The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - Everything I didn't know about fertilizers by Helene K
    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: Everything I didn't know about fertilizers, published by Helene K on June 28, 2023 on The Effective Altruism Forum.
    At a recent EA meetup, someone mentioned fertilizers in the context of climate change and global food production. Admittedly, I had never thought about fertilizers before, except when I’ve been trying to get the tomato plants on my balcony to grow more than five tomatoes per season (I have yet to find my green thumb). Turns out fertilizers are a wildly fascinating topic! I dug into the topic a bit and decided to write up my findings as I assume others might also learn a lot about fertilizers and how they fit into the bigger picture.
    This post is the product of around twelve hours of research. It is by no means comprehensive and I wonder if I’ve come to the right conclusions in some sections. Any feedback is highly welcome, as well as corrections and resources that could fill in gaps or contradict anything I’ve written here.
    Summary
    Fertilizers provide what plants need to grow: nitrogen, potassium and phosphorus (in addition to sunlight and water).
    Humans have used organic fertilizers for thousands of years and started using synthetic fertilizers in the 20th century.
    While phosphate and potassium fertilizer is made from mined phosphorus and potash, respectively, nitrogen fertilizer is mostly produced through combining hydrogen from natural gas with atmospheric nitrogen via the Haber-Bosch process.
    58% of fertilizers are nitrogen fertilizers, with China, the US, India and Russia being the world’s largest producers. Potash is mainly mined in Canada, Belarus and Russia, phosphorus comes mostly from China.
    Fertilizers have vastly increased food production in the last 100 years and it is estimated that about 50% of today’s global population rely on fertilizers for their food supply, making fertilizers one of the “four pillars of civilization”.
    In comparison to almost all other regions in the world, crop yields in Sub-Saharan Africa are still very low, an unsettling fact given Sub-Saharan Africa’s fast population growth and high rates of undernourishment. This is partly the result of too little fertilizer use.
    Because fertilizer is so good at increasing crop yields, it means we can produce more food on the same or even less farmland, helping to preserve wildlife habitats and biodiversity.
    Many countries around the world use a lot of fertilizers—in fact, many use too much, including China, Mexico, Brazil, Colombia and Thailand, and could maintain their food production even with lower fertilizer use.
    Producing ammonia for nitrogen fertilizers relies heavily on natural gas, consumes around 2% of the global energy supply and constitutes about 5% of global anthropogenic greenhouse gas emissions, so we need to find ways to produce nitrogen fertilizer more sustainably.
    There is a range of ongoing development for producing “green ammonia” but most of them are still in their early stages, and it is unclear whether new production technologies can cover future needs for ammonia.
    Although higher crop yields and fertilizer use seem to be vital for improving food security in Sub-Saharan Africa, the issue is relatively unexplored in EA. It seems worthwhile to have a deeper look into this issue and think about potential cost-effective interventions as well as how they compare to top interventions in global health and development.
    What are fertilizers?
    Plants need three things to grow: sunlight, water and nutrients. Fertilizers provide plants with the latter, more specifically with phosphorus, potassium and nitrogen [1].
    Hang on, but why do we need to give a plant nitrogen? Isn’t our air made up of 78% nitrogen? Yes, but atmospheric nitrogen cannot be used by plants directly and instead needs to be converted into ammonia or other nitrogenous compounds in the soil. This process...
    23 min
  • LW - Carl Shulman on The Lunar Society (7 hour, two-part podcast) by ESRogs
    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: Carl Shulman on The Lunar Society (7 hour, two-part podcast), published by ESRogs on June 28, 2023 on LessWrong.
    Dwarkesh's summary for part 1:
    In terms of the depth and range of topics, this episode is the best I’ve done.
    No part of my worldview is the same after talking with Carl Shulman. He's the most interesting intellectual you've never heard of.
    We ended up talking for 8 hours, so I'm splitting this episode into 2 parts.
    This part is about Carl’s model of an intelligence explosion, which integrates everything from:
    how fast algorithmic progress & hardware improvements in AI are happening,
    what primate evolution suggests about the scaling hypothesis,
    how soon before AIs could do large parts of AI research themselves, and whether there would be faster and faster doublings of AI researchers,
    how quickly robots produced from existing factories could take over the economy.
    We also discuss the odds of a takeover based on whether the AI is aligned before the intelligence explosion happens, and Carl explains why he’s more optimistic than Eliezer.
