are cheaper and more versatile than Contrapest could replace rodenticide in many (though probably not all) situations.
    Our runner-up recommendation is to run digital information campaigns to educate the public on the costs and dangers of rodenticides. Digital advertising is cheap, and can quickly reach millions of people without having to first develop personal relationships with voters. Although the relatively grassroots approach of extant anti-rodenticide activism may be a sign that more impersonal approaches would not work, there is value in testing how much can be accomplished through mass communication alone.
    Table 1: A summary of the overall ranking of the 10 interventions types analyzed above. We scored Impact, Neglectedness, and Tractability holistically using High (5), Medium-High (4), Medium (3), Medium-Low (2), and Low (1), which were converted to a 5-point scale. The rankings in the right-most column are a result of adding up the numeric values that Intervention Type received in Impact, Neglectedness, and Tractability. Ties were resolved by holistically considering which intervention we believed was overall higher priority.
    We acknowledge that some interventions that do not look promising on their own may increase the tractability of more promising interventions. For example, obtaining local- and state-level bans may be a hassle and the results may be imperfect, but legal pressure to find alternatives may spur investment in new technology that is both more effective and humane than rodenticides. Readers may also have a personal advantage in implementing certain interventions and therefore may want to prioritize implementing those even if other interventions are more highly ranked in this report.
    This research is a project of Rethink Priorities. It was written by Holly Elmore. If you’re...
    4 min
  • EA - Announcing Innovate Animal Ag (like GFI but for Animal Welfare Tech) by RobertY
    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: Announcing Innovate Animal Ag (like GFI but for Animal Welfare Tech), published by RobertY on April 14, 2023 on The Effective Altruism Forum.
    I’m very excited to announce the launch of Innovate Animal Ag, a new nonprofit whose mission is to support agricultural technologies that directly improve animal welfare. Our first focus is on helping introduce in-ovo egg sexing technologies into the US. We’re currently hiring and fundraising!
    The problem
    Companies improve their treatment of animals when the cost of the status quo is higher than the cost of change. Historically, pressure from activists, consumers, and regulators has been effective at increasing the cost of the inhumane practices that are part of the status quo. But significantly less work has been done on the other side of the equation: decreasing the cost of change.
    Similarly to how The Good Food Institute, New Harvest, and The Material Innovation Initiative support the field of alternative proteins, Innovate Animal Ag supports the field of animal welfare technology. Through education, networking building, and other ecosystem-level interventions, we aim to make it as easy as possible for companies to adopt new technologies that improve animal welfare.
    The technologies that we focus on will be guided by the EA principles of importance, tractability and neglectedness. Based on these criteria, some classes of technology we could be interested in include humane seafood slaughter machines, humane poultry euthanasia techniques, humane pest control technologies, and in-ovo egg sexing.
    Initial focus: in-ovo egg sexing
    Most of our work in the short term will be spent on in-ovo egg sexing technologies, specifically on helping introduce them to the US market. We chose this focus because it addresses an important problem that’s particularly tractable right now: The culling of male chicks in the egg industry. We hope to use this as a test-case for our overall approach.
    For each of the more than 6 billion egg-laying hens in the world, there was a male chick that was fertilized, incubated, hatched, manually identified by a human, sorted, and then immediately killed. This practice is inhumane, wasteful for the industry, and unpopular with the consumers that know about it.
    Fortunately, there are a number of solutions being developed by companies and labs around the world. Some are even in the early stages of commercialization in Europe, where governments have started to ban the practice of chick culling. This rollout is going better than many realize: Extrapolating from publicly available partnership announcements, we estimate that current in-ovo sexing companies already supply between 10 and 20% of the entire EU market, with a cost impact of 1–3 euro cents per table egg (similar to, if not lower than cage-free). This leads us to believe that current technology is already ready for the high-end specialty egg market in the US. A small market foothold at the high end is then the first step towards wider adoption. Once companies demonstrate that there is demand for these products, they can more easily invest in lowering costs and scaling.
    Eventually the goal is to fulfill The United Egg Producers promise to completely eliminate chick culling across the industry.
    In our conversations with the companies developing this technology, a common refrain is that they’re interested in the US market, but have little direct engagement with the US egg industry. For the most part, companies are focused on Europe because that's where governments are banning chick culling.
    Jumpstarting the market in the US will be challenging. Producers may not be aware that this technology is ready, and if they are, they may not be confident that consumers will be willing to pay a price premium. We aim to help solve these problems through interventions such as:...
    8 min
  • LW - List of slowdown/halt AI requests by Cleo Nardo
    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: List of slowdown/halt AI requests, published by Cleo Nardo on April 14, 2023 on LessWrong.
