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: "Should have been hired" Prizes, published by Larks on January 19, 2022 on The Effective Altruism Forum.

Retrospective Employment Bounties

Funding socially beneficial work in retrospect is an idea that has been frequently discussed (e.g. here) but has had a variety of problems in practice, including the difficulty in concretely identifying ahead of time what will be in retrospect eligible.

A possible solution to this would be to exploit an existing area where such organisations already make such deliberations: hiring. At present orgs evaluate researchers, estimating their expected research contributions to that organisation’s agenda. If this estimate is high, the researcher is hired. If it is low, the researcher is rejected.

Sometimes these estimates are wrong, however, and sometimes the candidate might have good reason to believe this. Perhaps he knows he underperformed his true potential at the interview, or struggled to explain his ideas, or simply thinks the interviewers were mistaken about some crucial consideration. In these cases it might be good if the researcher was incentivised to work on the research anyway, as if he had been hired.

This could be achieved if the organisation credibly promised to pay rejected applicants in retrospect if their work turned out to be sufficiently good that they should have been hired. In order to account for risk, the organisation might offer a multiple of the salary + benefits they would have offered. For example, if the organisation thinks their marginal hire (that they actually hired) has a 50% chance of producing work of quality X, they should be willing to pay 2x compensation for any rejected candidate who nonetheless produces quality X work.

The organisation could additionally hire the researcher at this point, but this is not a necessary component. Additionally, this system could be combined into a broader Impact Certificates program - it would essentially be a commitment to purchase certificates at a certain price - but can stand on its own. Some mechanism may be required to ensure that multiple organisations do not fund the same work.

Advantages of this approach:

Incentivise additional work

The prospect of compensation (and vindication!) may induce additional research effort on useful topics.

Guide research

Independent researchers will have some sense of what to work on: instead of ‘produce valuable things’, they have the slightly more concrete ‘things this organisations thinks are valuable’.

Pay for Performance

Only research successes require payment.

Existing Evaluation Criteria

Organisations already make hiring decisions, so there is relatively little novel evaluative work required.

Save overhead

If rejected candidates can produce good work independently, you are relieved of the costs of managing them.

Fail gracefully

If no organisations adopt this policy, or no spurned applicants pursue it, little is lost.

Disadvantages of this approach:

Waste Time

Low quality, correctly rejected candidates might waste their time on a quixotic quest for vindication.

Limited Scope

This can only induce counterfactual effort from rejected candidates, a relatively small pool.

If organisations are good at hiring, this may be a weak pool.

Dissatisfaction

Being rejected from such a prize could add insult to the injury of the initial rejection.

Some people might disagree with the retrospective evaluation and feel cheated.

Risk aversion / diminishing marginal utility

A guarantee of Y dollars is more valuable than a 50% chance of 2Y, so this produces less utility for researchers than if they had been hired (though this was not on the table as they were rejected).

Duplication of effort

Multiple researchers may require compensation for the same insights in a way that could be avoided if the organisation was directly allocating their labour.

F...

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: Omicron Post #15, published by Zvi on January 19, 2022 on LessWrong.

Omicron has fully taken over, and our uncertainty about it has mostly been resolved. The Omicron pandemic is now the Covid-19 pandemic, and we’re back in a kind of ‘normal mode’ albeit at the top of the Omicron wave.

A few weeks ago, the majority of the information and week’s discussion were in the Omicron posts. This week, the vast majority are in the normal weekly post, there’s only one pure Omicron post (although the Polymarket post was closely related) and it’s relatively short.

Going forward, after this week, my default plan is to incorporate the Omicron news and the Omicron-related predictions into the mainline weekly posts, and supplement with issue-specific posts rather than general Omicron updates.

If things continue to go well, I hope to use the resulting bandwidth to start doing less speed premium writing, and more longer term analysis, both on Covid and otherwise. There’s a lot I simply haven’t had the time to think about let alone write out properly.

Remember, this is a happy moment. We are no longer in (as much of) an emergency.

Severity

Thread updating what we know about severity of Omicron. Less severe in adults, less loss of taste and smell, mostly this is confirmation of the usual good news.

The issue of severity in children is less clear because Covid-19 is so non-severe in children it’s hard to tell the relative change.

