The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - What is the likelihood that civilizational collapse would cause technological stagnation? (outdated research) by Luisa Rodriguez
    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 is the likelihood that civilizational collapse would cause technological stagnation? (outdated research), published by Luisa Rodriguez on October 19, 2022 on The Effective Altruism Forum.
    Important note
    This is a rough draft I wrote between October 2019 and April 2020. It’s incomplete, and doesn’t reflect updates in my views in the 2+ years since I worked on it.
    I think there are serious downsides to sharing the draft publicly, because I think some parts of it are likely to be substantially wrong. I’m posting it anyway (with the hope that flagging the potential substantial wrongness will help people be especially skeptical of the conclusions) because I think the benefits outweigh those downsides. The benefits:
    Transparency: I’ve shared the draft with a number of researchers exploring civilizational collapse, and they’ve built off some of the research. It seems bad for an unpublished piece of research to be informing other research without being public (citable, scrutinizable) itself.
    Potential insights: To the extent that not everything in this post is wrong, it seems good for people to be able to easily draw/build on any good arguments in it (rather than have to start from scratch).
    Noticing bad arguments: As per Cunningham’s Law: “The best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer.”
    I think the probability of technological stagnation is somewhat higher than I did when I was working on this piece in earnest for a number of reasons — most of which I don’t have capacity to write up at the moment.
    The biggest reason is probably the risk of extreme, long-lasting climate change. It seems possible that anthropogenic climate change could cause global warming extreme enough that agriculture would become much more difficult than it was for early agriculturalists. Temperatures wouldn’t return to current levels for hundreds of thousands of years, so if the warmer temperatures were much less conducive to recovering agriculture and downstream technological developments, humanity might be stagnant for millennia.
    Acknowledgements
    This research was funded by the Forethought Foundation. It was written by Luisa Rodriguez under the supervision of Lewis Dartnell, and draws heavily on research and conversations with Lewis and Haydn Belfield. Thanks to Max Daniel, Matthew van der Merwe, Rob Wiblin, Howie Lempel, Aron Vallinder, and Kit Harris who provided valuable comments. Thanks also to Will MacAskill for providing guidance and feedback on the larger project. And thanks to Katy Moore for editing this piece, and for drafting the summary.
    All errors are my own.
    Summary
    In this post, I explore the probability a catastrophe that caused civilizational collapse might lead to indefinite technological stagnation (and eventual human extinction) — even if it didn’t cause extinction in the very short term (a topic I covered in What is the likelihood that civilizational collapse would directly lead to human extinction (within decades)?). To do this, I ask three key questions:
    1. If we “re-ran” history, would we see the agricultural and industrial revolutions again?
    If a catastrophe caused a return to hunter-gatherer levels of society, we’d have to undergo the agricultural and industrial revolutions all over again to get back to our current levels of technological civilization.
    How likely is it that we’d overcome those hurdles again?
    Because the first agricultural revolutions happened in multiple places following the stabilization of the climate after the last glacial period, I expect it’s very likely that we could expect to see subsequent agricultural revolutions within years (once the climate is suitable for agriculture).
    I feel less confident, but still fairly optimistic, that another industrial revolution — which has only happened o...
    1 hr 2 min
  • LW - How to Take Over the Universe (in Three Easy Steps) by Writer
    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: How to Take Over the Universe (in Three Easy Steps), published by Writer on October 18, 2022 on LessWrong.
    EA-Forum crosspost
    This is the script of Rational Animations' video linked above. It's about how to take over the universe with amounts of energy and resources that are small compared to what is at our disposal in the Solar System. It's based on this paper, by Anders Sandberg and Stuart Armstrong.
    This is our highest-quality video so far. Below, the script of the video.
    Let’s take over the universe in three easy steps
    Welcome. We’ve heard that you want to take over the universe. Well, you’ve come to the right place. In this video, we’ll show you how to reach as many as four billion galaxies with just a few relatively easy steps and six hours of the Sun’s energy.
    Here’s what you need to do:
    Disassemble mercury and build a Dyson swarm: a multitude of solar captors around the sun.
    Build self-replicating probes
    Launch the self-replicating probes to every reachable galaxy.
    In science fiction, humanity’s expansion into the universe usually starts within our galaxy, the Milky Way. After a new star system is occupied, humanity jumps to the next star, and so on, until we take the whole galaxy. Then, humanity jumps to the nearest galaxy, and the process is repeated.