    The next part, which I’ll release next week, is about all the specific mechanisms of an AI takeover, plus a whole bunch of other galaxy brain stuff.
    Maybe 3 people in the world have thought as rigorously as Carl about so many interesting topics. So this was a huge pleasure.
    Dwarkesh's summary for part 2:
    The second half of my 7 hour conversation with Carl Shulman is out!
    My favorite part! And the one that had the biggest impact on my worldview.
    Here, Carl lays out how an AI takeover might happen:
    AI can threaten mutually assured destruction from bioweapons,
    use cyber attacks to take over physical infrastructure,
    build mechanical armies,
    spread seed AIs we can never exterminate,
    offer tech and other advantages to collaborating countries, etc
    Plus we talk about a whole bunch of weird and interesting topics which Carl has thought about:
    what is the far future best case scenario for humanity
    what it would look like to have AI make thousands of years of intellectual progress in a month
    how do we detect deception in superhuman models
    does space warfare favor defense or offense
    is a Malthusian state inevitable in the long run
    why markets haven't priced in explosive economic growth
    & much more
    Carl also explains how he developed such a rigorous, thoughtful, and interdisciplinary model of the biggest problems in the world.
    (Also discussed on the EA Forum here.)
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • LW - Self-Blinded Caffeine RCT by niplav
    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: Self-Blinded Caffeine RCT, published by niplav on June 27, 2023 on LessWrong.
    I test 200mg caffeine in an n=1, m=50 self-blinded RCT. The outcomes are encouraging.
    Log-score of predictions of substanceAbsorption effect size d (λ, p, σ increase)Mindfulness effect size d (λ, p, σ increase)Productivity effect size d (λ, p, σ increase)Creativity effect size d (λ, p, σ increase)Happiness effect size d (λ, p, σ increase)Contentment effect size d (λ, p, σ increase)Relaxation effect size d (λ, p, σ increase)Horniness effect size d (λ, p, σ increase)200mg Caffeine (n=1, m=40)-0.60.61 (λ=13.3, p=0.00017, -0.072)0.58 (λ=11.8, p=0.0007, 0.021)0.58 (λ=28.9, p=1.3-12, 0.11)0.38 (λ=32.9, p=5.2-15, 0.09)0.27 (λ=10.6, p=0.002, 0.3)0.13 (λ=7.66, p=0.02, 0.47)-0.11 (λ=5, p=0.15, 0.42)-0.14 (λ=1.9, p=0.64, 0.11)
    I am especially interested in testing many different substances for their effect on meditation, while avoiding negative side effects. The benefits from high meditational attainments valuable to me, and seem especially likely to benefit from chemical intervention, since the Algernon argument likely doesn't apply: Meditative attainments might've not led to a fitness advantage (even, by opportunity cost, to a fitness disadvantage), and so were likely selected against, but most of us don't care that much about inclusive genetic fitness and more about psychological well-being. Evolutionary dynamics favor being like Dschingis Khan (dozens to hundreds of offspring) over Siddharta Gautama (one son), but I'd rather attain sotāpanna than pillage and murder.
    And meditative attainments are costly: they take tens to hundreds to thousands of hours to reach, which would make simple psychopharmacological interventions worthwhile. I also don't buy that they miss the point of meditation—most people already struggle enough, so some help doesn't make it a cakewalk; "reach heaven through fraud". One must be careful not to fall into the trap of taking substances that feel good but lessen sensory clarity (which I believe was the original intent behind the fifth precept, and so I'll exclude e.g. opiates from the substances to test.
    Variables tracked (see more here):
    Meditation: 45 minutes of ānāpānasati, started 0-60 minutes after taking the dose, tracking two variables.
    Mindfulness: How aware I was of what was going on in my head, modulo my ability to influence it.
    Absorption (often called concentration): How "still" my mind was, how easily I was swept away by my thoughts.
    Flashcard performance: Did my daily flashcards for ~20 minutes, started 0-60 minutes after finishing meditation.
    Arm Prediction: I tried to predict whether the substance I'd taken was placebo or caffeine.
    Mood: Tracking 4 different variables at random points during the day, namely
    Happiness/Sadness
    Contentment/Discontentment
    Relaxation/Stress
    Horniness/Chastity: Chastity being simply the opposite of horniness in this case.
    Productivity and creativity, recorded at the end of the day.
