The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - 80,000 Hours has been putting much more resources into growing our audience by Bella
    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: 80,000 Hours has been putting much more resources into growing our audience, published by Bella on February 27, 2023 on The Effective Altruism Forum.
    Since the start of 2022, 80,000 Hours has been putting a lot more effort and money into getting more people to hear of us and engage with our advice.
    This post aims to give some insight into what we’ve been up to and why.
    Why invest more in outreach?
    80,000 Hours has, we think, been historically cost-effective at causing more people to aim their careers at tackling pressing world problems. We've built a system of resources (website, podcast, job board, advising) that many people have found helpful for this end — and so we want more people to find them.
    Also, 80,000 Hours has historically been the biggest single source of people learning about the EA community. If we want to grow the community, increasing the number of people reached by 80k seems like one of the best available tools for doing that.
    Thirdly, outreach at the “top of the funnel” (i.e. getting people to subscribe after they hear about 80k’s ideas for the very first time) has unusually good feedback mechanisms & is particularly easy to measure. For the most part, we can tell if what we’re doing isn’t working, and change tack pretty quickly.
    Another reason is that a lot of these activities take relatively little staff time, but can scale quite efficiently with more money.
    Finally, based on our internal calculations, our outreach seems likely to be cost-effective as a means of getting more people into the kinds of careers we’re really excited about.
    What did we do to invest more in outreach?
    In 2020, 80k decided to invest more in outreach by moving one of their staff into a position focused on outreach, but it ended up not working out & that person left their role.
    Then in mid-2021, 80k decided to hire someone new to work on outreach full-time. They hired 1 staff member (me!), and I started in mid-January 2022.
    In mid-2022, we found that our initial pilots in this area looked pretty promising — by May we were on track to 4x our yearly rate of subscriber growth — and we decided to scale up the team and the resource investment. I ran a hiring round and made two hires, who started at the end of Nov 2022 and in Feb 2023; I now act as head of marketing.
    We also decided to formalise a “marketing programme” for 80k, which is housed within the website team. Since this project spends money so differently from the rest of 80k, and in 2022 was a large proportion of our overall spending, last year we decided to approach funders specifically to support our marketing spend (rather than draw from our general funds). The marketing programme has a separate fundraising cycle and decisions are made on it somewhat independently from the rest of 80k.
    In 2022, the marketing programme spent $2.65m (compared to ~$120k spent on marketing in 2021). The bulk of this spending was on sponsored placements with selected content creators ($910k), giving away free books to people who signed up to our newsletter ($1.04m), and digital ads ($338k). We expect to spend more in 2023, and are in conversation with funders about this.
    As a result of our efforts, more than 5x as many people subscribed to our newsletter in 2022 (167k) than 2021 (30k), and we had more website visitors in Q4 2022 than any previous quarter (1.98m).
    We can’t be sure how many additional people will change to a high-impact career as a result, in large part because we have found that “career plan changes” of this kind take, on average, about 2 years from first hearing about 80k.
    Still, our current best guess is that these efforts will have been pretty effective at helping people switch careers to more impactful areas.
    Partly this guess is based on the growth in new audience members that we’ve seen (plus 80k’s solid track record...
    12 min
  • EA - Very Briefly: The CHIPS Act by Yadav
    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: Very Briefly: The CHIPS Act, published by Yadav on February 26, 2023 on The Effective Altruism Forum.
    About six months ago, Congress passed the CHIPS Act. The "Creating Helpful Incentives to Produce Semiconductors for America Act", will spend $280 billion over the next ten years. $200 billion will go into scientific R&D and commercialization, $52.7 billion into semiconductor manufacturing, R&D, and workforce development, and $24 billion into tax credits (government subsidies) for chip production.
    Semiconductor production has been slipping in the United States for some time. While countries like China and Taiwan have maintained a strong foothold in the global chip market, the U.S. now produces just 12% of the world's semiconductors, down from 37% in the 1990s (source).