    This is not how we’re going to do it. Our method is much more efficient. We're going to send self-replicating probes to all the reachable galaxies at once. Getting to the furthest galaxies is not more difficult than getting to the nearest ones. It just takes more time. When a probe arrives at its destination galaxy, it will search for a planet to disassemble, build another Dyson swarm, and launch a new wave of probes to reach every star within the galaxy. And then, each probe in that galaxy will restart civilization.
    We already hear you protest, though: “this whole thing still seems pretty hard to me,” you say. “Especially the “disassembling mercury” part”.
    But actually, none of these steps are as hard as they first appear. If you analyze closely how they could be implemented you’ll find solutions that are much easier than you’d expect. And that’s exactly what Stuart Armstrong and Anders Sandberg do in their paper “Eternity in six hours: intergalactic spreading of intelligent life and sharpening the Fermi paradox.” This video is based on that paper.
    Exploratory engineering and assumptions
    What we mean by “easy” here, is that we will require amounts of energy and resources that are small compared to what is at our disposal in the solar system. Also, the technology required is not extremely far beyond our capabilities today, and the time required for the whole feat is insignificant on cosmic scales.
    Not every potential future technology will make sense to include in our plan to spread to the stars. We need to choose what technologies to use by reasoning in the style of exploratory engineering: trying to figure out what techniques and designs are physically possible and plausibly achievable by human scientists. The requirement “physically possible'' is much easier to comply with than “achievable by human scientists”, therefore, we introduce two assumptions that serve to separate the plausible from the merely possible:
    First: Any process in the natural world can be replicated with human technology. This assumption makes sense in light of the fact that humans have generally been successful at copying or co-opting nature.
    Second: Any task that can be performed can be automated. The rationale for this assumption is that humans have proven to be adept at automating processes, and with advances in AI, we will become even more so.
    Design of the Dyson swarm
    Now, we’ve said we are going to launch probes to every reachable galaxy. This means a hundred million to a hundred billion probes. Where do we get the energy to power...
    19 min
  • EA - My experience experimenting with a bunch of antidepressants I'd never heard of by Luisa Rodriguez
    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: My experience experimenting with a bunch of antidepressants I'd never heard of, published by Luisa Rodriguez on October 19, 2022 on The Effective Altruism Forum.
    This isn't professional medical advice, it's just my experience and amateur knowledge.
    Summary
    I got moderately (and occasionally, severely) depressed about four months into the pandemic.
    I tried a bunch of things to treat it, including therapy, antidepressants, meditation, and a long list of other things.
    I was prescribed the antidepressant sertraline (aka Zoloft) by my NHS GP, and it really helped! But it had a number of side effects that made me very unexcited to stay on it.
    I experimented with a bunch of different types of antidepressants to see if I could find one that worked well and didn’t give me bad side effects.
    I knew this would probably be a long and grueling experiment... and it was. It took over a year, and was really hard, both emotionally and physically.
    But I ended up finding a great one for me, and it now feels well worth the costs.
    I think this kind of experimentation might be a good approach for some people who also experience depression and anxiety.
    Caveats:
    This was expensive. It’s probably much harder to do this without savings or great insurance, or both.
    I have a particularly supportive work environment and partner. It’s probably much harder to do this without those things.
    Things besides antidepressants were also really important to feeling better (and also allowed me to stick with the experiment). Therapy and cognitive behavioral therapy (CBT) were extremely effective for me in treating my imposter syndrome and perfectionism, and in making my depression manageable. Searching for an antidepressant that worked for me probably wouldn’t have gone nearly as well if I hadn’t done it alongside these other practices.
    Antidepressants have not solved all of my problems — and it’s important to set reasonable expectations for how much they can help.
    Context
    I got moderately (and occasionally, severely) depressed about four months into the pandemic.
    This was triggered by:
    The pandemic — including things like: feeling very isolated, not having much personal space, shitty British weather, having to work from home, not being able to see my family for years.
    Genetics — my mom has suffered from moderate-severe depression since she was in her early thirties.
    Imposter syndrome at work — I’d gotten a dream job, but it felt like, sooner or later, my employer would realize I was a total phony. I was afraid to do really any task that might reveal I was actually a big idiot, which made me anxious for a lot of my day-to-day activities.