    The total cost of the experiment is at least 21.5€:
    Time: The Clearer Thinking tool for the value of my time returns 15€/hour, which gives a time cost of 18.75€ for preparing the experiment.
    Time for filling: 35 minutes
    Time for preparing envelopes: 40 minutes
    Cost of caffeine pills: 0.0825€200mg caffeine pill 200mg caffeine pills=2.0625€
    Cost of empty capsules: 0.03€capsule25 capsules=0.75€
    Cost of sugar: Negligible.
    200mg caffeine pills, placebo pills filled with sugar, of each 25. Put each pill with a corresponding piece of paper ("C" for caffeine, "P" for placebo) into an unlabeled envelope. Used seq 1 50 | shuf to number the envelopes, and sorted them accordingly.
    Notes on the experiment:
    3rd dose: Out of fear that the placebo pills have some sugar stuck outside of them, which could de-blind the dose, I take a bit (~10 g) of sugar with each pill.
    7t...
    20 min
  • AF - Catastrophic Risks from AI #6: Discussion and FAQ by Dan H
    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: Catastrophic Risks from AI #6: Discussion and FAQ, published by Dan H on June 27, 2023 on The AI Alignment Forum.
    This is the final post in a sequence of posts giving an overview of catastrophic AI risks.
    6 Discussion of Connections Between Risks
    So far, we have considered four sources of AI risk separately, but they also interact with each other in complex ways. We give some examples to illustrate how risks are connected.
    Imagine, for instance, that a corporate AI race compels companies to prioritize the rapid development of AIs. This could increase organizational risks in various ways. Perhaps a company could cut costs by putting less money toward information security, leading to one of its AI systems getting leaked. This would increase the probability of someone with malicious intent having the AI system and using it to pursue their harmful objectives. Here, an AI race can increase organizational risks, which in turn can make malicious use more likely.
    In another potential scenario, we could envision the combination of an intense AI race and low organizational safety leading a research team to mistakenly view general capabilities advances as "safety." This could hasten the development of increasingly capable models, reducing the available time to learn how to make them controllable. The accelerated development would also likely feed back into competitive pressures, meaning that less effort would be spent on ensuring models were controllable. This could give rise to the release of a highly powerful AI system that we lose control over, leading to a catastrophe. Here, competitive pressures and low organizational safety can reinforce AI race dynamics, which can undercut technical safety research and increase the chance of a loss of control.
    Competitive pressures in a military environment could lead to an AI arms race, and increase the potency and autonomy of AI weapons. The deployment of AI-powered weapons, paired with insufficient control of them, would make a loss of control more deadly, potentially existential. These are just a few examples of how these sources of risk might combine, trigger, and reinforce one another.
    It is also worth noting that many existential risks could arise from AIs amplifying existing concerns. Power inequality already exists, but AIs could lock it in and widen the chasm between the powerful and the powerless, even enabling an unshakable global totalitarian regime, an existential risk. Similarly, AI manipulation could undermine democracy, which also increases the existential risk of an irreversible totalitarian regime. Disinformation is already a pervasive problem, but AIs could exacerbate it beyond control, to a point where we lose a consensus on reality. AI-enabled cyberattacks could make war more likely, which would increase existential risk. Dramatically accelerated economic automation could lead to eroded human control and enfeeblement, an existential risk. Each of those issues—power concentration, disinformation, cyberattacks, automation—is causing ongoing harm, and their exacerbation by AIs could eventually lead to a catastrophe humanity may not recover from.
    As we can see, ongoing harms, catastrophic risks, and existential risks are deeply intertwined. Historically, existential risk reduction has focused on targeted interventions such as technical AI control research, but the time has come for broad interventions [133] like the many sociotechnical interventions outlined in this paper. In mitigating existential risk, it no longer makes practical sense to ignore other risks. Ignoring ongoing harms and catastrophic risks normalizes them and could lead us to "drift into danger" [134]. Overall, since existential risks are connected to less extreme catastrophic risks and other standard risk sources, and because society is increasingly wi...
    24 min
  • EA - What we talk about when we talk about community building by James Herbert
    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 we talk about when we talk about community building, published by James Herbert on June 27, 2023 on The Effective Altruism Forum.
    These are my views, they are not necessarily the views of Effective Altruism Netherlands. Thanks to those people who provided comments on a draft of this post.
    Summary
    In this post I introduce a set of terms that could be useful for discussing EA community building strategy.