    The United States' dwindling position in the global semiconductor market, coupled with concerns about reliance on foreign suppliers - especially China and Taiwan - likely played a role in the introduction of the CHIPS Act. In a recent speech, Commerce Secretary Gina Raimondo spoke about how the CHIPS Act could help the U.S. regain its position as the top destination for innovation in chip design, manufacturing, and packaging. According to her, the U.S. "will be the premier destination in the world where new leading-edge chip architectures can be invented in our research labs, designed for every end-use application, manufactured at scale, and packaged with the most advanced technologies".
    An obvious reason I am concerned about this Act is that the increased investment in the U.S. semiconductor industry could enable AI capabilities companies in the US, such as OpenAI, to overcome computing challenges they may face right now. Additionally, other countries, such as the U.K. and the Member States of the EU, seem to be following suit. For example, the U.K. recently launched a research project aimed at building on the country's strengths in design, compound semiconductors, and advanced technologies. The European Chips Act also seeks to invest €43 billion in public and private funding to support semiconductor manufacturing and supply chain resilience. Currently, the Members of the European Parliament, are preparing to initiate talks on the draft Act.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - Remote Health Centers In Uganda - a cost effective intervention? by NickLaing
    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: Remote Health Centers In Uganda - a cost effective intervention?, published by NickLaing on February 27, 2023 on The Effective Altruism Forum.
    TLDR: Operating basic health centers in remote rural Ugandan communities looks more cost-effective than top GiveWell interventions on early stage analysis - with huge uncertainty.
    I’m Nick, a medical doctor who is co-founder and director of OneDay Health (ODH). We operate 38 nurse-led health centers in healthcare “black holes,” remote rural areas more than 5 km from government health facilities. About 5 million Ugandans live in these healthcare black holes and only have bad options when they get sick. ODH health centers provide high-quality primary healthcare to these communities at the lowest possible cost. We train our talented nurses to use protocol based guidelines and equip them with over 50 medications to diagnose and treat 30 common medical conditions. In our 5 years of operation, we have so far treated over 150,000 patients – including over 70,000 for malaria.
    Since we started up 5 years ago, we’ve raised about $290,000 of which we’ve spent around $220,000 to date. This year we hope to launch another 10-15 OneDay Health centers in Uganda and we're looking to expand to other countries which is super exciting!
    If you’re interested in how we select health center sites or more details about our general ops, check our website or send me a message I’d love to share more!
    Challenges in Assessing Cost-Effectiveness of OneDay Health
    Unfortunately, obtaining high-quality effectiveness data requires data from an RCT or a cohort study that would cost 5-10 times our current annual budget. So we've estimated our impact by estimating the DALYs our health centers avert through treating four common diseases and providing family planning. I originally evaluated this as part of my masters dissertation in 2019 and have updated it to more recent numbers. As we’re assessing our own organisation, the chance of bias here is high.
    Summary of Cost-Effectiveness Model
    To estimate the impact of our health centers, we estimated the DALYs averted through treating individual patients for 4 conditions: malaria, pneumonia, diarrhoea, and STIs. We started with Ugandan specific data on DALYs lost to each condition. We then adjusted that data to account for the risk of false diagnosis and treatment failure (in which case the treatment would have no effect). We then added impact from family planning. Estimating impact per patient isn’t a new approach. PSI used a similar method to evaluate their impact (with an awesome online calculator), but has now moved to other methods. Inputs for our approach
    Headline findings
    For each condition, we multiplied the DALYs averted per treatment by the average number of patients treated with that condition in one health center in one month. When we added this together that each ODH health center averted 13.70 DALYs per month, predominantly through treatment of malaria in all ages, and pneumonia in children under 5.ODH health centers are inexpensive to open and operate. Each health center currently needs only $137.50 per month in donor subsidies to operate. The remaining $262.50 in expenses are covered by small payments from patients. Many of these patients would have counterfactually received treatment, but would have incurred significantly greater expense to do so (mainly for travel). In addition, about 40% of patient expenses were for treating conditions not included in the cost-effectiveness analysis.
    We estimate that In one month, each health center averts 13.70 DALYs and costs $137.50 in donor subsidies. This is roughly equivalent to saving a life for $850, or more conservatively for $1766 including patient expenses. However, there is huge uncertainty in our analysis.