    About this document
    There has been a recent flurry of letters/articles/statements/videos which endorse a slowdown or halt of colossal AI experiments via (e.g.) regulation or coordination.
    This document aspires to collect all examples into a single centralised list. I'm undecided on how best to order and subdivide the examples, but I'm open to suggestions.
    As a disclaimer, this list is...
    Living — I'll try to update the list over time.
    Non-exhaustive — There are almost certainly examples I've missed.
    Non-representative — The list is biased, at least initially, towards things that I have been shown personally.
    Please mention in the comments any examples I've missed so I can add them!
    List of slowdown/halt AI requests
    Last updated: April 14th 2023.
    (Note that I'm also including surveys.)
    Pause Giant AI Experiments: An Open Letterby Future of Life Institute
    Pausing AI Developments Isn't Enough. We Need to Shut it All Downby Eliezer Yudkowsky
    We must slow down the race to God-like AIby Ian Hogarth
    The case for slowing down AI by Sigal Samuel
    The Case for Halting AI Developmentby Max Tegmark, Lex Fridman
    Lennart Heim on Compute Governanceby Lennart Heim, Future of Life Institute
    Let’s think about slowing down AIby KatjaGrace
    The 0.2 OOMs/year targetby Cleo Nardo
    AI Summer Harvestby Cleo Nardo
    Instead of technical research, more people should focus on buying timeby Akash, Olivia Jimenez, Thomas Larsen
    Slowing down AI progress is an underexplored alignment strategyby Michael Huang
    Slowing Down AI: Rationales, Proposals, and Difficulties by Simeon Campos, Henry Papadatos, Charles M
    In the Matter of OpenAI (FTC 2023) by Center for AI and Digital Policy
    Dangers of AI and the End of Human Civilizationby Eliezer Yudkowsky, Lex Fridman
    We’re All Gonna Die with Eliezer Yudkowskyby Eliezer Yudkowsky, Bankless
    The public supports regulating AI for safetyby Zach Stein-Perlman
    New survey: 46% of Americans are concerned about extinction from AI; 69% support a six-month pause in AI developmentby Akash
    Credit to Zach Stein-Perlman.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • AF - Concave Utility Question by Scott Garrabrant
    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: Concave Utility Question, published by Scott Garrabrant on April 15, 2023 on The AI Alignment Forum.
    This post will just be a concrete math question. I am interested in this question because I have recently come tor reject the independence axiom of VNM, and am thus playing with some weaker versions.
    Let Ω be a finite set of deterministic outcomes. Let L be the space of all lotteries over these outcomes, and let ⪰ be a relation on L. We write A∼B if A ⪰ B and B ⪰ A. We write A≻B if A⪰B but not A∼B.
    Here are some axioms we can assume about ⪰:
    A1. For all A,B∈L, either A⪰B or B⪰A (or both).
    A2. For all A,B,C∈L, if A⪰B, and B⪰C, then A⪰C.
    A3. For all A,B,C∈L, if A⪰B, and B⪰C, then there exists a p∈[0,1] such that B∼pA+(1−p)C.
    A4. For all A,B∈L, and p∈[0,1] if A⪰B, then pA+(1−p)B⪰B.
    A5. For all A,B∈L, and p∈[0,1], if p>0 and B⪰pA+(1−p)B, then B⪰A.
    Here is one bonus axiom:
    B1. For all A,B,C∈L, and p∈[0,1], A⪰B if and only if pA+(1−p)C⪰pB+(1−p)C.
    (Note that B1 is stronger than both A4 and A5)
    Finally, here are some conclusions of successively increasing strength:
    C1. There exists a function u:L[0,1] such that A⪰B if and only if u(A)≥u(B).
    C2. Further, we require u is quasi-concave.
    C3. Further, we require u is continuous.
    C4. Further, we require u is concave.
    C5. Further, we require u is linear.
    The standard VNM utility theorem can be thought of as saying A1, A2, A3, and B1 together imply C5.
    Here is the main question I am curious about:
    Q1: Do A1, A2, A3, A4, and A5 together imply C4? [ANSWER: NO]
    (If no, how can we salvage C4, by adding or changing some axioms?)
    Here are some sub-questions that would constitute significant partial progress, and that I think are interesting in their own right:
    Q2: Do A1, A2, A3, and A4 together imply C3? [ANSWER: NO]
    Q3: Do C3 and A5 together imply C4? [ANSWER: NO]
    (Feel free to give answers that are only partial progress, and use this space to think out loud or discuss anything else related to weaker versions of VNM.)