    My main symptoms were:
    Having continuous low mood or sadness
    Feeling hopeless
    Having very low self-esteem, feeling like I was letting everyone down all the time
    Feeling tearful
    Feeling guilt-ridden
    Feeling irritable and intolerant of others
    Having little motivation or interest in things — finding myself and everyone else really boring
    Finding it difficult to make decisions
    Not getting much enjoyment out of life, including out of things that were previously some of my favorite activities (hiking, seeing close friends)
    Feeling anxious and worried
    Non-pharmacological things I did to treat my depression:
    Weekly therapy with an excellent therapist, who taught me a bunch about CBT (helped a lot)
    Weekly therapy with a couples therapist (sometimes my depression made me especially angst-y about my relationship) (helped a bit)
    Taking time off work (sometimes helped, sometimes made things a bunch worse)
    My depression was a factor in switching jobs (kind of helped)
    Changing my role at work significantly to focus on tasks I found especially exciting and satisfying (helped a lot)
    Spending time in sunnier countries (helped a fair bit)
    Regular exercise (helped a fair bit when I actually had ...
    15 min
  • EA - Does the Trajectory of Pain Matter? by William McAuliffe
    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: Does the Trajectory of Pain Matter?, published by William McAuliffe on October 18, 2022 on The Effective Altruism Forum.
    Summary
    Humans might prefer bad events giving way to good outcomes over good events giving way to bad outcomes, holding net welfare constant.
    If this preference reflects a valid moral intuition that applies to both humans and non-human animals, then painful events at the end of life deserve additional moral weight because there is no scope for a better future.
    It would follow that improving the humaneness of the transport and slaughter of farmed animals is more urgent than the duration and severity of pain they cause would imply.
    There are at least two plausible explanations for why humans would value the temporal order of valenced experiences even if it had no intrinsic moral value:
    People might implicitly assume that improvements in welfare result in greater total welfare because pain relief is itself rewarding (and, as such, might produce more pleasure than is initially apparent).
    Human compassion likely evolved to be sensitive to sudden need, not absolute need
    Unless future research indicates that farmed animals themselves intrinsically value the trajectory of welfare, it is not clear that the welfare issues posed by transport and slaughter deserve additional attention in virtue of their occurence at the end of life.
    Epistemic Status
    This report was the product of about three weeks of research, and as such may overlook relevant research or philosophical considerations that we did not find or have time to consider. Before finalizing the report, we spent about a week's worth of work revising it in light of new comments, but did not add any new sections.
    Introduction
    This report is a postscript to "The Relative Importance of the Severity and Duration of Pain", which was intended to contextualize cause prioritization decisions when sources of suffering they seek to ameliorate differ in duration and severity. The present report addresses an issue that has elicited some discussion in the philosophical literature, but to our knowledge has only recently surfaced in animal ethics: Does the temporal ordering of positive and negative valence states matter beyond its influence on the total amount of pleasure and pain experienced?
    Browning and Veit (2020, 2022) say "yes." Their main premise is that end-of-life welfare is particularly important: "even when two lives contain an equal total sum of positive and negative experiences, it is still worse to have suffering at the end of life" (2022, p. 9). Therefore, the transport and slaughter of farmed animals deserve special consideration over and above the duration and severity of suffering they cause. The following provides an exposition and evaluation of Browning and Veit's argument that the temporal ordering of suffering is morally relevant. To make the implications of their view clear, we contrast it with a simple hedonistic approach in which all that matters morally is the severity and duration of pain and pleasure. That said, none of our arguments against the moral relevance of the trajectory of valenced experience depend on accepting hedonism.
    Shape of a Life
    On a simple hedonistic approach to welfare, the temporal ordering of events per se has no intrinsic importance (de Lazari-Radek & Singer, 2014, Chapter 5, section 3). Timing only matters to the extent it affects future welfare. Browning and Veit (2020, p. 7) give an example of previously higher living standards making current low living standards seem even worse:
    enclosure space is an important determinant of welfare for captive animals. It is likely to be worse for an animal to have the positive experience of a larger enclosure space before the negative experience of a smaller one, because the perceived decrease in quality will add to the negative f...
    20 min
  • EA - Call for applications for Zanzibar residency by Anne Nganga
    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: Call for applications for Zanzibar residency, published by Anne Nganga on October 19, 2022 on The Effective Altruism Forum.