    The Analysis
    Measuring Impact by Estimating DALYs...
    16 min
  • LW - Respect Chesterton-Schelling Fences by shminux
    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: Respect Chesterton-Schelling Fences, published by shminux on February 27, 2023 on LessWrong.
    This post is inspired by the recent Ziz-revelations posted here and elsewhere.
    Chesterton Fence injunction: Do not remove a protective barrier unless you know why it was put there in the first place.
    Schelling fence injunction: Do not cross a generally agreed upon guardrail, lest you end up sliding down an increasingly slippery slope, not noticing until it is too late.
    I think a term like a Chesterton-Schelling Fence injunction might be useful: Respect an ethical injunction even if you think you know why it was put there in the first place.
    A somewhat simplified example: There is a rather strong Schelling fence against, say, killing someone. Suppose the stated reasoning behind it is "God commanded so". Some day, you deconvert and start questioning the tenets of your faith, throwing one injunction after another, assuming you know why it was there, not realizing that this particular Chesterton fence is fake, the real reason is an unstated Schelling fence that has little to do with religion, but a lot with living in a society.
    I said "respect" not "obey", because it is often hard to tell whether there is a hidden Schelling fence behind a Chesterton fence, and how strong the former is. Or vice versa. Or how many of the various hidden fences are there. Is it okay to cheat in an unhappy marriage? Maybe, maybe not, but noticing that this is an unsafe territory, that respecting the societal norms is generally a safe default, and that crossing it is likely yo backfire in both expected and unexpected ways can be quite useful.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • LW - Learning How to Learn (And 20+ Studies) by maxa
    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: Learning How to Learn (And 20+ Studies), published by maxa on February 26, 2023 on LessWrong.
    I have been interested in cognitive science and effective learning methods for years. I've read multiple books and articles and put many ideas to test. In this post, I aim to synthesize my notes and provide references to scientific studies.
    Summary (TL;DR)
    Effective Learning Strategies
    Distributed learning. Study less in each session but more frequently.
    Active recall. Actively test your knowledge and skills.
    Distributed recall. Space the tests in time and adjust the intervals based on performance.
    Interleaving. Practice multiple related yet different skills/concepts simultaneously.
    Elaborative interrogation (quiz-and-recall). Ask yourself questions and use the material you've learned to answer them.
    Self-explanation and the Feynman technique. Explain what you've just learned in simple terms.
    Physiology and Brain's Health
    Sleep
    Exercise
    Nutrition
    Disclaimer and Introduction
    I have no formal background in cognitive science or neuroscience, and this has been more of a side interest. My understanding is limited, and I still need to learn how to effectively and consistently apply all these ideas to practice.
    That being said, I found some of the methods described in this article very useful. For example, I've used them to learn foreign languages, the basics of programming, and various disciplines covered during the two-year MBA program.
    Effective Learning Strategies
    Strategy #1: Distributed (Spaced) Learning Practice
    In short, it's better to distribute one's practice over a period of time than cram it into one day.
    In one study, elementary school students were asked to study in one of the three ways: massed, clumped, and spaced.
    Massed = four lessons at a time
    Clumped = two lessons on one day and two lessons on the next day
    Spaced = one lesson per day for four days
    The “spaced” group performed best, followed by the “clumped” group:
    Another study compared comprehension scores under three different conditions:
    Read a text once (“single”)
    Read a text twice (“massed”)
    Read a text twice with a week-long gap (“distributed”)
    When tested immediately, the second group performed best. But when tested with a delay of two days, the third group performed best.
    This method is also superior for learning motor skills.
    How to apply this in practice:
    Create a learning schedule or find time to practice a little bit every day or every few days instead of cramming all your learning into one or just a few days.
    If you’d like to learn more, read the Wikipedia article on distributed practice.
    Strategy #2: Active Recall (Retrieval) Practice
    It might be more effective to actively retrieve the information you’ve already learned than passively re-read or try to learn it once again.
    One study that compared a method that emphasized study sessions with a method that emphasized tests and found the latter to be more effective for delayed recall.