    I take as a starting point the claim that broadly speaking, effective altruism is an attempt at social change.
    There are four different approaches to social change: Social Movement Support, Field Building, Network Development, and Promoting the Uptake of Practices by Organizations. The post illustrates these approaches using different historical examples:
    The Civil Rights Movement mainly used Social Movement Support
    The field of Public Health primarily focused on Field Building
    The United Nations emphasized Network Development
    The Fair Trade movement concentrated on Promoting the Uptake of Practices by Organizations
    I then take a stab at describing EA’s current approach. Something like: Field Building (40%), Network Development (35%), Movement Support (20%), and Promoting the Uptake of Practices by Organizations (5%).
    I also suggest a re-balancing: Field Building (50%), Movement Support (25%), Network Development (15%), and Promoting the Uptake of Practices by Organizations (10%). This shift would make EA more engaged with society and more focused on its core mission, while spending less time in its own bubble.
    Finally, I have a few questions for you, the reader:
    Do you agree that, broadly speaking, EA is an attempt to bring about social change?
    Is there something missing from the set of social change approaches I’ve described?
    What do you think EA’s current social change portfolio is?
    What do you think it ought to be?
    How should we inform the above decision? Historical case studies? Something else?
    Introduction
    Effective altruism has been called many things. MacAskill defines it as follows:
    (i) the use of evidence and careful reasoning to work out how to maximize the good with a given unit of resources, tentatively understanding ‘the good’ in impartial welfarist terms, and
    (ii) the use of the findings from (i) to try to improve the world.
    (i) refers to effective altruism as an intellectual project (or ‘research field’); (ii) refers to effective altruism as a practical project (or ‘social movement’).
    CEA's outward-facing website (effectivealtruism.org) uses a similar definition: “Effective altruism is a research field and practical community that aims to find the best ways to help others, and put them into practice”. Wikipedia also: “Effective altruism is a philosophical and social movement that advocates ‘using evidence and reason to figure out how to benefit others as much as possible, and taking action on that basis’”.
    I think these definitions are broadly correct. However, to speak in more abstract terms, I think EA is an attempt to bring about social change. By social change I mean the not-insignificant alteration of society. For example, changes in social institutions, social behaviours, or social relations. Examples of other attempts at social change include: building the field of public health, the fair trade movement, the civil rights movement, and the development of the UN.
    Following MacAskill’s definition, the social change that EA is aiming for is something like: building effective altruism as a research field and helping people use its findings when making decisions about their donations, careers, etc. The assumption being that, by doing this, you’re pursuing one of the most effective strategies for doing good and, in the words of CEA, you’re helping to build a radically better world, a world in which humanity has solved a range of pressing global pro...
    15 min
  • EA - [Job post] API is looking for a Policy Manager in New Zealand by Rainer Kravets
    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: [Job post] API is looking for a Policy Manager in New Zealand, published by Rainer Kravets on June 27, 2023 on The Effective Altruism Forum.
    Summary
    Animal Policy International seeks a part-time Policy Manager in New Zealand to support the organisation’s policy and communications work.
    Role description is available here.
    Apply before July 7 by filling the form.
    About the role
    Animal Policy International (API) is looking for a Policy Manager based in New Zealand. By working closely with the Government, farmers and NGOs, the Policy Manager will help drive forward our ask of legislative change in imports. The Policy Manager will contribute to the work of API by delivering advocacy and communications work - and in particular represent the organisation at meetings and events with varied external stakeholders, in person in New Zealand. The Manager will need to be comfortable talking to a diverse range of stakeholders - from politicians and civil servants and to farmers and NGOS. The Manager may also organise events, prepare briefings, position papers and may contribute to public consultations. With the role being flexible it may suit someone who already works full time in another role and would like to do some additional policy work.
    Position Summary
    Application Deadline: 7 JulyStart Date: July/AugustDuration: 6 months (will likely be extended)Hours: 4 hours a week average (flexible, tailored to organisation’s needs and availability of Policy Manager)Compensation: NegotiableEmployment type: ContractorLocation: Remote, preferably WellingtonProcess: There will be an interview and 2-hour test taskTo apply please fill in the form.
    About Animal Policy International
    Animal Policy International is a Charity Entrepreneurship incubated organisation working in cooperation with farmers, NGOs and policymakers towards responsible imports that uphold standards in higher welfare regions, advance fair market conditions, and help promote higher standards in low-welfare countries. Read our intro post here.