All we know is there’s a lot more cases, and a lot more cases means a lot more child cases, and a lot more ‘with Covid’ cases means proportionally more child cases, and none of that is particular reason to doubt that the severity effect in children is different from that in adults, but we don’t know for sure because we lack the data, which results in for example the UK’s report saying about severity in children we have ‘low confidence.’ I agree that we are less confident here, but our baseline should mostly be that it roughly matches the adult changes, rather than a baseline of no change from Delta. But even if there is no change from Delta, that’s still not high risk.

Spread

BA.2 is the other sub-variant of Omicron, whereas BA.1 is the one that shows up on S-gene deletion tests and it most common in the USA and UK. This thread discusses whether we should worry about BA.2, and offers this chart, you can get more at this website.

If we take this at face value, it was 10% of Omicron cases on 21 December, and is about 30% of Omicron cases on 8 January, which is about two doublings in three weeks. Even with fast generations that edge is substantial, and is one reason to expect this to blunt downward trends somewhat, but the rates of decline of BA.1 are faster than this in places where BA.1 is declining, so this is unlikely to be additionally scary in those areas. If BA.2 ‘gets there first’ in a given area, it would lead to a higher peak, and it makes containment by China potentially that much harder.

Wastewater surveillance is available in a few places other than Boston so here’s some data from California.

Whereas Boston continues to be a straight line rapidly dropping down (interesting note is that the ratio of north Boston to south Boston doesn’t seem very constant?), the California data represents a gentle peak that doesn’t seem to be continuing to drop, in a way that doesn’t have an obvious physical explanation. I’m not sure what to make of it.

A basic but well-executed thread from Bob Wachter that spells out how much Covid is out there right now, and thus how much less risky activity will soon become. The difference will be extreme, although I’m not convinced it will be quite this extreme.

That’s a factor of 50, a 98% decline in relative risk of a given activity (e.g. microCovids), within a month. I’d be surprised if we get a 98% decline within a month, but I’d b...

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: Seattle: The physics of dynamism (and AI alignment), published by Alex Flint on January 19, 2022 on The AI Alignment Forum.

I will be giving a talk on the physics of dynamism with relevance to AI alignment (abstract below). I am hoping to get feedback on a new thread of research and am also eager to meet folks in Seattle. Afterward the talk there will be some time to hang out with the group.

Location is room 5, Phinney Center (see address below) ——— What does it mean to be intellectually alive? What is it that distinguishes a creative and vibrant company/lab/community from one that has been taken of by “suits”? In this talk I will attempt a formalization of dynamism in terms of physical systems that unboundedly avoid stasis. I will then ask what kind of AI system would exhibit dynamism in this sense.

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

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: Thought Experiments Provide a Third Anchor, published by jsteinhardt on January 18, 2022 on LessWrong.

Previously, I argued that we should expect future ML systems to often exhibit "emergent" behavior, where they acquire new capabilities that were not explicitly designed or intended, simply as a result of scaling. This was a special case of a general phenomenon in the physical sciences called More Is Different.

I care about this because I think AI will have a huge impact on society, and I want to forecast what future systems will be like so that I can steer things to be better. To that end, I find More Is Different to be troubling and disorienting. I’m inclined to forecast the future by looking at existing trends and asking what will happen if they continue, but we should instead expect new qualitative behaviors to arise all the time that are not an extrapolation of previous trends.

Given this, how can we predict what future systems will look like? For this, I find it helpful to think in terms of "anchors"---reference classes that are broadly analogous to future ML systems, which we can then use to make predictions.

The most obvious reference class for future ML systems is current ML systems---I'll call this the current ML anchor. I think this is indeed a pretty good starting point, but we’ve already seen that it fails to account for emergent capabilities.

What other anchors can we use? One intuitive approach would be to look for things that humans are good at but that current ML systems are bad at. This would include:

Mastery of external tools (e.g. calculators, search engines, software, programming)

Very efficient learning (e.g. reading a textbook once to learn a new subject)

Long-term planning (e.g. being able to successfully achieve goals over months)

Models sufficiently far in the future will presumably have these sorts of capabilities. While this still leaves unknowns---for instance, we don't know how rapidly these capabilities will appear---it's still a useful complement to the current ML anchor. I'll call this the human anchor.