    We are delighted to announce the 2023 Effective Altruism Africa Residency Fellowship, to be hosted on the island of Zanzibar. The Program shall host up to 20 fellows for a 10 week program beginning January 15, 2023 through to March 31, 2023.
    The objective of the residency is to provide structured support and develop community connections between Effective Altruists working on projects improving wellbeing across the African continent. We especially encourage Effective Altruists with personal roots in Africa to apply.
    Fellows will be provided with free modern accommodation and working space with high-speed fibre internet. The residency will be located in the same community as the growing Silicon Zanzibar collective of tech companies, which will facilitate collaboration with existing organizations who have scaled up operations across the African continent.
    If you are interested in applying, please complete the application form here. For additional questions, submit your queries here below.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • EA - ‘Dissolving’ AI Risk – Parameter Uncertainty in AI Future Forecasting by Froolow
    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: ‘Dissolving’ AI Risk – Parameter Uncertainty in AI Future Forecasting, published by Froolow on October 18, 2022 on The Effective Altruism Forum.
    1 - Summary
    This is an entry into the Future Fund AI Worldview contest. The headline figure from this essay is that I calculate the best estimate of the risk of catastrophe due to out-of-control AGI is approximately 1.6%.
    However, the whole point of the essay is that “means are misleading” when dealing with conditional probabilities which have uncertainty spanning multiple orders of magnitude (like AI Risk). My preferred presentation of the results is as per the diagram below, showing it is more probable than not that we live in a world where the risk of Catastrophe due to out-of-control AGI is <3%.
    I completely understand this is a very radical claim, especially in the context of the Future Fund contest considering error bars of 7%-35% to be major updates. I will defend the analysis to a degree that I think suits such a radical claim, and of course make my model available for public scrutiny. All of my analysis is generated with this spreadsheet, which is available to download if you would like to validate any of my results.
    Some general comments on the methods involved in this essay:
    My approach is to apply existing methods of uncertainty analysis to the problem of AI Risk to generate new findings, which I believe is a novel approach in AI Risk but a standard approach in other disciplines with high levels of uncertainty (like cost-effectiveness modelling).
    Rather than a breakthrough insight about AI itself, this essay makes the case that a subtle statistical issue about uncertainty analysis means low-risk worlds are more likely than previously believed. This subtle statistical issue has not been picked up previously because there are systematic weaknesses in applying formal uncertainty analysis to problems in EA / rationalist-adjacent spaces, and the issue is subtle enough that non-systematised intuition alone is unlikely to generate the insight.
    The intuitive explanation for why I report such different results to everyone else is that people’s intuitions are likely to mislead them when dealing with multiple conditional probabilities – the probability of seeing a low-risk output is the probability of seeing any low-risk input when you are stacking conditional probabilities. I avoid my intuitions misleading me by explicitly and systematically investigating uncertainty with a statistical model.
    The results pass several sensitivity and validation checks, so I am confident that the mechanism I describe is real, and should affect AI funding decisions going forward. There are limitations with the exact parameterisation of the model, and I will explain and contextualise those limitations to the extent that I don’t think they fundamentally alter the conclusion that distributions matter a lot more than has previously been understood.
    The conclusion of this essay is that for the average AI-interested individual nothing much will change; everyone was already completely aware that there was at least order-of-magnitude uncertainty in their beliefs, so this essay simply updates people towards the lower end of their existing beliefs. For funding bodies, however, I make some specific recommendations for applying these insights into actionable results:
    We should be devoting significantly more resources to identifying whether we live in a high-risk or low-risk world. The ‘average risk’ (insofar as such a thing actually exists) is sort of academically interesting, but doesn’t help us design strategies to minimise the harm AI will actually do in this world.
    We should be more concerned with systematic investigation of uncertainty when producing forecasts. In particular, the radical results contained in this essay only hold under quite specific...
    1 hr 5 min
  • AF - Response to Katja Grace's AI x-risk counterarguments by Erik Jenner
    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: Response to Katja Grace's AI x-risk counterarguments, published by Erik Jenner on October 19, 2022 on The AI Alignment Forum.
    This is a response to the recent Counterarguments to the basic AI x-risk case ("Counterarguments post" from here on). Based on its reception, it seems that the Counterarguments post makes points that resonate with many, so we're glad that the post was written. But we also think that most of the gaps it describes in the AI x-risk case have already been addressed elsewhere or vanish when using a slightly different version of the AI x-risk argument. None of the points we make are novel, we just thought it would be useful to collect all of them in one reply.