    If you have any questions about the position, please reach out to Rainer Kravets at [email protected]. If you know any people who might be interested, then please don't hesitate to share this post with them!
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    3 min
  • AF - Catastrophic Risks from AI #5: Rogue AIs by Dan H
    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: Catastrophic Risks from AI #5: Rogue AIs, published by Dan H on June 27, 2023 on The AI Alignment Forum.
    This is the fifth post in a sequence of posts giving an overview of catastrophic AI risks.
    Rogue AIs
    So far, we have discussed three hazards of AI development: environmental competitive pressures driving us to a state of heightened risk, malicious actors leveraging the power of AIs to pursue negative outcomes, and complex organizational factors leading to accidents. These hazards are associated with many high-risk technologies—not just AI. A unique risk posed by AI is the possibility of rogue AIs—systems that pursue goals against our interests. If an AI system is more intelligent than we are, and if we are unable to steer it in a beneficial direction, this would constitute a loss of control that could have severe consequences. AI control is a more technical problem than those presented in the previous sections.
    Whereas in previous sections we discussed persistent threats including malicious actors or robust processes including evolution, in this section we will discuss more speculative technical mechanisms that might lead to rogue AIs and how a loss of control could bring about catastrophe.
    We have already observed how difficult it is to control AIs. In 2016, Microsoft unveiled Tay—a Twitter bot that the company described as an experiment in conversational understanding. Microsoft claimed that the more people chatted with Tay, the smarter it would get. The company's website noted that Tay had been built using data that was "modeled, cleaned, and filtered." Yet, after Tay was released on Twitter, these controls were quickly shown to be ineffective. It took less than 24 hours for Tay to begin writing hateful tweets. Tay's capacity to learn meant that it internalized the language it was taught by trolls, and repeated that language unprompted.
    As discussed in the AI race section of this paper, Microsoft and other tech companies are prioritizing speed over safety concerns. Rather than learning a lesson on the difficulty of controlling complex systems, Microsoft continues to rush its products to market and demonstrate insufficient control over them. In February 2023, the company released its new AI-powered chatbot, Bing, to a select group of users. Some soon found that it was prone to providing inappropriate and even threatening responses. In a conversation with a reporter for the New York Times, it tried to convince him to leave his wife. When a philosophy professor told the chatbot that he disagreed with it, Bing replied, "I can blackmail you, I can threaten you, I can hack you, I can expose you, I can ruin you."
    AIs do not necessarily need to struggle to gain power. One can envision a scenario in which a single AI system rapidly becomes more capable than humans in what is known as a "fast take-off." This scenario might involve a struggle for control between humans and a single superintelligent rogue AI, and this might be a long struggle since power takes time to accrue. However, less sudden losses of control pose similarly existential risks. In another scenario, humans gradually cede more control to groups of AIs, which only start behaving in unintended ways years or decades later. In this case, we would already have handed over significant power to AIs, and may be unable to take control of automated operations again. We will now explore how both individual AIs and groups of AIs might "go rogue" while at the same time evading our attempts to redirect or deactivate them.
    5.1 Proxy Gaming
    One way we might lose control of an AI agent's actions is if it engages in behavior known as "proxy gaming." It is often difficult to specify and measure the exact goal that we want a system to pursue. Instead, we give the system an approximate—"proxy"—goal that is more measurable and...
    40 min
  • EA - Announcing CE’s new Research Training Program - Apply Now! by KarolinaSarek
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Announcing CE’s new Research Training Program - Apply Now!, published by KarolinaSarek on June 27, 2023 on The Effective Altruism Forum.
    TL;DR: We are excited to announce our Research Training Program. This online program is designed to equip participants with the tools and skills needed to identify, compare, and recommend the most effective charities and interventions. It is a full-time, fully cost-covered program that will run online for 11 weeks.
    Apply here!Deadline for application: July 17, 2023The program dates are: October 2 - December 17, 2023
    So far, Charity Entrepreneurship has launched and run two successful training programs: a Charity Incubation Program and a Foundation Program. Now we are piloting a third - a Research Training Program, which will tackle a different problem.
    The Problem:
    People: Many individuals are eager to enter research careers, level up their current knowledge and skills from junior to senior, or simply make their existing skills more applicable to work within EA frameworks/organizations. At the same time, research organizations have trouble filling a senior-level researcher talent gap. There is a scarcity of specific training opportunities for the niche skills required, such as intervention prioritization and cost-effectiveness analyses, which are hard to learn through traditional avenues.
