The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - EAs interested in EU policy: Consider applying for the European Commission’s Blue Book Traineeship by EU Policy Careers
    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: EAs interested in EU policy: Consider applying for the European Commission’s Blue Book Traineeship, published by EU Policy Careers on January 6, 2023 on The Effective Altruism Forum.
    Inspired by a series of recent forum posts highlighting early career opportunities in US policy, this post summarises why and how to apply to the Blue Book Traineeship. This paid, five-month internship programme with the European Commission, the executive body of the EU, is one of the main pathways into an EU policy career. The last section of this post also outlines some other options to get started in EU policy.
    There are two Blue Book sessions each year, with applications opening in January for the session starting in October and in August for a start in March of the following year. Application deadlines can be found here. Applications for the October 2023 session are open now and close on January 31.
    As the programme is suitable for people in different stages of their career and from various backgrounds (see below) and EU policy is arguably still neglected within the EA community, an application could be a good option for many EAs. The initial application is fairly low-cost, as you only need to upload your CV and documents without writing a motivation letter.
    The programme is not only relevant to students or recent graduates, as many trainees have some years of previous work experience (around 30% of all trainees are 30+ and only 5% younger than 25, see full statistics here). Work experience can even be a significant advantage for finding full-time positions after the traineeship, and it can be a strategic decision to only start the traineeship after gaining some work experience to increase the chances of being able to stay on.[1]
    Epistemic status: This post is mostly based on my experience completing the traineeship last year and now working full time at the Commission, including conversations with around 20 people before and during the traineeship about both the application process and getting full-time employment afterwards. The post was greatly improved by the contributions of four other EAs with expertise on EU policy.
    Eligibility
    The programme is mostly directed towards EU citizens.[2] The minimum educational requirement is a completed undergraduate degree. However, a master’s degree is sometimes necessary to pass through the first stage of the selection process (especially for ‘competitive’ nationalities below) and increases employment opportunities after the traineeship.
    The programme is open to graduates of all disciplines, not just people holding policy-related degrees—even rewarding applicants from ‘rare fields of study’ in the selection process. Most degrees outside of policy, law and economics should fall into this, as the majority of trainees (around 70%) hold degrees in one of these fields. It is therefore a good opportunity for EAs with no previous policy experience interested in testing their fit and learning more about impact in the sector.
    All eligibility requirements are detailed here, including the requirement to prove very good knowledge (B2 level) of two official EU languages (English plus one other is sufficient).
    Background on the European Commission
    The Commission is the executive body of the EU. It draws up initial legislative proposals (which are then amended and adopted jointly by the European Parliament and the Council of the EU, the other two main EU institutions) and implements EU policies (e.g., deciding how to allocate funds, outlining the technical details of legislation, monitoring the implementation at member state level). It employs around 30,000 people mostly based in Brussels. The Commission’s main departments according to policy areas, the Directorate Generals (DGs), are comparable to government ministries at the national level. A Blue Book Tr...
    31 min
  • LW - The Limit of Language Models by DragonGod
    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: The Limit of Language Models, published by DragonGod on January 6, 2023 on LessWrong.
    Epistemic Status
    Highlighting a thesis in Janus' "Simulators" that I think is insufficiently appreciated.
    Thesis
    In the limit, models optimised for minimising predictive loss on humanity's text corpus converge towards general intelligence.
    Preamble
    From Janus' Simulators:
    Something which can predict everything all the time is more formidable than any demonstrator it predicts: the upper bound of what can be learned from a dataset is not the most capable trajectory, but the conditional structure of the universe implicated by their sum (though it may not be trivial to extract that knowledge).
    Introduction
    I affectionately refer to the above quote as the "simulators thesis". Reading and internalising that passage was an "aha!" moment for me. I was already aware (at latest July 2020) that language models were modelling reality. I was persuaded by arguments of the below form:
    Premise 1: Modelling is transitive. If X models Y and Y models Z, then X models Z.
    Premise 2: Language models reality. "Dogs are mammals" occurs more frequently in text than "dogs are reptiles" because dogs are in actuality mammals and not reptiles. This statistical regularity in text corresponds to a feature of the real world. Language is thus a map (albeit flawed) of the external world.
    Premise 3: GPT-3 models language. This is how it works to predict text.
    Conclusion: GPT-3 models the external world.
    But I hadn't yet fully internalised all the implications of what it means to model language and hence our underlying reality. The limit that optimisation for minimising predictive loss on humanity's text corpus will converge to. I belatedly make those updates.