    SSSS = four study sessions
    SSST = three study sessions, followed by one test
    STTT = one study session, followed by three tests
    Even imagining that you might be tested on the material you’re learning might help improve the recall.
    How to apply this in practice:
    If a few days ago you learned how past tense works in the Spanish language, try to remember the rules or even test yourself on your knowledge — instead of simply re-reading the same material once again.
    You can read more about the active recall practice on Wikipedia.
    Strategy #3: Distributed (Spaced) Recall Practice
    Distributed recall practice is basically a combination of the two ideas above. You test yourself frequently and modify the test intervals depending on how familiar you’re with the material or how strong your skill is.
    How to apply this in practice:
    Many apps simplify the process by tracking one’s perfor...
    12 min
  • EA - Let's Fund: Better Science impact evaluation. Registered Reports now available in Nature by Hauke Hillebrandt
    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: Let's Fund: Better Science impact evaluation. Registered Reports now available in Nature, published by Hauke Hillebrandt on February 26, 2023 on The Effective Altruism Forum.
    Cross-posted from my blog - inspired by the recent call for more monitoring and evaluation
    Hi, it's Hauke, the founder of Let's Fund. We research pressing problems, like climate change or the replication crisis in science, and then crowdfund for particularly effective policy solutions.
    Ages ago, you signed up to my newsletter. Now I've evaluated the $1M+ in grants you donated, and they had a big impact. Below I present the Better Science / Registered Report campaign evaluation, but stay tuned for the climate policy campaign impact evaluation (spoiler: clean energy R&D increased by billions of dollars).
    Let's Fund: Better Science
    Chris Chambers giving a talk on Registered Reports
    We crowdfunded ~$80k for Prof. Chambers to promote Registered Reports, a new publication format, where research is peer-reviewed before the results are known. This fundamentally changes the way research is done across all scientific fields. For instance, one recent Registered Report studied COVID patients undergoing ventilation1 (but there are examples in other areas including climate science,2 development economics,3 biosecurity, 4 farm animal welfare,5 etc.).
    Registered Reports have higher quality than normal publications,6 because they
    make science more theory-driven, open and transparent
    find methodological weaknesses and also potential biosafety failures of dangerous dual-use research prior to publication (e.g. gain of function research)7
    get more papers published that fail to confirm the original hypothesis
    increase the credibility of non-randomized natural experiments using observational data
    If Registered Reports become widely adopted, it might lead to a paradigm shift and better science. 300+ journals have already adapted Registered Reports. And just last week Nature, the most prestigious academic journal, adopted it:
    Chris Chambers on Twitter: "10 years after we created Registered Reports, the thing critics told us would never happen has happened: @Nature is offering them Congratulations @Magda_Skipper & team. The @RegReports initiative just went up a gear and we are one step closer to eradicating publication bias.
    This is big and Registered Reports might soon become the gold standard.
    Why? Imagine you’re a scientist with a good idea for an experiment with high value of information (think: a simple cure for depression). If that has a low chance of working out (say 1%), then previously you had little incentive to run it.
    Now, if your idea is really good, and based on strong theory, Registered Reports derisks running the experiment. You can first submit the idea and methodology to Nature and the reviewers might say: ‘This idea is nuts, but we agree there’s a small chance it might work and really interested in this works. If you run the experiment, we’ll publish this independent of results!’ Now you can go ahead spend a lot of effort on running the experiment, because even if it doesn’t work, you still get a Nature paper (which you wouldn’t with null results).
    This will lead to more high risk, high reward research (share this post or the tweet with academics! They might thank you for the Nature publication).
    Many people were integral to this progress, but I think Chambers, the co-inventor and prime proponent of Registered Reports deserves special credit. In turn he credited:
    Chris Chambers @chrisdc77: 'You. That's right. Some of the most useful and flexible funding I've received has been donated by hundreds of generous members of public (& small orgs) via our @LetsFundOrg-supported crowd sourcing fund'
    You may feel smug.
    If you want to make a bigger donation (>$1k), click here. There are proposals to improve Regis...