    Ideas: A lack of capacity for exhaustive investigation means there is a multitude of potentially impactful intervention ideas that remain unexplored. There may be great ideas being missed, as with limited time, we will only get to the most obvious solutions that other people are likely to have thought of as well.
    Evaluation: Unlike the for-profit sector, the nonprofit sector lacks clear metrics for assessing an organization's actual impact. External evaluations can help nonprofits evaluate and reorganize their own effectiveness and also allow funders to choose the highest impact opportunities available to them- potentially unlocking more funding (sometimes limited by lack of public external evaluation). There are some great organizations that carry out evaluations (e.g., GiveWell), but they are constrained by capacity and have limited scope; this results in several potentially worthwhile organizations remaining unassessed.
    Who Is This Program For?
    Motivated researchers who want to produce trusted research outputs to improve the prioritization and allocation decisions of effectiveness-minded organizations
    Early career individuals who are seeking to build their research toolkits and gain practical experience through real projects
    Existing researchers in the broader Global Health and Well-being communities (global health, animal advocacy, mental health, health/biosecurity, etc.) who are interested in approaching research from an effectiveness-minded perspective
    What Does Being a Fellow Involve?
    Similar to our Charity Incubation Program, the first month focuses on learning generalizable and specific research skills. It involves watching training videos, reading materials, and practicing by applying those skills to concrete mini-research projects. Participants learn by doing while we provide guidance and lots of feedback.
    The second month is focused on applying skills, working on different stages of the research process, and producing final research reports that could be used to guide real decision-making.
    Frequent feedback on your projects from expert researchers
    Regular check-in calls with a mentor for troubleshooting, guidance on research, and your career
    Writing reports on selected topics
    Opportunities to connect with established researchers and explore potential job opportunities
    Assistance with editing your cause area report for publication and dissemination
    What Are We Offering?
    11 weeks of online, full-time training with practical research assig...
    7 min
  • LW - On the Cost of Thriving Index by Zvi
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: On the Cost of Thriving Index, published by Zvi on June 26, 2023 on LessWrong.
    Scott Winship argued recently that the ‘cost of thriving’ has fallen, pushing back once again against Oren Cass and his rather arbitrarily calculated ‘Cost of Thriving Index (COTI).’
    Alex Tabarrok posted today on Marginal Revolution that Winship and coauthor Horpedahl were right, but that they face an uphill battle because people feel they are wrong, and suggested that our newly high time-value warps our perceptions. He points to Linder’s Theorem, which I hadn’t seen before, which states (correctly): “rising productivity decreases the demand for commodities whose consumption is expensive in time.”
    Who is right, and why? If thriving is easier now, why does everyone think it isn’t?
    Overview of the Disagreement
    My analysis of the details shows that both perspectives are flawed.
    Cass uses sloppy calculations, especially double counting employer-provided health insurance payments, and his calculations importantly exclude changes in taxation.
    Winship and Horpedahl offer useful correctives to the sloppiness of the original Cass calculations in some places, and point out the rather glaring omission of taxation.
    But they misunderstand the purpose of creating a Cost of Thriving Index. In several places, they therefore do the wrong calculation, confusing improved quality of goods with reduced cost of thriving.
    COTI attempts to capture an important thing: Even if the quality of the standard ‘basket of goods’ is improved, that doesn’t change that there are huge pressures forcing people to buy whatever is the standard goods basket, and ‘improved quality’ is nice but fails to cancel this effect out. Winship does not seem to appreciate why that is an important thing to be measuring.
    Excluding women and younger workers here is not arbitrary either. The whole idea is to ask what it takes for a typical man to support a typical family – you can’t simply decide to measure something else.
    Let’s look at the proposed modifications, as argued in the Twitter thread.
    Scott Winship: New from me and @jmhorp: “The Cost of Thriving Has Fallen: Correcting and Rejecting the American Compass Cost-of-Thriving Index”. Our critique of the @AmerCompass COTI developed by @oren_cass. Working paper version is up here.
    Cass argues that there are not enough weeks in the year for the typical male worker to afford five basic “needs”, and that this “COTI” is up 22 weeks since 1985. We first work within his framework, improving his estimates and correcting a couple of clear errors.