    Interlude: The Requisite Capabilities for Language Modelling
    Janus again:
    If loss keeps going down on the test set, in the limit – putting aside whether the current paradigm can approach it – the model must be learning to interpret and predict all patterns represented in language, including common-sense reasoning, goal-directed optimization, and deployment of the sum of recorded human knowledge.
    Its outputs would behave as intelligent entities in their own right. You could converse with it by alternately generating and adding your responses to its prompt, and it would pass the Turing test. In fact, you could condition it to generate interactive and autonomous versions of any real or fictional person who has been recorded in the training corpus or even could be recorded (in the sense that the record counterfactually “could be” in the test set).
    Implications
    The limit of predicting text is predicting the underlying processes that generated said text. If said underlying processes are agents, then sufficiently capable language models can predict agent (e.g., human) behaviour to arbitrary fidelity. If it turns out to be the case that the most efficient way of predicting the behaviour of conscious entities (as discriminated via text records) is to instantiate conscious simulacra, then such models may perpetuate mindcrime.
    Furthermore, the underlying processes that generate text aren't just humans, but the world which we inhabit. That is, a significant fraction of humanity's text corpus reports on empirical features of our external environment or the underlying structure of reality:
    Timestamps
    And other empirical measurements
    Log files
    Database files
    Including CSVs and similar
    Experiment records
    Research findings
    Academic journals in quantitative fields
    Other reports
    Etc.
    Moreover, such text is often clearly distinguished from other kinds of text (fiction, opinion pieces, etc.) via its structure, formatting, titles, etc. In the limit of minimising predictive loss on such text, language models must learn the underlying processes that generated them — the conditional str...
    10 min
  • EA - Metaculus Beginner Tournament for New Forecasters by Anastasia
    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: Metaculus Beginner Tournament for New Forecasters, published by Anastasia on January 6, 2023 on The Effective Altruism Forum.
    Are you new to forecasting? Join the latest version of the Metaculus Beginner Tournament! The tournament runs for the first quarter of 2023. Each week, we'll have two new questions on relevant topics that week. And the two questions from the previous week will resolve, so you can see how you did on them! At the end of the quarter, you'll be a veteran forecaster with a real forecasting track record.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min
  • LW - Childhood Roundup #1 by Zvi
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Childhood Roundup #1, published by Zvi on January 6, 2023 on LessWrong.
    I have a few shorter more focused posts in the works, including my practical short version of On Bounded Distrust. In the meantime, it makes sense to continue to clean out the backlog of roundup style things.
    Let Your Children Play
    We have gone insane on this.
    Sane 2022 parents of 10-year-olds: I would like to let you go outside without me. I am terrified that someone will call the cops and they will take you away from me.
    That is a thing now. As in parents being thrown in jail for letting their eight year old child walk home from school on their own.
    As they stood on her porch, the officers told Wallace that her son could have been kidnapped and sex trafficked. “‘You don’t see much sex trafficking where you are, but where I patrol in downtown Waco, we do,'” said one of the cops, according to Wallace.
    Did things get more dangerous since 1980, when we were mostly sane about this? No. They got vastly less dangerous, in all ways other than the risk of someone calling the cops.
    The numbers on ‘sex trafficking’ and kidnapping by strangers are damn near zero.
    The incident caused ‘ruin your life’ levels of damage.
    Child services had the family agree to a safety plan, which meant Wallace and her husband could not be alone with their kids for even a second. Their mothers—the children’s grandmothers—had to visit and trade off overnight stays in order to guarantee the parents were constantly supervised. After two weeks, child services closed Wallace’s case, finding the complaint was unfounded.
    Wallace’s sister has started a GoFundMe for her. She is in debt after losing her job and paying for the lawyer and the diversion program. She also hopes to hire a lawyer to get her record expunged so that she can work with kids again.
    One of my closest friends here in New York is strongly considering moving to the middle of nowhere so that his child will be able to walk around outside, because it is not legally safe to do that anywhere there are people.
    Claim that UNICEF thinks Dutch children are the happiest in the world and that this is because they bike everywhere, things are designed to make this work, and they therefore have freedom to move around.
    Childless World
    This is also related to the thing where a lot of places give off the vibe that life does not involve children, and children are a weird, shameful and irresponsible lifestyle choice.
    I sense this type of attitude ebb and flow in power as I move between places.
    School Daze
    It’s not like the kids were attending school that much anyway (MR). The chart is San Francisco, but it is far from unique in this. This headline says ‘41% of NYC students were chronically absent’ last year, versus a historical expectation of ~25%.