A problem with the human anchor is that it risks anthropomorphising ML by over-analogizing with human behavior. Anthropomorphic reasoning correctly gets a bad rap in ML, because it's very intuitively persuasive but has a mixed at best track record. This isn't a reason to abandon the human anchor, but it means we shouldn't be entirely satisfied with it.

This brings us to a third anchor, the optimization anchor, which I associate with the "Philosophy" or thought experiment approach that I've described previously. Here the idea is to think of ML systems as ideal optimizers and ask what a perfect optimizer would do in a given scenario. This is where Nick Bostrom's colorful description of a paperclip maximizer comes from, where an AI asked to make paperclips turns the entire planet into paperclip factories. To give some more prosaic examples:

The optimization anchor would correctly predict imitative deception (Lin et al., 2021), since a system optimized to produce high-probability outputs has no intrinsic reason to be truthful.

It also would observe that power-seeking is instrumentally useful for many different goals, and so predict that optimal policies (as well as sufficiently powerful neural networks) will tend to do so (Turner et al., 2021).

Ideas produced by the optimization anchor are often met with skepticism, because they often contradict the familiar current ML anchor, and they don't benefit from the intuitive appeal of the human anchor. But the differences from these other two anchors are precisely what make the optimization anchor valuable. If you (like me) feel that both the current ML and human anchors paint an incomplete picture, then you should want a third independent perspective.

The optimization anchor does have limitation...

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: You Can Get Fluvoxamine, published by AppliedDivinityStudies on January 18, 2022 on LessWrong.

[TLDR: I paid $95 for a 10 minute video consultation with a doctor, told them I was depressed and wanted fluvoxamine, and got my prescription immediately.]

I'm not a doctor, and this isn't medical advice. If you want information on the status of fluvoxamine as a Covid treatment, you can see the evidence base in the appendix, but interpreting those results isn't my business.

I'm just here to tell you that if you want fluvoxamine, you can get it.

Years ago, some of my friends were into downloading apps that would get you a 10 minute consultation with a doctor in order to quickly acquire a prescription for medical marajuana. Today, similar apps exist for a wide range of medications, and with a bit of Googling, you can find one that will prescribe you fluvoxamine.

What's required on your end? In my case, $95, 10 minutes of my time, and some white lies about my mental health. Fluvoxamine is only prescribed right now for depression and anxiety, so if you want it, my advice is to say that:

You have an ongoing history of moderate depression and anxiety

You have taken Fluvoxamine in the past, and it's helped

And that's basically it. Because there are many other treatments for depression, you do specifically have to ask for Fluvoxamine by name. If they try to give you something else, say that you've tried it before and didn't like the side effects (weight gain, insomnia, headaches, whatever).

One more note, and this is critical: unless you are actually suicidal, do not tell your doctor that you have plans to commit suicide, to hurt yourself or others, or do anything that sounds like an immediate threat. This puts you at risk of being put involuntarily in an inpatient program, and you don't want that.

Finally, you might ask: isn't this super unethical? Aren't you not supposed to lie to doctors to get drugs? Maybe, I don't know, this isn't medical advice, and it's not really ethical advice either. I think the only real potential harms here are we consume so much fluvoxamine that there isn't enough for depressed people, or that doctors start taking actual depressed patients who want fluvoxamine less seriously. As far as I can tell, there isn't currently a shortage, as to the latter concern, I couldn't really say.

Appendix

Again, this isn't medical advice. You shouldn't take any of these results or pieces of news coverage as evidence that fluvoxamine works and that the benefits outweigh the costs. I'm literally only adding this to cover my own ass and make the point that fluvoxamine is a normal mainstream thing and not some weird conspiracy drug.

Here's the Lancet article, and the JAMA article.

Here's Kelsey Piper at Vox:

One medication the TOGETHER trial found strong results for, fluvoxamine, is generally used as an antidepressant and to treat obsessive-compulsive disorder. But it appears to reduce the risk of needing hospitalization or medical observation for Covid-19 by about 30 percent, and by considerably more among those patients who stick with the 10-day course of medication. Unlike monoclonal antibodies, fluvoxamine can be taken as a pill at home --- which has been an important priority for scientists researching treatments, because it means that patients can take their medication without needing to leave the home and without straining a hospital system that is expected to be overwhelmed.