    5 min
  • LW - The Preference Fulfillment Hypothesis by Kaj Sotala
    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 Preference Fulfillment Hypothesis, published by Kaj Sotala on February 26, 2023 on LessWrong.
    Short version
    Humans have an innate motivation ("preference fulfillment", PF) to fulfill the preferences of those they care about. It corresponds to at least some of the senses of the word "love", as well as related words such as "kindness" and "compassion".
    I hypothesize that it works by simulating the other person and predicting what they would want or how they would like to be treated. PF is when you take your simulation of what other people would want and add an extra component that makes you intrinsically value outcomes that your simulation predicts the other people would prefer.
    I also hypothesize that this is the same kind of simulation that forms our ability to work as a social species in the first place. A mental simulation process is active in virtually every situation where we interact with other people, such as in a grocery store. People use masks/roles/simulations to determine the right behavior in any social situation, running simulations of how others would react to various behaviors. These simulations involve both the actual people present in the situation as well as various other people whose opinions we've internalized and care about. The simulations generally allow people to engage in interactions by acting the way a normal person would in a given situation.
    Once you have this kind of a simulation, constantly running in basically any social situation, it’s likely already exhibiting the PF drive to a weak degree. Doing things that we expect to fulfill other people’s preferences often feels intrinsically nice, even if the person in question was a total stranger. So does wordless coordination in general, as evidenced by the popularity of things like dance.
    If this is true, capabilities progress may then be closely linked to alignment progress. Getting AIs to be better at following instructions requires them to simulate humans better. Once you have an AI that can simulate human preferences, you already have most of the machinery required for having PF as an intrinsic drive. This is contrary to the position that niceness is unnatural. The preference fulfillment hypothesis is that niceness/PF is a natural kind that will be relatively easy to get out of any AI smart enough to understand what humans want it to do. This implies that constructing aligned AIs might be reasonably easy, in the sense that most of the work necessary for it will be a natural part of progress in capabilities.
    Long version
    The preference fulfillment hypothesis
    Imagine someone who you genuinely care about. You probably have some kind of a desire to fulfill their preferences in the kind of way that they would like their preferences to be fulfilled.
    It might be very simple ("I like chocolate but they like vanilla, so I would prefer for them to get vanilla ice cream even when I prefer chocolate"), but it might get deep into pretty fundamental differences in preferences and values ("I'm deeply monogamous and me ever being anything else would go against my sacred value, but clearly non-monogamy is what works for my friend and makes them happy so I want them to continue living that way").
    It's not necessarily absolute - some things you might still find really upsetting and you'd still want to override the other person’s preferences in some cases - but you can at least feel the "I want them to satisfy their preferences the way they themselves would like their preferences to be satisfied" thing to some extent.
    I think this kind of desire is something like its own distinct motivation in the human mind. It can easily be suppressed by other kinds of motivations kicking in - e.g. if the other person getting what they wanted made you feel jealous or insecure, or if their preferences involved actively...
    17 min
  • AF - The Preference Fulfillment Hypothesis by Kaj Sotala
    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 Preference Fulfillment Hypothesis, published by Kaj Sotala on February 26, 2023 on The AI Alignment Forum.
    Short version
    Humans have an innate motivation ("preference fulfillment", PF) to fulfill the preferences of those they care about. It corresponds to at least some of the senses of the word "love", as well as related words such as "kindness" and "compassion".
    I hypothesize that it works by simulating the other person and predicting what they would want or how they would like to be treated. PF is when you take your simulation of what other people would want and add an extra component that makes you intrinsically value outcomes that your simulation predicts the other people would prefer.
    I also hypothesize that this is the same kind of simulation that forms our ability to work as a social species in the first place. The "virtual bargaining" model of cooperation suggests that people can coordinate without communication by behaving based on what they would agree to do if they were to explicitly bargain, provided that the resulting agreement is commonly known. A mental simulation process is active in virtually every situation where we interact with other people, such as in a grocery store. People use masks/roles/simulations to determine the right behavior in any social situation, running simulations of how others would react to various behaviors. These simulations involve actual people and various people whose opinions we've internalized and care about. The simulations generally allow people to engage in interactions by acting the way a normal person would in a given situation.