    Note that in school parlance, ‘chronic absentee’ means you missed 18 days of school out of 180. Thus, someone who reports as demanded 90% of the time is still ‘chronically absent,’ no matter the cause.
    Public school enrollment continues to be 3% below pre-pandemic levels. Many responded with ‘at least some good has come of this.’
    The degree of learning in schools, even high end ones, leaves something to be desired.
    I simply don’t know how to update far enough in the appropriate directions.
    A paper compared students answers on homework questions to their later answers to the same question posed on an exam. They found that in the past this correlation was high, and said this meant that the homework helped students. One obvious alternative explanation is that knowing X is correlated with knowing X.
    None of this is ‘benefit.’ Then, the study notes that this correlation has gone down, due to a large group of students who don’t show this pattern – they get the homework right by copying it, then get the exam question wrong anyway.
    Via MR, paper from ...
    13 min
  • AF - Categorizing failures as “outer” or “inner” misalignment is often confused by Rohin Shah
    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: Categorizing failures as “outer” or “inner” misalignment is often confused, published by Rohin Shah on January 6, 2023 on The AI Alignment Forum.
    Pop quiz: Are the following failures examples of outer or inner misalignment? Or is it ambiguous?
    (I’ve spoilered the scenarios so you look at each one separately and answer before moving on to the next. Even if you don’t want to actually answer, you should read through the scenarios for the rest of the post to make sense.)
    Scenario 1: We train an AI system that writes pull requests based on natural language dialogues where we provide specifications. Human programmers check whether the pull request is useful, and if so the pull request is merged and the AI is rewarded. During training the humans always correctly evaluate whether the pull request is useful. At deployment, the AI constructs human farms where humans are forced to spend all day merging pull requests that now just add useless print statements to the codebase.
    Scenario 2: We train an AI system to write pull requests as before. The humans providing feedback still correctly evaluate whether the pull request is useful, but it turns out that during training the AI system noticed that it was in training, and so specifically wrote useful pull requests in order to trick the humans into thinking that it was helpful, when it was already planning to take over the world and build human farms.
    I posed this quiz to a group of 10+ full-time alignment researchers a few months ago, and a majority thought that Scenario 1 was outer misalignment, while almost everyone thought that Scenario 2 was inner misalignment.
    But that can’t be correct, since Scenario 2 is just Scenario 1 with more information! If you think that Scenario 2 is clearly inner misalignment, then surely Scenario 1 has to be either ambiguous or inner misalignment! (See the appendix for hypotheses about what’s going on here.)
    I claim that most researchers’ intuitive judgments about outer and inner misalignment are confused. So what exactly do we actually mean by “outer” and “inner” misalignment? Is it sensible to talk about separate “outer” and “inner” misalignment problems, or is that just a confused question?
    I think it is misguided to categorize failures as “outer” or “inner” misalignment. Instead, I think “outer” and “inner” alignment are best thought of as one particular way of structuring a plan for building aligned AI.
    I’ll first discuss some possible ways that you could try to categorize failures, and why I’m not a fan of them, and then discuss outer and inner alignment as parts of a plan for building aligned AI.
    Possible categorizations
    Objective-based. This categorization is based on distinctions between specifications or objectives at different points in the overall system. This post identifies three different notions of specifications or objectives:
    Ideal objective (“wishes”): The hypothetical objective that describes good behavior that the designers have in mind.
    Design objective (“blueprint”): The objective that is actually used to build the AI system. This is the designer's attempt to formalize the ideal objective.
    Revealed objective (“behavior”): The objective that best describes what actually happens.
    On an objective-based categorization, outer misalignment would be a discrepancy between the ideal and design objectives, while inner misalignment is a discrepancy between the design and revealed objectives.
    (The mesa-optimizers paper has a similar breakdown, except the third objective is a structural objective, rather than a behavioral one. This is also the same as Richard’s Framing 1.)
    With this operationalization, it is not clear which of the two categories a given situation falls under, even when you know exactly what happened. In our scenarios above, what exactly is the design objective? Is it “how ...
    13 min
  • AF - Definitions of “objective” should be Probable and Predictive by Rohin Shah
    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: Definitions of “objective” should be Probable and Predictive, published by Rohin Shah on January 6, 2023 on The AI Alignment Forum.
    Introduction
    Core arguments about existential risk from AI misalignment often reason about AI “objectives” to make claims about how they will behave in novel situations. I often find these arguments plausible but not rock solid because it doesn’t seem like there is a notion of “objective” that makes the argument clearly valid.