"We would not expect it to be affected by which variants" a person is sick with, Angela Reiersen, a psychiatrist at Washington University in St. Louis whose research turned up fluvoxamine as a promising anti-Covid candidate, told me.

And here's a Wall Street Journal article headlined "Is Fluvoxamine the Covid Drug We've Been Waiting For?" with subheading "A 10-day treatment costs only $4 and app...

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: Personal Perspective: Is EA particularly promising?, published by SebastianSchmidt on January 18, 2022 on The Effective Altruism Forum.

(Edited)

Summary: During a personal decision-making process, I wanted to examine my understanding of the promise of the community and decided that it was useful for me to make it into an okay-ish post. The tentative conclusion of the promise of EA from my point of view: 6/10 (where 10 is an incredibly flourishing state).

Intro

I’m currently in the phase of making some important personal/professional decisions related to the EA mission. Therefore, I wanted to examine my current understanding of its promise. However, in the process of doing so, I observed that I was too sloppy in my thinking and it became clear that approaching it from a perspective of wanting to post it was very helpful for me.

I’d be deeply grateful if you could strengthen my understanding e.g., by pointing me to important discussions, giving me a different framework to think about value, and pointing out suboptimal assumptions.

Also, while I think there's an important distinction to be drawn between the EA mission and the EA community, I'm mainly lumping the two together in this post.

Tentative conclusion:

Intuitive rating: 6/10 (where 10 is an incredibly flourishing state). EA has made outstanding progress in terms of understanding value, influencing according to those values, and being antifragile. However, when grounded in the bigger scheme of things (e.g., Elon Musk’s net worth) and considering existing problems (e.g., lack of individual flourishing and the presence of mental health issues) my previous unconditional enthusiasm decreases. I'm in the process of thinking about the implications but this post has already evolved more than I expected so coming up with constructive solutions is beyond present-me.

Three (abstract) conditions for bringing about long-term value

What has to hold true in order for something like a community to be promising? There are probably excellent models out there but for pragmatic reasons, I generated a framework.

Understanding true value

Understanding what's actually valuable (and preferably what's less valuable) is needed. The vector must be pointing in the right direction.

Influencing / creating according to values

Understanding isn't enough - a capacity for influencing / creating according to the values is also needed. The vector must have a non-trivial length.

Being antifragile (improving in the face of mistakes + being sustainable)

It's highly unlikely that any process will be working well from its inception. Therefore, being antifragile in the face of uncertainty and mistakes is crucial. My vector model is too simplistic and breaks down now but I think this is an important notion as Nassim Taleb describes this quality: “Wind extinguishes a candle but energizes fire”.

Assessing my impressions of the community

Understanding true value (excellent contributions to the world or qualities of the community)

Intuitive rating: 8/10 based on below.

Normative

Moral uncertainty. Moral parliament.

Impartial altruism/impartial welfarism.

Expanding circle of compassion / anti-tribalism/ anti-speciesism.

Long-termism.

In-between normative and epistemic

Global priorities research.

Existential risks.

Suffering risks.

Transcending important unfortunate tendencies of the mind

Scope insensitivity.

Confirmation bias.

Tribalism.

Crucial considerations.

Epistemic

A massive window of discourse.

Cruxes and scout mindset.

Probabilities.

Predictions and betting.

Finding mistakes.

Lacks (or growth areas)

Understanding human flourishing / psychological well-being

It seems to me that most people aren’t flourishing by having a strong sense of agency, hope, and purpose while pursuing what they believe is the most important thing they could do with their life. ...

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 longtermist AI governance landscape: a basic overview, published by SamClarke on January 18, 2022 on The Effective Altruism Forum.

Aim: to give a basic overview of what is going on in longtermist AI governance.

Audience: people who have limited familiarity with longtermist AI governance and want to understand it better. I don’t expect this to be helpful for those who already have familiarity with the field.