    EDIT:
    AlexMennen actually resolved the question in the negative as stated, but my curiosity is not resolved, since his argument is violating continuity, and I really care about concavity. My updated main question is now:
    Q4: Do A1, A2, A3, A4, and A5 together imply that there exists a concave function u:L[0,1] such that A⪰B if and only if u(A)≥u(B)? [ANSWER: NO]
    (i.e. We do not require u to be continuous.)
    This modification also implies interest in the subquestion:
    Q5: Do A1, A2, A3, and A4 together imply C2?
    EDIT 2:
    Here is another bonus axiom:
    B2. For all A,B∈L, if A≻B, then there exists some C∈L such that A≻C≻B.
    (Really, we don't need to assume C is already in L. We just need it to be possible to add a C, and extend our preferences in a way that satisfies the other axioms, and A3 will imply that such a lottery was already in L. We might want to replace this with a cleaner axiom later.)
    Q6: Do A1, A2, A3, A5, and B2 together imply C4? [ANSWER: NO]
    EDIT 3:
    We now have negative answers to everything other than Q5, which I still think is pretty interesting. We could also weaken Q5 to include other axioms, like A5 and B2. Weakening the conclusion doesn't help, since it is easy to get C2 from C1 and A4.
    I would still really like some axioms that get us all the way to a concave function, but I doubt there will be any simple ones. Concavity feels like it really needs more structure that does not translate well to a preference relation.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    5 min
  • EA - Healthier Hens Y1.5 Update and scaledown assessment by lukasj10
    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: Healthier Hens Y1.5 Update and scaledown assessment, published by lukasj10 on April 14, 2023 on The Effective Altruism Forum.
    TL;DR
    Healthier Hens (HH) has scaled down due to not being able to secure enough funding to provide a sufficient runway to pilot dietary interventions effectively. We will continue through mini-projects and refining our plan for a feed pilot on the ground until our next organisational assessment at the end of summer 2023. Most efforts will now be spent on reporting, dissemination and fundraising. In this post we share updates, show what went well, less so and what others can learn from our attempts.
    Our mission and updated approach
    Keel bone fractures (KBF) are the second biggest source of hens’ suffering after behavioural deprivation related to cages, and the biggest in cage-free systems. Our mission remains to reduce the suffering of hens by addressing this source of pain. Numerous studies indicate that dietary interventions can reduce bone fractures. Our goal is to find ways to help hens get adequate nutrition and experience less pain. Besides research and piloting activities, we are doing that by outreach and collaboration with Kenyan cage-free egg farming stakeholders (including farmers, feed mills, universities and regulators) to improve on-farm hen welfare. Please read our introductory, 6M and 1Y update posts to learn more about the background and progress of HH. Due to funding constraints, we have downscaled and now focus on building capacity through mini-projects before piloting an intervention on the ground.
    HH Y1.5 update
    We decided to scale down and as of the first of March 2023, HH is a volunteer-led organisation. We will have our next evaluation point at the end of summer 2023.
    Despite major delays brought on by the presidential elections in Kenya, HH is finally a registered entity in both the US and our country of pilot operations.
    We carried out two additional farmer workshops in Murang’a and Nakuru counties. A report outlining this work will be published in May 2023. The workshops revealed significant knowledge gaps coupled with expressed interest by the farmers to learn and improve. To retain engagement, we began providing free resources and formed a WhatsApp group to stay connected with motivated cage-free farmers.
    Cage-free farmer workshop in Nakuru county, October, 2022.
    Cage-free coalition strategic session in Naivasha in February, 2023.
    We have published the third volume of our hen feed fortification literature review. This edition focused on nutrient level recommendations and the potential of other additives to improve bone health. We have also published a report outlining our feed sampling findings. It confirms potential issues with Kenyan feed quality and compositional consistency.
    We are engaged in the formation of a regional cage-free coalition aimed at accelerating transition campaigns, led by African Network for Animal Welfare.
    We surveyed our staff in February. We generally felt positive about the organization's culture and values. The survey also reveals that we were generally satisfied with our team and work, but were concerned about the impact of the recent decision to downscale. We also felt that there was a lack of critical feedback and constant communication within the team.
    Our collaboration with the University of Bern indicates negative results of a split-feeding intervention. While we expected a decrease in KBFs without compromising egg quality, surprisingly, poorer egg quality, digestive issues and neutral-to-negative effect on KBFs are observed.
    To explore alternative fundraising avenues, we looked into several social enterprise concepts, where hen welfare benefits could be attained while also improving the livelihoods of African cage-free farmers. The chosen concept (a hybridised adaptation of the wor...
    21 min