    Before we begin, let us clarify what we are arguing for: we think that current alignment techniques are likely insufficient to prevent an existential catastrophe, i.e. if AI development proceeds without big advances in AI alignment, this would probably lead to an existential catastrophe eventually. In particular, for now, we are not discussing
    how hard open alignment problems are,
    whether these problems will be solved anyway without efforts from longtermists,
    whether an existential catastrophe would happen within a particular time frame,
    what a reasonable all-things-considered p(doom) is.
    These are all important questions, but the main thrust of the Counterarguments post seems to be "maybe this whole x-risk argument is wrong and things are actually just fine", so we are focusing on that aspect.
    Another caveat: we're attempting to present a minimal high-level case for AI x-risk, not to comprehensively list all arguments. This means there are some concepts we don't discuss even though they might become crucial on further reflection. (For example, we don't talk about the concept of coherence, but this could become important when trying to argue that AI systems trained for corrigibility will not necessarily remain corrigible). Anticipating all possible counterarguments and playing out the entire debate tree isn't feasible, so we focus on those that are explicitly made in the Counterarguments post. That said, we do think the arguments we outline are broadly correct and survive under more scrutiny.
    Summary
    Our most important points are as follows:
    The ambiguity around "goal-directedness" can be resolved if we consistently think about AI systems that are able to achieve certain difficult-to-achieve objectives.
    While certain small differences between AI and human values would be fine, there are good reasons to believe that the types of differences we'll get are not the type of small differences we can accept.
    It is not enough to align slightly superhuman AI systems if those systems don't let us reach a stable state where no one builds even more powerful unaligned AI.
    Those would roughly be our one-sentence responses to sections A, B, and C of the Counterarguments post respectively. Below, we first describe some modifications we would make to the basic case for AI x-risk and then go through the points in the Counterarguments post in more detail.
    Notes on the basic case for AI x-risk
    To recap, the Counterarguments post gives the following case for AI x-risk:
    I. If superhuman AI systems are built, any given system is likely to be ‘goal-directed’.
    II. If goal-directed superhuman AI systems are built, their desired outcomes will probably be about as bad as an empty universe by human lights.
    III. If most goal-directed superhuman AI systems have bad goals, the future will very likely be bad.
    This is roughly the case we would make as well, but there are a few modifications and clarifications we'd like to make.
    First, we will focus on a purely behavioral definition of "goal-directed" throughout this post: an AI system is goal-directed if it ensures fairly reliably that some objective will be achieved. For poin...
    24 min
  • EA - [Link post] AI could fuel factory farming—or end it by BrianK
    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: [Link post] AI could fuel factory farming—or end it, published by BrianK on October 18, 2022 on The Effective Altruism Forum.
    Despite the fact that artificial intelligence is already transforming everyday life, most people don’t know much about it.
    Simply put, AI is a method of making software “think” intelligently like the human mind. It can make decisions using real-time data, learning and adapting as it compiles new information. Netflix’s algorithm, for example, stores and analyzes your viewing history to predict what you may be most interested in watching next. Every time Netflix suggests a movie, it uses the data it receives based on whether or not you click play to improve the next time. Soon enough, it might know your preferences better than you do.
    Discussions over the ethics of technology have been going on since long before virtually everyone had a smartphone in their pocket. Some argue that technology itself is value neutral—essentially, that a given technology is only as bad or good as the intentions of the person using it. Others argue that values are built into technologies by the people who design them.
    The ethics of AI is a hot topic these days. Will it make the world better for people, or create more problems than it solves? Nobody knows for sure. But technology affects nonhuman animals as well, and we would be remiss to leave them out of our considerations.
    Read the rest in Fast Company:.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • LW - How To Make Prediction Markets Useful For Alignment Work by johnswentworth
    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: How To Make Prediction Markets Useful For Alignment Work, published by johnswentworth on October 18, 2022 on LessWrong.