This post outlines the different kinds of work happening in longtermist AI governance. For each kind of work, I’ll explain it, give examples, sketch some stories for how it could have a positive impact, and list the actors I’m aware of who are currently working on it.[1]

Firstly, some definitions:

AI governance means bringing about local and global norms, policies, laws, processes, politics, and institutions (not just governments) that will affect social outcomes from the development and deployment of AI systems.[2]

Longtermist AI governance, in particular, is the subset of this work that is motivated by a concern for the very long-term impacts of AI. This overlaps significantly with work aiming to govern transformative AI (TAI).

It’s worth noting that the field of longtermist AI governance is very small. I’d guess that there are around 60 people working in AI governance who are motivated by a concern for very long-term impacts.

Short summary

On a high level, I find it helpful to consider there being a spectrum between foundational and applied work. On the foundational end, there’s strategy research, which aims to identify good high-level goals for longtermist AI governance; then there’s tactics research which aims to identify plans that will help achieve those high-level goals. Moving towards the applied end, there’s policy development work that takes this research and translates it into concrete policies; work that advocates for those policies to be implemented, and finally the actual implementation of those policies (by e.g. civil servants).

There’s also field-building work (which doesn’t clearly fit on the spectrum). Rather than contributing directly to the problem, this work aims to build a field of people who are doing valuable work on it.

Of course, this classification is a simplification and not all work will fit neatly into a single category.

You might think that insights mostly flow from the more foundational to the more applied end of the spectrum, but it’s also important that research is sensitive to policy concerns, e.g. considering how likely your research is to inform a policy proposal that is politically feasible.

We’ll now go through each of these kinds of work in more detail.

Research

Strategy research

Longtermist AI strategy research ultimately aims to identify high-level goals we could pursue that, if achieved, would clearly increase the odds of eventual good outcomes from advanced AI, from a longtermist perspective (following Muehlhauser, I’ll sometimes refer to this aim as ‘getting strategic clarity’).

This research can itself vary on a spectrum between targeted and exploratory as follows:

Targeted strategy research answers questions which shed light on some other specific, important, known question

e.g. “I want to find out how much compute the human brain uses, because this will help me answer the question of when TAI will be developed (which affects what high-level goals we should pursue)”

Exploratory strategy research answers questions without a very precise sense of what other important questions they’ll help us answer

e.g. “I want to find out what China’s industrial policy is like, because this will probably help me answer a bunch of important strategic questions, although I don't know precisely which ones”

Examples

Work on TAI forecasting, e.g. biological anchors and scaling laws for neural language models.

Example of strategic rele...

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: Long-Term Future Fund: July 2021 grant recommendations, published by abergal on January 18, 2022 on The Effective Altruism Forum.

Introduction

The Long-Term Future Fund made the following grants through July 2021:

Total grants: $1,427,019

Number of grantees: 18

Payout date: July 2021

Report authors: Asya Bergal (Chair), Oliver Habryka, Adam Gleave, Evan Hubinger, Luisa Rodriguez

This payout report is substantially delayed, but we’re hoping to get our next payout report covering grants through November out very soon. Updates since our last payout report:

We took on Luisa Rodriguez as a guest manager in June and July.

We switched from a round-based system to rolling applications.

We received $675,000 from Jaan Tallinn through the Survival and Flourishing Fund.

Public payout reports for EA Funds grantees are now optional.

Consider applying for funding from the Long-Term Future Fund here.

Grant reports

Note: Many of the grant reports below are very detailed. Public reports are optional for our grantees, and we run all of our payout reports by grantees before publishing them. We think carefully about what information to include to maximize transparency while respecting grantees’ preferences. We encourage anyone who thinks they could use funding to positively influence the long-term trajectory of humanity to apply for a grant.

Grant reports by Asya Bergal

Any views expressed below are my personal views and not the views of my employer, Open Philanthropy. In particular, receiving funding from the Long-Term Future Fund should not be read as an indication that an organization or individual has an elevated likelihood of receiving funding from Open Philanthropy. Correspondingly, not receiving funding from the Long-Term Future Fund (or any risks and reservations noted in the public payout report) should not be read as an indication that an organization or individual has a diminished likelihood of receiving funding from Open Philanthropy.