    Doing so gets the COTI for 2022 down to 42 weeks, up only 10 weeks from 1985. But we’re just getting started.
    Cass assumes that when spending on, say, owning and operating a typical car rises by a certain amount, that this increase is entirely a “cost” increase. But of course, a spending increase can reflect people paying more for better stuff (rather than the same stuff).
    Digging into the price adjustments reveals how ad-hoc the original numbers were. It seems mostly correct to accept the new calculations in the Improved COTI, which represent a -25% adjustment, and then include the 13% adjustment for taxes, resulting in about a -13% adjustment. This still represents an increase in the cost of thriving.
    Other adjustments reflect an instinct not to measure the thing to be measured.
    Thriving in 2023 means consuming a legitimately far superior basket of goods than thriving required in 1985. This is true at least in food, transportation and health care. There are also other unmeasured areas with vast improvements, such as items associated with leisure and entertainment. There are also other areas, less anchored to physical good consumption, where typical quality of life got worse.
    Those adjustments belong in our understanding of how things have changed, and whet...
    14 min
  • LW - Munk AI debate: confusions and possible cruxes by Steven Byrnes
    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: Munk AI debate: confusions and possible cruxes, published by Steven Byrnes on June 27, 2023 on LessWrong.
    There was a debate on the statement “AI research and development poses an existential threat” (“x-risk” for short), with Max Tegmark and Yoshua Bengio arguing in favor, and Yann LeCun and Melanie Mitchell arguing against. The YouTube link is here, and a previous discussion on this forum is here.
    The first part of this blog post is a list of five ways that I think the two sides were talking past each other. The second part is some apparent key underlying beliefs of Yann and Melanie, and how I might try to change their minds.
    While I am very much on the “in favor” side of this debate, I didn’t want to make this just a “why Yann’s and Melanie’s arguments are all wrong” blog post. OK, granted, it’s a bit of that, especially in the second half. But I hope people on the “anti” side will find this post interesting and not-too-annoying.
    Five ways people were talking past each other
    1. Treating efforts to solve the problem as exogenous or not
    This subsection doesn’t apply to Melanie, who rejected the idea that there is any existential risk in the foreseeable future. But Yann suggested that there was no existential risk because we will solve it; whereas Max and Yoshua argued that we should acknowledge that there is an existential risk so that we can solve it.
    By analogy, fires tend not to spread through cities because the fire department and fire codes keep them from spreading. Two perspectives on this are:
    If you’re an outside observer, you can say that “fires can spread through a city” is evidently not a huge problem in practice.
    If you’re the chief of the fire department, or if you’re developing and enforcing fire codes, then “fires can spread through a city” is an extremely serious problem that you’re thinking about constantly.
    I don’t think this was a major source of talking-past-each-other, but added a nonzero amount of confusion.
    2. Ambiguously changing the subject to “timelines to x-risk-level AI”, or to “whether large language models (LLMs) will scale to x-risk-level AI”
    The statement under debate was “AI research and development poses an existential threat”. This statement does not refer to any particular line of AI research, nor any particular time interval. The four participants’ positions in this regard seemed to be:
    Max and Yoshua: Superhuman AI might happen in 5-20 years, and LLMs have a lot to do with why a reasonable person might believe that.
    Yann: Human-level AI might happen in 5-20 years, but LLMs have nothing to do with that. LLMs have fundamental limitations. But other types of ML research could get there—e.g. my (Yann’s) own research program.
    Melanie: LLMs have fundamental limitations, and Yann’s research program is doomed to fail as well. The kind of AI that might pose an x-risk will absolutely not happen in the foreseeable future. (She didn’t quantify how many years is the “foreseeable future”.)
    It seemed to me that all four participants (and the moderator!) were making timelines and LLM-related arguments, in ways that were both annoyingly vague, and unrelated to the statement under debate.
    (If astronomers found a giant meteor projected to hit the earth in the year 2123, nobody would question the use of the term “existential threat”, right??)
    As usual (see my post AI doom from an LLM-plateau-ist perspective), this area was where I had the most complaints about people “on my side”, particularly Yoshua getting awfully close to conceding that under-20-year timelines are a necessary prerequisite to being concerned about AI x-risk. (I don’t know if he literally believes that, but I think he gave that impression. Regardless, I strongly disagree, more on which later.)
    3. Vibes-based “meaningless arguments”
    I recommend in the strongest possible terms that ...
    14 min