    Two examples of these core arguments:
    AI risk from power-seeking. This is often some variant of “because the AI system is pursuing an undesired objective, it will seek power in order to accomplish its goal, which causes human extinction”. For example, “The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.” This is a prediction about a novel situation, since “causing human extinction” is something that only happens at most once.
    AI optimism. This is often some variant of “we will use human feedback to train the AI system to help humans, and so it will learn to pursue the objective of helping humans.” Implicitly, this is a prediction about what AI systems do in novel situations; for example, it is a prediction that once the AI system has enough power to take over the world, it will continue to help humans rather than execute a treacherous turn.
    When we imagine powerful AI systems built out of large neural networks, I’m often somewhat skeptical of these arguments, because I don’t see a notion of “objective” that can be confidently claimed is:
    Probable: there is a good argument that the systems we build will have an “objective”, and
    Predictive: If I know that a system has an “objective”, and I know its behavior on a limited set of training data, I can predict significant aspects of the system’s behavior in novel situations (e.g. whether it will execute a treacherous turn once it has the ability to do so successfully).
    Note that in both cases, I find the stories plausible, but they do not seem strong enough to warrant confidence, because of the lack of a notion of “objective” with these two properties. In the case of AI risk, this is sufficient to justify “people should be working on AI alignment”; I don’t think it is sufficient to justify “if we don’t work on AI alignment we’re doomed”.
    The core difficulty is that we do not currently understand deep learning well enough to predict how future systems will generalize to novel circumstances. So, when choosing a notion of “objective”, you either get to choose a notion that we currently expect to hold true of future deep learning systems (Probable), or you get to choose a notion that would allow you to predict behavior in novel situations (Predictive), but not both.
    This post is split into two parts. In the first part, I’ll briefly gesture at arguments that make predictions about generalization behavior directly (i.e. without reference to “objectives”), and why they don’t make me confident about how future systems will generalize. In the second part, I’ll demonstrate how various notions of “objective” don’t seem simultaneously Probable and Predictive.
    Part 1: We can’t currently confidently predict how future systems will generalize
    Note that this is about what we can currently say about future generalization. I would not be shocked if in the future we could confidently predict how the future AGI systems will generalize.
    My core reasons for believing that predicting generalization is hard are that:
    We can’t predict how current systems will generalize to novel situations (of similar novelty to the situations that would be encountered when deliberately causing an existential catastrophe)
    There are a ridiculously huge number of possible programs, including a huge number of possible programs that are consistent with a ...
    21 min
  • AF - 200 COP in MI: Techniques, Tooling and Automation by Neel Nanda
    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: 200 COP in MI: Techniques, Tooling and Automation, published by Neel Nanda on January 6, 2023 on The AI Alignment Forum.
    This is the seventh post in a sequence called 200 Concrete Open Problems in Mechanistic Interpretability. Start here, then read in any order. If you want to learn the basics before you think about open problems, check out my post on getting started. Look up jargon in my Mechanistic Interpretability Explainer
    I’ll make another post every 1-2 days, giving a new category of open problems. If you want to read ahead, check out the draft sequence here!
    Motivating papers: Causal Scrubbing, Logit Lens
    Motivation
    In Mechanistic Interpretability, the core goal is to form true beliefs about what’s going on inside a network. The search space of possible circuits is extremely large, and even once a circuit is found, we need to verify that that’s what’s really going on. These are hard problems, and having good techniques and tooling is essential to making progress. This is particularly important in mech interp because it’s such a young field that there isn’t an established toolkit and standard of evidence, and each paper seems to use somewhat different and ad-hoc techniques (pre-paradigmatic, in Thomas Kuhn’s language).,
    Getting better at this is particularly important to enable us to get traction interpreting circuits at all, even in a one layer toy language model! But it’s particularly important for dealing with the problem of scale. Mech interp can be very labour intensive, and involve a lot of creativity and well honed research intuitions. This isn’t the end of the world with small models, but ultimately we want to understand models with hundreds of billions to trillions of parameters! We want to be able to leverage researcher time as much as possible, and the holy grail is to eventually automate the finding of circuits and understanding models. My guess is that the most realistic path to really understanding superhuman systems is to slowly automate more and more of the work with weaker systems, while making sure that we understand those systems, and that they’re aligned with what we want.
    This can be somewhat abstract, so here are my best guesses for what progress could look like:
    Refining understanding: In what contexts are current techniques most useful, and where are they misleading? How can we notice misapplications? Which techniques fail in the same way?
