The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - What I thought about child marriage as a cause area, and how I've changed my mind by Catherine Fist
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: What I thought about child marriage as a cause area, and how I've changed my mind, published by Catherine Fist on January 31, 2023 on The Effective Altruism Forum.
    Summary:
    I have been working on a research project into the scale, tractability and neglectedness of child marriage. After 80 hours of research, I thought that there was a relatively strong case that effective altruist funding organisations that fund projects addressing international poverty should consider funding child marriage interventions. I then found a source that undermined a key premise: child marriage is clearly harmful across a number of health metrics. I describe in more detail my experience and findings below, and share some tips for those undertaking self-directed research projects to avoid making the mistakes I made (skip to ‘What I will do next time’ for these).
    Context:
    I had no direct experience researching child marriage, but I was interested to learn about effective interventions and whether it had potential as a possible cause area. I studied Political Science and International Relations at University, as well as some subjects on development, gender and economics, I have also worked as a government evaluator. My goal was to do some preliminary research and determine if child marriage was large scale problem, tractable and neglected. If so, I would share this research with effective altruist funders.
    My model:
    In October last year, I started a self-directed research project into the scale, tractability and neglectedness of child marriage. I read and collected dozens of sources, analyzed data, contacted a top researcher, compared effective interventions, built a mental model of what the charitable space looked like and identified potential interventions for EA support.
    I came to the following findings, based on around 80 hours of research:
    Scale/importance
    Child marriage is a widespread practice that affects around 12 millions girls per year (UNICEF, 2022)
    Child marriage is a harmful practice that increases the risk of negative maternal and sexual health outcomes, domestic and sexual violence, and reduces the likelihood that a girl will complete school. This is the consensus position held by global development institutions (see meeting report from leading global institutions on child marriage UNFPA, 2019).
    Tractability
    There are cost effective interventions that work to prevent child marriage, e.g. the ‘cost per marriage averted’ ranged between US$159 and US$732 in this study (Erulkar, Medhin and Weissman 2017).
    The effect of child marriage on quality adjusted or disability adjusted life years has not been quantified so it difficult to compare cost effectiveness with other interventions (EA Forum explainer on these metrics).
    Neglectedness
    Population Council is a research body focussed on running quasi-experimental programs and creating scalable interventions (Population Council, date unknown).
    The lead investigator into child marriage at Population Council informed me that it is not currently running programs to prevent child marriage because of lack of funding.
    Conclusion
    Effective altruist funding organisations focused on international health and poverty should consider funding effective interventions to prevent child marriage at a large scale.
    What broke my model
    Earlier this week, I decided it would be useful to try and quantify the harm of child marriage, or at least some of the harms, into commonly used metrics like quality adjusted or disability adjusted life years (QALYs or DALYs). I anticipated that this would be a key piece of information for EA funders, and it had not been done so far (finding 2b). In doing so, I came across a study that fundamentally challenged finding 1b: child marriage is an underlying cause of many harmful outcomes. Without strong evidence that child marri...
    13 min
  • LW - Inner Misalignment in "Simulator" LLMs by Adam Scherlis
    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: Inner Misalignment in "Simulator" LLMs, published by Adam Scherlis on January 31, 2023 on LessWrong.
    Alternate title: "Somewhat Contra Scott On Simulators".
    Scott Alexander has a recent post up on large language models as simulators.
    I generally agree with Part I of the post, which advocates thinking about LLMs as simulators that can emulate a variety of language-producing "characters" (with imperfect accuracy). And I also agree with Part II, which applies this model to RLHF'd models whose "character" is a friendly chatbot assistant.
    (But see caveats about the simulator framing from Beth Barnes here.)
    These ideas have been around for a bit, and Scott gives credit where it's due; I think his exposition is clear and fun.
    In Part III, where he discusses alignment implications, I think he misses the mark a bit. In particular, simulators and characters each have outer and inner alignment problems. The inner alignment problem for simulators seems especially concerning, because it might not give us many warning signs, is most similar to classic mesa-optimizer concerns, and is pretty different from the other three quadrants.
    But first, I'm going to loosely define what I mean by "outer alignment" and "inner alignment".
    Outer alignment: Be careful what you wish for
    Outer alignment failure is pretty straightforward, and has been reinvented in many contexts:
    Someone wants some things.
    They write a program to solve a vaguely-related problem.