    Once you have this kind of a simulation, constantly running in basically any social situation, it’s likely already exhibiting the PF drive to a weak degree. Doing things that we expect to fulfill other people’s preferences often feels intrinsically nice, even if the person in question was a total stranger. So does wordless coordination in general, as evidenced by the popularity of things like dance.
    If this is true, capabilities progress may then be closely linked to alignment progress. Getting AIs to be better at following instructions requires them to simulate humans better. Once you have an AI that can simulate human preferences, you already have most of the machinery required for having PF as an intrinsic drive. This is contrary to the position that niceness is unnatural. The preference fulfillment hypothesis is that niceness/PF is a natural kind that will be relatively easy to get out of any AI smart enough to understand what humans want it to do. This implies that constructing aligned AIs might be reasonably easy, in the sense that most of the work necessary for it will be a natural part of progress in capabilities.
    Long version
    The preference fulfillment hypothesis
    Imagine someone who you genuinely care about. You probably have some kind of a desire to fulfill their preferences in the kind of way that they would like their preferences to be fulfilled.
    It might be very simple ("I like chocolate but they like vanilla, so I would prefer for them to get vanilla ice cream even when I prefer chocolate"), but it might get deep into pretty fundamental differences in preferences and values ("I'm deeply monogamous and me ever being anything else would go against my sacred value, but clearly non-monogamy is what works for my friend and makes them happy so I want them to continue living that way").
    It's not necessarily absolute - some things you might still find really upsetting and you'd still want to override the other person’s preferences in some cases - but you can at least feel the "I want them to satisfy their preferences the way they themselves would like their preferences to be satisfied" thing to some extent.
    I think this kind of desire is something like its own distinct motivation in t...
    17 min
  • EA - Some more projects I’d like to see by finm
    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: Some more projects I’d like to see, published by finm on February 25, 2023 on The Effective Altruism Forum.
    I recently wrote about some EA projects I’d like to see (also on the EA Forum). This went well!
    I suggested I’d write out a few more half-baked ideas sometime. As with the previous post, I make no claim to originating these ideas, and I’ll try to attribute them where possible. I also make no claim to being confident that all the ideas are any good; just that they seem potentially good without much due diligence. Since many of these are based on shallow dives, I’ve likely missed relevant ongoing projects.
    If you’re considering writing a similar list, at the end of this post I reflect on the value of writing about speculative project ideas in public.
    The order of these ideas is arbitrary and you can read any number of them (i.e. there’s no thread running through them).
    Summary
    Fermi games
    BOTEC tools
    Billionaire impact list
    Forecasting guide
    Short stories about AI futures
    Technical assistance with AI safety verification
    Infosec consultancy for AI labs
    Achievements ledger
    World health dashboard
    The Humanity Times
    Fermi games
    Many people are interested in getting good at making forecasts, and spreading good forecasting practice. Becoming better (more accurate and better calibrated) at forecasting important outcomes — and being willing to make numerical, testable predictions in the first place — often translates into better decisions that bear on those outcomes.
    A close (and similarly underappreciated) neighbor of forecasting is the Fermi estimate, or BOTEC. This is the skill of considering some figure you’re uncertain about, coming up with some sensible model or decomposition into other figures you can begin guessing at, and reaching a guess. It is also the skill of knowing how confident you should be in that guess; or how wide your uncertainty should be. If you have interviewed for some kind of consulting-adjacent job you have likely been asked to (for example) size a market for whiteboard markers; that is an example.
    As well as looking ahead in time, you can answer questions about how the past turned out (‘retrocasting’). It’s hard to make retrocasting seriously competitive, because Google exists, but it is presumably a way to teach forecasting: you tell people about the events that led up to some decision in a niche of history few people are familiar with, and ask: did X happen next? How long did Y persist for? And so on. You can also make estimates without dates involved. Douglas Hofstadter lists some examples in Metamagical Themas:
    How many people die per day on the earth?
    How many passenger-miles are flown each day in the U.S.?