    So, I’m an alignment researcher. And I have a lot of prediction-flavored questions which steer my day-to-day efforts. Which recent interpretability papers will turn out, in hindsight a year or two from now, to have been basically correct and high-value? Will this infrabayes stuff turn out to be useful for real-world systems, or is it the kind of abstract math which lost contact with reality and will never reconnect? Is there any mathematical substance to Predictive Processing other than “variational inference is a thing” plus vigorous hand-waving? How about Shard Theory? What kinds of evidence will we see for/against the Natural Abstraction Hypothesis over the next few years? Will HCH-style amplification ever be able to usefully factor problems which the human operator/programmer doesn’t immediately see how to factor? Will some version of this conjecture be proven? Will it be proven by someone else if I don’t focus on it?
    Are there any recent papers/posts which lots of other people expect to have high value in hindsight, but which I haven’t paid attention to?
    Or, to sum that all up in two abstract questions: what will I (or someone else whose judgement I at least find informative) think I should have paid more attention to, in hindsight? What will it turn out, in hindsight, that I should have ignored or moved on from sooner?
    Hmm, I wonder if the prediction markets have anything useful to say here? Let’s go look for AI predictions on Manifold.
    hollywood-level AI-generated feature film by 2026?
    Will an AI get gold on any International Math Olympiad by 2025?
    Will AI wipe out humanity before the year 2100
    Will any Fortune 500 corporation mostly/entirely replace their customer service workforce with AI by 2026?
    So, we’ve got about a gazillion different flavors of AI capabilities questions, with a little bit of dabbling into how-society-will-react-capabilities questions. On the upside, operationalizing things in lots of different ways is exactly what we want. On the downside, approximately-all of the effort is in operationalizing one thing (AI capabilities timelines), and that thing is just not particularly central to day-to-day research decisions. It’s certainly not on the list of questions which first jump to mind when I think of things which would be useful to know to steer my research. Sure, somebody is going to argue in the comments that timelines are relevant for some particular decision, like whether to buy up GPU companies or something, but there’s no way in hell that focusing virtually all alignment-related prediction-market effort on operationalizations of capabilities timelines is actually the value-maximizing strategy here.
    (And while I happened to open up Manifold first, the situation is basically the same in other prediction markets. To a first approximation, the only alignment-adjacent question prediction markets ever weigh in on is timelines.)
    So, here’s my main advice to someone who wants to use prediction markets to help alignment work: imagine that you are an alignment researcher/grantmaker/etc. Concretely imagine your day-to-day: probing weight matrices in nets, conjecturing, reading papers/posts, reviewing proposals, etc. Then, ask what kind of predictions have the highest information value for your work. If the answer is “yet another operationalization of timelines”, then you have probably fucked up somewhere.
    Of course you might also try asking some researchers or grantmakers the same question, though keep in mind the standard user-interview caveat: users do not actually know what they want or would like.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    4 min
  • LW - Untapped Potential at 13-18 by belkarx
    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: Untapped Potential at 13-18, published by belkarx on October 18, 2022 on LessWrong.
    The Western world is facing a crisis of stretched adolescence, keeping children exclusively "learning" for longer than necessary.
    Imagine that you are a student, around 14 years old. You fall into the top 5%, not the extremely motivated startup founder or math prodigy that you hear about on the news, but comfortably more intelligent and curious than the others. The system sucks some of your free time with busywork, but you build up a knowledge base, and start working on small projects related to your area of interest. Now what?
    Young people are capable. Their neurons are fully myelinated, their worldview is fresh, their passion is strong. But we don't use them.
    In fact we make it difficult for them to thrive: labor laws make it bothersome for companies to take on anyone under 18, even if they can demonstrate competence, so that path is rocky. Research internships and freelancing are more viable, but still extremely underused. Perhaps it's because the school system drowns people in busywork, forcing them to study things just to forget them within a few days, and personal projects are stifled under the workload. Competent teens fall into apathy because the work is too easy, application of anything more difficult is inaccessible, and there's only so much knowledge-gathering they can do before they want to do something.
    We need to put frameworks in place to take advantage of capable high school students. The current assumption is that the most qualified will find a way, and while that may be true, that leaves behind a huge untapped resource. Not everyone is hyper-ambitious but if provided engaging work or more exposure to interesting problems, they can be both useful and satisfied, and the world benefits.
    The Internet provides some of these opportunities, however working with "professionals" on larger problems is an experience that cannot be replicated. The combination of years of wisdom and accumulated knowledge with a modern view of a problem without the inset biases of current models has serious potential.
    TL;DR: Teenagers are underappreciated and a vast and unique source of cognitive power that should be more cleanly introduced into the working world.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min