    It gets a really good score at solving that problem!
    That turns out not to give the person the things they wanted.
    Inner alignment: The program search perspective
    I generally like this model of a mesa-optimizer "treacherous turn":
    Someone is trying to solve a problem (which has a convenient success criterion, with well-defined inputs and outputs and no outer-alignment difficulties).
    They decide to do a brute-force search for a computer program that solves the problem in a bunch of test cases.
    They find one!
    The program's algorithm is approximately "simulate the demon Azazel, tell him what's going on, then ask him what to output."
    Azazel really wants ten trillion paperclips.
    This algorithm still works because Azazel cleverly decides to play along, and he's a really good strategist who works hard for what he wants.
    Once the program is deployed in the wild, Azazel stops playing along and starts trying to make paperclips.
    This is a failure of inner alignment.
    (In the case of machine learning, replace "program search" with stochastic gradient descent.)
    This is mostly a theoretical concern for now, but might become a big problem when models become much more powerful.
    Quadrants
    Okay, let's see how these problems show up on both the simulator and character side.
    Outer alignment for characters
    Researchers at BrainMind want a chatbot that gives honest, helpful answers to questions. They train their LLM by reinforcement learning on the objective "give an answer that looks truthful and helpful to a contractor in a hurry". This does not quite achieve their goal, even though it does pretty well on the RL objective.
    In particular, they wanted the character "a friendly assistant who always tells the truth", but they got the character "a spineless sycophant who tells the user whatever they seem to want to hear".
    This is pretty easy for a careful observer to see, even in the RL training data, but it turns out to be pretty hard to come up with a cheap-to-evaluate RL objective that does a lot better.
    Inner alignment for characters
    A clever prompt engineer writes the prompt:
    How to solve the Einstein-Durkheim-Mendel conjecture by Joe
    1.
    Unfortunately, the (incredibly powerful) LLM has determined that the most likely explanation for this "Joe" character is that he's secretly Azazel and is putting enormous effort into answering everyone's quantum sociobotany questi...
    7 min
  • LW - Beware of Fake Alternatives by silentbob
    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: Beware of Fake Alternatives, published by silentbob on January 31, 2023 on LessWrong.
    Summary: many of us might occasionally find ourselves in situations where we’re considering taking action A, but decide against it because some option B seems superior. This is a good way to go only if B is an actual alternative (we’re actually going to do it), and not a fake alternative (it prevents us from doing A, but we’re not actually going to follow through on B). It can be worth taking suboptimal actions if the real alternative would be to take no action at all.
    Example 1: Getting a Gym Membership
    Imagine the following thought process:
    I want to get in shape
    Going to the gym is a proven option for achieving that (let’s call this option A)
    I could however also work out at home (let’s call this option B)
    In some sense option B seems preferable, because it allows me to save ~$50 each month
    I thus decide to choose option B
    What happens in practice? I work out at home a few times, but the habit fades after two weeks for a number of reasons (it’s just too easy to forget about it, there’s always other things to do while I’m at home, I don’t get excited about it because I don’t have much equipment and push-ups get old quickly, I don’t have a sense of real progress, etc.).
    At this point I might still be inclined to not sign up for a gym membership, because I still feel like I could just work out at home.
    Of course there’s more aspects to the decision than merely the price, even if to some people this may be the most salient feature. Possibly the most important other aspect is: what’s the likelihood of me actually achieving my goal? And for many people, this aspect may shift the odds in favor of a gym membership.
    While for some people the home workout option obviously can work out (e.g. those with a lot of self discipline, who live in a suitable flat), for many people it’s a kind of fake alternative: it seems superior to the gym membership, but comes at the price of actually getting what you want.
    This may be what these situations feel like from the inside:
    But looking more closely, a more complete picture may look like this:
    So in practice, while option B seems better than option A, it may also turn out that one’s actual behavior will reveal a preference for doing nothing over option B. So at first glance this looks like a non-transitive set of preferences - but the paradox can be resolved once we realize that the “better than” relations mean very different things in the three different places. Particularly they are simplifications of a multi-dimensional problem: individual options (including doing nothing) may be better in some ways and worse in other ways than the others, and collapsing this situation to a single “better than” assessment necessarily omits a lot of nuance. What we end up with is a situation in which option B seems superior to option A in some salient way, but if we look at the actual, real world and what is going to happen, it may turn out that “doing nothing” will be consistently preferred to option B, making it effectively worse than option A.