    How many square miles are there in the U.S.? How many of them have you been in?
    How many syllables have been uttered by humans since 1400 A.D.?
    How many moving parts are in the Columbia space shuttle?
    What volume of oil is removed from the earth each year?
    How many barrels of oil are left in the world?
    How many meaningful, grammatical, ten-word sentences are there in English?
    How many insects [.] are now alive? [.] Tigers? Ostriches? Horseshoe crabs?
    How many tons of garbage does New York City put out each week?
    How fast does your hair grow (in miles per hour)?
    What is the weight of the Empire State Building? Of the Hoover Dam? Of a fully loaded jumbo jet?
    Again, most forecasts have a nice feature for evaluation and scoring, which is that before the time where a forecast resolves nobody knows the answer for sure, and after it resolves everyone does, and so there is no way to cheat other than through prophecy.
    This doesn’t typically apply for other kinds of Fermi estimation questions. In particular, things get really interesting where nobody really knows the correct answer, though a correct answer clearly exists. This pays when ‘ground ...
    40 min
  • LW - "Rationalist Discourse" Is Like "Physicist Motors" by Zack M Davis
    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: "Rationalist Discourse" Is Like "Physicist Motors", published by Zack M Davis on February 26, 2023 on LessWrong.
    Imagine being a student of physics, and coming across a blog post proposing a list of guidelines for "physicist motors"—motor designs informed by the knowledge of physicists, unlike ordinary motors.
    Even if most of the things on the list seemed like sensible advice to keep in mind when designing a motor, the framing would seem very odd. The laws of physics describe how energy can be converted into work. To the extent that any motor accomplishes anything, it happens within the laws of physics. There are theoretical ideals describing how motors need to work in principle, like the Carnot engine, but you can't actually build an ideal Carnot engine; real-world electric motors or diesel motors or jet engines all have their own idiosyncratic lore depending on the application and the materials at hand; an engineer who worked on one, might not the be best person to work on another. You might appeal to principles of physics to explain why some particular motor is inefficient or poorly-designed, but you would not speak of physicist motors as if that were a distinct category of thing—and if someone did, you might quietly begin to doubt how much they really knew about physics.
    As a student of rationality, I feel the same way about guidelines for "rationalist discourse." The laws of probability and decision theory describe how information can be converted into optimization power. To the extent that any discourse accomplishes anything, it happens within the laws of rationality.
    Rob Bensinger proposes "Elements of Rationalist Discourse" as a companion to Duncan Sabien's earlier "Basics of Rationalist Discourse". Most of the things on both lists are, indeed, sensible advice that one might do well to keep in mind when arguing with people, but as Bensinger notes, "Probably this new version also won't match 'the basics' as other people perceive them."
    But there's a reason for that: a list of guidelines has the wrong type signature for being "the basics". The actual basics are the principles of rationality one would appeal to explain which guidelines are a good idea: principles like how evidence is the systematic correlation between possible states of your observations and possible states of reality, how you need evidence to locate the correct hypothesis in the space of possibilities, how the quality of your conclusion can only be improved by arguments that have the power to change that conclusion.
    Contemplating these basics, it should be clear that there's just not going to be anything like a unique style of "rationalist discourse", any more than there is a unique "physicist motor." There are theoretical ideals describing how discourse needs to work in principle, like Bayesian reasoners with common priors exchanging probability estimates, but you can't actually build an ideal Bayesian reasoner. Rather, different discourse algorithms (the collective analogue of "cognitive algorithm") leverage the laws of rationality to convert information into optimization in somewhat different ways, depending on the application and the population of interlocutors at hand, much as electric motors and jet engines both leverage the laws of physics to convert energy into work without being identical to each other, and with each requiring their own engineering sub-specialty to design.
    Or to use another classic metaphor, there's also just not going to be a unique martial art. Boxing and karate and ju-jitsu all have their own idiosyncratic lore adapted to different combat circumstances, and a master of one would easily defeat a novice of the other. One might appeal to the laws of physics and the properties of the human body to explain why some particular martial arts school was not teaching their st...
    16 min