    Eg, what’s up with backup heads, which take over when the main head is ablated
    How to find circuits?: What are the right mindsets and approaches for finding a novel circuit? Common traps? What tools are best to apply in an unfamiliar situation, and how to interpret their output?
    I expect this to look like a mix of a refined understanding of current techniques, building great infrastructure, and building practical intuitions and experience.
    Building a better toolkit: What kind of new, general techniques are out there for really understanding model internals? One of the joys of mech interp is that we have full control over the model’s internals, and can edit whatever weights and activations we want. I expect there’s many ways to use this that no one’s tried!
    Causal tracing/activation patching from ROME is a great example of this! It’s such an elegant, powerful and generally applicable technique, that I’d just never thought of before
    Gold standards of evidence: What does it mean to have truly understood a circuit? Are there generally applicable approaches
    Redwood’s Causal Scrubbing is a solid attempt here, and I’m excited to see how well it works in practice
    Good infrastructure: The right software and tooling can massively increase the rate of being able to do research - if common operations can be done in a few lines of code and run quickly, you can focus on actually doing research.
    My TransformerLens li...
    25 min
  • EA - Foundation Entrepreneurship - How the first training program went by Aidan 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: Foundation Entrepreneurship - How the first training program went, published by Aidan Alexander on January 6, 2023 on The Effective Altruism Forum.
    TLDR:
    Charity Entrepreneurship (CE) has run our first foundation training program. This program looked less like traditional philanthropic advising, and more like our classic charity incubation program. Instead of recommending specific charities, we aimed to teach skills that would allow foundation leaders to work as informed, independent, full-time grantmakers, equipped with a suite of decision-making tools. Despite seeing many areas where we could improve the program, we consider the pilot to have been very successful. We are excited to run this program again and believe it could provide impact comparable to that of our charity incubation program.
    Background
    At the beginning of 2022 we announced that CE would be running a pilot training program for grantmaking foundations. The pilot program finished several months ago, and so we wanted to follow up on how it went and what we learned.
    Why
    Starting an impactful foundation and being a strong grantmaker is hard. Few high-quality resources exist (at least in a publicly available form) that teach how to run an evidence-based and highly effective foundation.
    Over the past four years, CE has helped launch 23 new nonprofits, several of which are on track to becoming field leaders in their respective areas (e.g., recommended by GiveWell). An unpredicted learning from this work is that our content, approach, and handbook have been useful to a number of foundations and grantmakers. They have improved their prioritization, decision-making, their clarity of mission and, ultimately, their impact (particularly in the early stages of setting up a foundation or granting department).
    With this in mind, we incubated one grantmaking foundation in 2021, which proved successful. This, combined with the number of interested foundations, led us to pilot a program focused on grantmaking foundations in 2022.
    What
    The pilot program ran for four weeks (three remote and one in person in London) and required participants’ full-time commitment. It was high intensity relative to other programs typically aimed at philanthropists.
    During the remote weeks there were readings from a 300+ page handbook written for the purpose, ~3x daily video lectures, 2x daily projects and 1x daily group discussion.
    During the in-person week, in addition to group discussions, we ran mock grantmaking and donor coordination exercises. We supported participants to work towards a concrete final deliverable that they could use to better their grantmaking going forward.
    The content of the program was focused on (a) equipping the foundations with decision-making tools, e.g., cost-effectiveness analyses, (b) making key decisions regarding their scope, structure and strategy, and (c) sharing best practices on vetting processes in the contexts of hiring and grantmaking.
    Who
    Five foundations (represented by six people) participated in the pilot. Once these foundations have launched and scaled, their combined annual donations are expected to be approximately $60M/yr. Participants had a wide range of grantmaking experience (from none to many years), foundation maturity (from yet-to-exist to fairly established), EA knowledge (from novice to expert) and expected annual dollars granted (from $100k to >$10M).
    How it went
    What went well
    The program was a pilot, and as a result it was quite rough around the edges. Despite this, we think the program was very successful! From our perspective, the aspects of the program that went particularly well were the group discussions, the mock grantmaking exercises and the final project at the end.
    Feedback has suggested that the group discussion sessions were extremely valuable. These sessions gave part...
    11 min
  • AF - AI Safety Camp, Virtual Edition 2023 by Linda Linsefors
    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 Camp, Virtual Edition 2023, published by Linda Linsefors on January 6, 2023 on The AI Alignment Forum.