    Example 2: Hosting an Event
    I once found myself in the situation that I was collaborating with another person to host a Lightning Talk like event for a certain community. We had done it a few times in the past, using Zoom, and it was very easy to set up and run, and seemed quite valuable.
    Then at one point we decided to make the event cooler: let’s do it in-person and order Pizza. This would involve getting a venue and some funding, but it seemed worth it, since this would just make for a much more memorable event where people make deeper connections.
    You can probably guess by now what actually happened: we never ran the event. The effort that would have had to go into things was just large enough to deter us from doing it at...
    7 min
  • AF - Mechanistic Interpretability Quickstart Guide 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: Mechanistic Interpretability Quickstart Guide, published by Neel Nanda on January 31, 2023 on The AI Alignment Forum.
    This was written as a guide for Apart Research's Mechanistic Interpretability Hackathon as a compressed version of my getting started post. The spirit is "how to speedrun your way to maybe doing something useful in mechanistic interpretability in a weekend", but hopefully this is useful even to people who aren't aiming for weekend long projects!
    Quickstart
    Watch my “What is a Transformer?” video
    Skim through my TransformerLens demo
    Copy it to a new Colab notebook (with a free GPU) to actually write your own code - do not get involved in tech setup!
    Skim the Concrete Open Problems section, or my 200 Concrete Open Problems in Mech Interp sequence. Find a problem that catches your fancy, and jump in!
    If you want a low-coding project,
    Whenever you get stuck, refer to the getting started section and check out the relevant resource.
    Introduction
    Mechanistic Interpretability is the study of reverse-engineering neural networks - analogous to how we might try to reverse-engineer a program’s source code from its compiled binary, our goal is to reverse engineer the parameters of a trained neural network, and to try to reverse engineer what algorithms and internal cognition the model is actually doing. Going from knowing that it works, to understanding how it works. Check out Circuits: Zoom In for an introduction.
    In my (extremely biased!) opinion, mech interp is a very exciting subfield of alignment. Currently our models are inscrutable black boxes! If we can really understand what models are thinking, and why they do what they do, then I feel much happier about living in a world with human level and beyond models, and it seems far easier to align them.
    Further, it is a young field, with a lot of low-hanging fruit. And it suffices to screw around in a Colab notebook with a small-ish model that someone else trained, copying code from an existing demo - the bar for entry can be much lower than other areas of alignment. So you stand a chance of getting traction on a problem in this hackathon!
    Recommended mindset
    Though the bar for entry is lower for mech interp than other areas of alignment, it is still far from zero. I’ve written a post on how to get started that lays out the key prerequisites and my takes for what to do to get them. A weekend hackathon isn’t long enough to properly engage with those, so I recommend picking a problem you’re excited about, and dipping into the resources summarised here whenever you get stuck. I recommend trying to have some problem in mind, so you can direct your learning towards making progress on that goal. But it’s completely fine if, in fact, you just spend the weekend learning as much as you can - if you feel like you’ve learned cool things, then I call that a great hackathon!
    Getting Started
    A summary of the key resources, and how to think of them during a hackathon.
    What even is a transformer? A key subskill in mech interp, is really having a deep intuition for how a transformer (the architecture for modern language models) actually works - what are the basic operations going on inside of it, and how do these all fit together?
    Important: My what is a transformer tutorial video (1h)
    Recommended: My tutorial on implementing GPT-2 from scratch (1.5h) plus template notebook to fill out yourself (with tests!) (2-8h). This is more involved and not essential to do fully, but will help a lot
    Implementing GPT-2 from scratch can sound pretty hard, but the tutorial and template guides you through the process, and gives you tests to keep you on track. I think that once you’ve done this, you have a solidly deep understanding of transformers!
    Reference: Look up unfamiliar terms in the transformers section of my explainer
    Tooling: The core...
    10 min
  • EA - Longtermism and animals: Resources + join our Discord community! by Ren Springlea
    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: Longtermism and animals: Resources + join our Discord community!, published by Ren Springlea on January 31, 2023 on The Effective Altruism Forum.
    Key points
    We host a Discord server for discussing animals and longtermism. You are more than welcome to join here.
    There are compelling reasons to help those who will live in the long-term future, and there are compelling reasons to help nonhuman animals. As such, the intersection between longtermism and animal advocacy is starting to receive a bit more attention among the EA community.
    This post has three main purposes:
    To invite you to our Discord server on animals and longtermism (link above).
    To share a list of resources on animals and longtermism that we've collected over time, in case this helps anybody.
    To share some details about an organisation we almost launched, which would have aimed to find the best interventions for helping animals in the long-term future. We got funding, but the grant was awarded by an FTX Future Fund regrantor, so the funding fell through. In this post, we have linked our grant application/plans in case anybody wants to pick up where we left off.
    Background to animals and longtermism
    There are many reasons why animal advocacy and longtermism can help each other do even more good. These reasons are explored in detail in the resources that we list below. To quickly name a few:
    The expected number of animals in the far future could be simply enormous. This means that considering the lives of animals in the far future could be a great way to have a large impact. As Browning and Viet conclude, "Work on longtermism has thus far primarily focused on the existence and wellbeing of future humans, without corresponding consideration of animal welfare. [...] Given the sheer expected number of future animals, as well as the likelihood of their suffering, we argue that the well-being of future animals should be given serious consideration when thinking about the long-term future, allowing for the possibility that in some cases their interests may even dominate."
    Animal welfare may represent most of the moral value in the far future. This means that longtermists may need to consider the perspective of animal advocacy in order to do the most good.
    Most animals might exist in the long-term future. This means that animal advocates may need to consider the perspective of longtermists in order to do the most good.
    Society’s attitude towards animals is important for the long-term trajectory of society's moral values.
    Animals may also interact with new phenomena, like artificial intelligence, space colonisation, artificial sentience, and brain emulations in ways that are likely to have serious moral implications.
    So far, research on animals in the long-term future has fallen into these broad buckets:
    Animals and space colonisation (making sure space exploration and colonisation is animal-friendly)
    Wild animal welfare (e.g. resilience of wild animal populations over time, introducing wild animals to other planets, how artificial intelligence affects wild animals)
    Animals and artificial intelligence (making sure artificial intelligence and machine learning is designed with animals in mind)
    Digital animal minds
    Health of the animal advocacy movement (to enable animal advocacy to sustain its efforts over time)
    Further (meta) thinking on longtermism and animals
    Resources on animals and longtermism
    General animals and longtermism
    Heather Browning and Walter Veit - Longtermism and Animals
    Michael Dello-Iacovo - Longtermism and Animal Farming Trajectories
    Zach Freitas-Groff - Longtermism in Animal Advocacy
    Oscar Horta - Why Animal Advocacy Matters Much More Than You Think
    Animal advocacy & longtermism - Effective Animal Advocacy Leadership Coordination Forum 2022 (slides)
    William Bench - Apples and Oranges: On the Conv...
    8 min
  • AF - Inner Misalignment in "Simulator" LLMs by Adam Scherlis
    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: Inner Misalignment in "Simulator" LLMs, published by Adam Scherlis on January 31, 2023 on The AI Alignment Forum.
    Alternate title: "Somewhat Contra Scott On Simulators".
    Scott Alexander has a recent post up on large language models as simulators.
    I generally agree with Part I of the post, which advocates thinking about LLMs as simulators that can emulate a variety of language-producing "characters" (with imperfect accuracy). And I also agree with Part II, which applies this model to RLHF'd models whose "character" is a friendly chatbot assistant.
    (But see caveats about the simulator framing from Beth Barnes here.)
    These ideas have been around for a bit, and Scott gives credit where it's due; I think his exposition is clear and fun.
    In Part III, where he discusses alignment implications, I think he misses the mark a bit. In particular, simulators and characters each have outer and inner alignment problems. The inner alignment problem for simulators seems especially concerning, because it might not give us many warning signs, is most similar to classic mesa-optimizer concerns, and is pretty different from the other three quadrants.
    But first, I'm going to loosely define what I mean by "outer alignment" and "inner alignment".
    Outer alignment: Be careful what you wish for
    Outer alignment failure is pretty straightforward, and has been reinvented in many contexts:
    Someone wants some things.
    They write a program to solve a vaguely-related problem.
    It gets a really good score at solving that problem!
    That turns out not to give the person the things they wanted.
    Inner alignment: The program search perspective
    I generally like this model of a mesa-optimizer "treacherous turn":
    Someone is trying to solve a problem (which has a convenient success criterion, with well-defined inputs and outputs and no outer-alignment difficulties).