    AI Safety Camp connects you with an experienced research lead to collaborate on a research project – helping you try your fit for a potential career in AI Safety research.
    The applications for AI Safety Camp’s Virtual Edition in 2023 are now open!
    AI Safety Camp Virtual 8 will be a 3.5-month long online research program from 4 March to 18 June 2023, where participants form teams to work on pre-selected projects.
    We value people with diverse backgrounds and skillsets, such as cybersecurity or cognitive science. Not all projects require participants to have prior experience in AI Safety, mathematics or machine learning. You will be able to read in detail about the research topics & each project’s skill requirements for our upcoming edition by following the Project Proposal links below.
    Projects
    Projects you can apply to.
    Conceptualise AGI dynamics
    Investigate transformer models
    Finetune language transformers
    Create annotation tools
    Review and analyse literature
    Propose public policy/communication
    Apply to AISC Virtual 2023
    Apply if you.
    want to try out & consider ways you could help ensure that future AI performs safely and in line with what people value upon reflection;
    are ready to dig into our research leads’ research topics and are able to write down a few clear arguments for why you’d research one in particular & how you might start;
    previously studied a topic or practiced skills unique to your perspective/background that can bolster your new research team’s progress;
    can block off hours to focus on research from March to June 2023 on normal workdays and the weekends (at least 10 hours per week).
    Application timeline
    5 Jan 202300:01 UTCProject proposals are posted on our website. Application form opens. Reviews start right away .19 Jan23:59 AoEDeadline to apply to teams closes. Late submissions might not get a response.1 March23:59 AoELast applicant admitted or declined (most will be informed of our decision earlier).
    First virtual edition – a spontaneous collage
    Timeline
    January 5: Accepted proposals are posted on the AISC website. Application to join teams open.
    January 19: Application to join teams closes.
    Until February end: Organisers pre-filter applications. RLs interview potential members and pick their team.
    March 4-5: Opening weekend.From here, teams meet weekly, and plan in their own work hours.
    June 17-18: Closing weekend.All teams present their results.
    Team structure
    Every team will have:
    one Research Lead
    one Team Coordinator
    other team members
    All team members are expected to work at least 10 hours per week on the project, which includes joining weekly team meetings, and communicating regularly (between meetings) with other team members about their work.
    As of yet, we cannot commit to offering stipends compensation for team members, because a confirmed grant fell through. Another grantmaker is in the midst of evaluating a replacement grant for AI Safety Camp. If confirmed, team members can opt in to receive a minimum of $500 gross per month (up to $2000 for full-time work).
    Research Lead (RL)
    The RL is the person behind the research proposal. If a group forms around their topics, the RL will guide the research project, and keep track of relevant milestones. When things inevitably don’t go as planned (this is research after all) the RL is in charge of setting the new course.
    The RL is part of the research team and will be contributing to research the same as everyone else on the team.
    Team Coordinator (TC)
    The TC is the ops person of the team. They are in charge of making sure meetings are scheduled, checks in with individuals on their task progress, etc. TC and RL can be the same person.
    The role of the TC is important but not expected to...
    5 min
  • EA - Bill Burr on Boiling Lobsters (also manliness and AW) by Lixiang
    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: Bill Burr on Boiling Lobsters (also manliness and AW), published by Lixiang on January 4, 2023 on The Effective Altruism Forum.
    I think there's a vibe out there in many cultures (including American) that being vegetarian/vegan or certain kinds of sympathy towards animals is unmanly and just kind of lame. This is probably more true in the right-wing demographic. I'm guessing this has been discussed in the animal welfare movement somewhere, so I won't attempt to delve into the issue further in this post. Instead, I merely want to favorably acknowledge some commentary by comedian Bill Burr about boiling lobsters alive.
    Bill Burr is a super-famous comedian and one of the most prominent cultural icons of masculinity in the U.S (perhaps in some respects the most prominent). Although I would say he is a party-neutral comedian, his comedic themes have included anti-wokeness and challenges to certain aspects of feminism, and probably has a huge following among working-class right-wing men.
    Here is his commentary on boiling lobsters alive (6 min), excerpted from his podcast.
    Edit: Actually I will go into it for a minute. If anyone wants to see a great example of how to deal with this sort of thing in a different context, take a look at how Ford dealt with the issue of environmentalism/green-politics being considered soft, lefty, snowflake stuff when they wanted to advertise the fuel economy on their F150 pickup truck: "you won't be put in a chokehold everytime you fill up".
    Other versions of the commercial here and here.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min