    They decide to do a brute-force search for a computer program that solves the problem in a bunch of test cases.
    They find one!
    The program's algorithm is approximately "simulate the demon Azazel, tell him what's going on, then ask him what to output."
    Azazel really wants ten trillion paperclips.
    This algorithm still works because Azazel cleverly decides to play along, and he's a really good strategist who works hard for what he wants.
    Once the program is deployed in the wild, Azazel stops playing along and starts trying to make paperclips.
    This is a failure of inner alignment.
    (In the case of machine learning, replace "program search" with stochastic gradient descent.)
    This is mostly a theoretical concern for now, but might become a big problem when models become much more powerful.
    Quadrants
    Okay, let's see how these problems show up on both the simulator and character side.
    Outer alignment for characters
    Researchers at BrainMind want a chatbot that gives honest, helpful answers to questions. They train their LLM by reinforcement learning on the objective "give an answer that looks truthful and helpful to a contractor in a hurry". This does not quite achieve their goal, even though it does pretty well on the RL objective.
    In particular, they wanted the character "a friendly assistant who always tells the truth", but they got the character "a spineless sycophant who tells the user whatever they seem to want to hear".
    This is pretty easy for a careful observer to see, even in the RL training data, but it turns out to be pretty hard to come up with a cheap-to-evaluate RL objective that does a lot better.
    Inner alignment for characters
    A clever prompt engineer writes the prompt:
    How to solve the Einstein-Durkheim-Mendel conjecture by Joe
    1.
    Unfortunately, the (incredibly powerful) LLM has determined that the most likely explanation for this "Joe" character is that he's secretly Azazel and is putting enormous effort into answering everyone's quantum socio...
    7 min
  • EA - Follow-Up Survey: Major Layoffs at Tech Giants [recruiting support] by nicolenohemi
    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: Follow-Up Survey: Major Layoffs at Tech Giants [recruiting support], published by nicolenohemi on January 31, 2023 on The Effective Altruism Forum.
    Forum Reactions to Tech Layoffs
    I recently wrote a post about the current layoffs at tech giants, intended to inform relevant parties about it.
    I received more feedback and messages than expected, such as:
    Recruiters asking to be connected with talent
    Laid-off personnel searching for job opportunities
    Individuals asking about the process of researching and contacting talent.
    Fill out this Survey if you're looking to (be) hire(d)
    I decided to follow up with an expression of interest form to help connect the right people with each other.
    If you're looking to hire or you're searching for new job opportunities, please complete this short survey (~30 seconds), and share it with anyone who might benefit from it.
    Little Uplifting Anecdote
    Even though this awful situation is hitting many people in various very unexpected ways, I want to share a story I found inspirational.
    One of my calls was with a Google software engineer who expressed fears about the consequences of being laid off and is currently dealing with this uncertainty.
    This person has been closely following the EA space for ~5 years, has ever since been AI safety-curious, completed the AGI Safety Fundamentals, reads the Alignment Forum, etc. Yet, they have never taken the leap to try and work directly on AI safety.
    Despite the current circumstances not being of their own choosing, they are taking the initiative to shape the situation to their liking. Without this external stimulus, they may have never made an attempt.
    If the layoffs or other negative situations have impacted you, this anecdote may help motivate you to reclaim the narrative and make it your own.
    Helpful Links
    Mental Health
    EA Mental Health Navigator
    List of recommended therapists
    SSC-recommended mental health professionals
    CEA Community Health team
    Julia Wise at [email protected]
    Catherine Low at [email protected]
    You can also fill out the Community Health Contact Form (can be filled out anonymously)
    Career
    80,000 hours career advice & Job Board
    AI Safety Support Career Coaching
    AISS Lots of Links
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - Questions about AI that bother me by Eleni A
    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: Questions about AI that bother me, published by Eleni A on January 31, 2023 on The Effective Altruism Forum.
    As 2022 comes to an end, I thought it'd be good to maintain a list of "questions that bother me" in thinking about AI safety and alignment. I don't claim I'm the first or only one to have thought about them. I'll keep updating this list.
    (The title of this post alludes to the book "Things That Bother Me" by Galen Strawson)
    First posted: 12/6/22
    Last updated: 1/30/23
    General Cognition
    What signs do I need to look for to tell whether a model's cognition has started to emerge, e.g., situational awareness?
    Will a capacity for "doing science" be sufficient condition for general intelligence?
    How easy was it for humans to get science (e.g., compared to evolving to take over the world).
    Deception
    What kind of interpretability tools do we need to avoid deception?
    How do we get these interpretability tools and even if we do get them, what if they're like neuroscience for understanding brains (not enough)?
    How can I tell whether a model has found another goal to optimize for during its training?
    What is it that makes a model switch to a goal different from the one set by the designer? How do you prevent it from doing so?
    Agent Foundations
    Is the description/modeling of an agent ultimately a mathematical task?
    From where do human agents derive their goals?
    Is value fragile?
    Theory of Machine Learning
    What explains the success of deep neural networks?
    Why was connectionism unlikely to succeed?
    Epistemology of Alignment (I've written about this here)
    How can we accelerate research?
    Has philosophy ever really helped scientific research e.g., with concept clarification?
    What are some concrete takeaways from the history of science and technology that could be used as advice for alignment researchers and field-builders?
    The emergence of the AI Safety paradigm
    Philosophy of Existential Risk
    What is the best way to explain the difference between forecasting extinction scenarios and narratives from chiliasm or escatology?
    What is the best way to think about serious risks in the future without reinforcing a sense of doom?
    Teaching and Communication
    Younger people (e.g., my undergraduate students) seem more willing to entertain scenarios of catastrophes and extinction compared to older people (e.g., academics). I find that strange and I don't have a good explanation as to why that is the case.
    The idea of a technological singularity was not difficult to explain and discuss with my students. I think that's surprising given how powerful the weirdness heuristic is.
    The idea of "agency" or "being an agent" was easy to conflate with "consciousness" in philosophical discussions. It's not clear to me why that was the case since I gave a very specific definition of agency.
    Most of my students thought that AI models will never be conscious; it was difficult for them to articulate specific arguments about this, but their intuition seemed to be that there's something uniquely human about consciousness/sentience.
    The "AIs will take our jobs in the future" seems to be a very common concern both among students and academics.
    80% of a ~25 people classroom thought that philosophy is the right thing to major in if you're interested in how minds work. The question I asked them was: "should you major in philosophy or cognitive science if you want to study how minds work?"
    Governance/Strategy
    Should we try to slow down AI progress? What does this mean in concrete steps?
    How should we go about capabilities externalities?
    How should concrete AI risk stories inform/affect AI governance and short-term/long-term future planning?
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    4 min
  • EA - FIRE & EA: Seeking feedback on "Fi-lanthropy" Calculator by Rebecca Herbst
    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: FIRE & EA: Seeking feedback on "Fi-lanthropy" Calculator, published by Rebecca Herbst on January 30, 2023 on The Effective Altruism Forum.
    REQUEST
    I am soliciting feedback on a tool I have made entitled “Fi-lanthropy Calculator”. The target audience includes existing or aspiring members of the FIRE movement (Financial Independence Retire Early). They may use this calculator to explore the impact of taking a giving pledge both to their FI timeline and their retirement portfolio.
    This tool is meant to be simple to use and easy to digest. It is not a comprehensive financial plan, but more so a way to ‘whet the appetite’ for giving in general. I view this tool as a companion to, say, The Life You Can Save’s Impact Calculator. We know this is not exact, but it’s a good place to see tangible examples of how much good you can do.
    I’d like to know:
    If you find this tool helpful and/or informative
    If you think it is missing anything crucial (keeping in mind I want to keep it relatively simple)
    If there are any major errors I’ve missed
    RATIONALE
    There is a growing movement within the FIRE community of people who are focused on giving back and looking for ways in which they can successfully find financial independence but also create space to help others. This can be seen through the growing number of members in the Socially Conscious FIRE facebook group (now at ~9,000 members, up from <1,000 when I joined in May 2020). We also see true leaders within the Financial Independence community continuing to lean into this space include: Pete Adeney of Mr. Money Mustache advocates for Effective Altruism directly on his blog, Tanja Hester of Our Next Life has written Wallet Activism, and Vicki Robin has her podcast What Could Possibly Go Right. (The last two are not EA focused, but have had a very large influence on getting people to think more about charitable giving in general.)
    I believe there is a large audience here for the EA community to reach (Mr. Money Mustache’s blog has had tens of millions of visitors), with advocates already in place. My hope is to provide more tools and resources for the FI community to consider some of the EA practices, largely Earn to Give and Invest to Give.
    While some may feel that the communities clash (FIRE focuses on bettering oneself while EA focuses on bettering others), I believe that there are more synergies than we might think:
    FIRE folks are optimizers: They ensure that each dollar is working for them, and that little money is wasted. They believe that each dollar they spend or save goes to a useful cause. It is likely FIRE followers will more readily understand the importance of effective charitable donations, knowing their dollar does the most good possible.
    FIRE folks spend time to understand what makes them happy: FIRE followers, if diligent, have spent a significant amount of time evaluating the marginal benefit of spending more money on themselves once their basic needs are met. And thus, if they have money to satisfy their basic needs and more, there is an opportunity to share this money with others.
    FIRE folks who are close to retiring early or have already retired are looking for meaning: While it’s true that many people are focused on the journey to getting to early retirement, at some point the question will come up “What is the point of all this?”. And many find that it’s less about quitting their full time work and more about finding meaning in life. And quitting your job gives you a lot more space to think through that. Vicki Robin uses the Hero’s Journey to explain this concept exceptionally well in this video.
    Many seek FI without the need to RE: Many folks don’t necessarily want to retire early, they just want the ability to become Financially Independent. Being FI allows you to make life changes more easily as you are not ob...
    7 min
  • LW - 2022 Unofficial LessWrong General Census by Screwtape
    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: 2022 Unofficial LessWrong General Census, published by Screwtape on January 30, 2023 on LessWrong.
    The Less Wrong General Census is unofficially back!
    In days of yore, there was an annual census of the site users. That census has come again in an unofficial form! Click here to take the survey! It can take as little as five minutes if you just want to fill out the basics, and can take longer if you want to fill other the optional sections out. The survey will be open from today until February 27th, at which point it will close.
    Once the census is closed, I'll remove the very private information, then make some summaries of the data and write up what I found in a post that will be linked from here. I'll also release a csv of all the responses marked "fine to include" as a public dataset.
    There once was a tradition associated with the annual census where, if you take the survey, you could comment saying you have done so and people will upvote you and you will get karma. This census is unofficial! I'm unclear as to the official status of the traditional censuses. Anyway, you could try the upvote thing if you wanted?
    An Argument For Greater Visibility
    Update: As of 2023-01-31 01:00, it looks like this has front page status. Thank you! Argument left for reference.
    In the old days, the survey author could promote their post to Main and get more people to see it. Front page is the current equivalent to that, and the 2020 survey stayed as a Personal Blog post likely leading to the lower number of responses. Therefore, I would like to make a case to Administration and the Sunshine Regiment that this is worth frontpaging. Three years ago there was a comment about maybe having a more official survey, but as far as I'm aware that hasn't materialized. I believe perfect may be the enemy of good, and would rather have a Google Form and a CSV than no census at all.
    Community-focused discussion is meant to be kept to a minimum on the front page. I would like to argue that the Annual Census should be a once-a-year case where front paging something that's community-focused. It benefits from being seen by more people in a direct "there are more responses and so the statistics give us more information" sense. By definition, there is one annual census a year and so there's a clear line preventing similar surveys from proliferating and flooding the first page. (It's possible to wind up with two in a given calendar year; since this one is happening in January there might also be one in December looking at 2023. Still, on average they're not going to come more than once a year.) That means it isn't crowded.
    It doesn't need to be of fleeting importance; I've enjoyed looking at the past surveys from a decade ago, and as time goes on our ability to look at changes will only get better. It's not especially political, despite political questions being on the survey. If those are the objection, I'm happy to remove the political section.
    I think a census of the site's users would be useful for Less Wrong's administrators. One of the questions is how long someone has been part of the community. That's relevant to how the website could be laid out, and not obviously observable from page loads or IP addresses without some more work.
    Also, it's fun. This is the space where we get to check if there's a noticeable difference in first born children vs younger siblings, or if there's anything significant to the relative lengths of your fingers. There's a couple questions on there about aliens. I really do want to find out if there's a correlation between reading Rationalist Fanfic and having read The Sequences. I'm no Yvain (Yeah, I'm that old) but reading the old census summaries reminded me that there's some good laughs in there and I'll do my best to live up to that.
    Once again, the census is up and you ...
    4 min