Ezra Karger, Pavel Atanasov, Philip Tetlock ($572,000)

Existential risk forecasting tournaments.

This grant is to Ezra Karger, Pavel Atanasov, and Philip Tetlock to run an existential risk forecasting tournament.

Philip Tetlock is a professor at the University of Pennsylvania; he is known in part for his work on The Good Judgment Project, a multi-year study of the feasibility of improving the accuracy of probability judgments, and for his book Superforecasting: The Art and Science of Prediction, which details findings from that study. Pavel Atanasov is a decision psychologist currently working as a Co-PI on two NSF projects focused on predicting the outcomes of clinical trials. He previously worked as a post-doctoral scholar with Philip Tetlock and Barbara Mellers at the Good Judgement Project, and as a consultant for the SAGE research team that won the last season of IARPA’s Hybrid Forecasting Competition.

Ezra Karger is an applied microeconomist working in the research group of the Federal Reserve Bank of Chicago; he is a superforecaster who has participated in several IARPA-sponsored forecasting tournaments and has worked with Philip Tetlock on some of the methods proposed in the tournament below.

Paraphrasing from the proposal, the original plan for the tournament was as follows:

Have a panel of subject-matter experts (SMEs) choose 10 long-run questions about existential risks and 20 short-run, resolvable, early warning indicator questions as inputs for the long-run questions.

Ask the SMEs to submit forecasts and rationales for each question.

Divide the superforecasters into two groups. Ask each person to forecast the same questions as the SMEs and explain those forecasts. For short-run questions, evaluate forecasts using a proper scoring rule, like Brier or logarithmic scores. For long-run questions, use reciprocal scoring to incentivize ac...

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: Animal Welfare Fund: November 2021 grant recommendations, published by kierangreig on January 18, 2022 on The Effective Altruism Forum.

Introduction

The Animal Welfare Fund made the following grants as part of its 2021 Q4 grant cycle:

Total grants: $1,115,000

Number of grantees: 21

Acceptance rate (excluding desk rejections): 29%

Payout date: November-December, 2021

Report authors: Kieran Greig (Chair), Alexandria Beck, Karolina Sarek, Marcus Davis, Mikaela Saccoccio

The three primary areas we granted to in this round were:

Large-scale and neglected animal populations (for instance, farmed fish).

Large-scale and neglected geographies (for instance, work in medium or large Asian countries).

Exploratory work regarding policy and/or alternative proteins (for instance, attempting to secure government funding for research and development of alternative proteins).

Would you like to get funded? Apply for funding.

Grant recipients

See below for a list of grantees' names, grant amounts, and brief grant descriptions:

Fish Welfare Initiative ($170,000): Funding ~40% of general operating budget for 2022, mostly to expand producer/corporate/policy work on fish in India.

Animals Aotearoa ($103,000): Supporting the only NGO working on chicken corporate welfare commitments in New Zealand.

Vegetarian Society of Denmark ($90,000): Securing government funding for R&D on plant-based alternatives.

Aquatic Life Institute ($80,000): Salaries for existing positions and potentially towards hiring a research lead for work on farmed fish.

Advocates for Animals ($75,000): Increasing capacity of an organization providing legal support for EAA organizations in the UK, and conducting judicial review related to salmon farming and fast-growing broilers.

Animal Rights Organization EVA ($67,000): Campaigning for implementation of Animal Protection Law in Bosnia & Herzegovina.

Animal Advocacy Africa ($66,000): Supporting African animal advocacy organisations with fundraising and piloting similar work in Asia.

Animal Empathy Philippines ($64,000): A year of funding to do community building on effective animal advocacy in the Philippines.

Shannon Campion ($63,000): Seed funding to advise the movement on political advocacy and develop a plan to build greater political power.

Sentient Media ($60,000): A year of funding for providing digital marketing/training/content production for high-impact farmed animal advocacy groups.

Fórum Nacional de Proteção e Defesa Animal ($50,000): Continue accountability with companies that have already committed to banning chicken cages in Brazil.

Veganuary ($50,000): Upgrading Latin American staff from freelance to full-time to increase corporate engagement success.

Animal Advocacy Careers ($36,000): Three months of general operating costs while they separate from their fiscal sponsor and await a large grant.

Tanzania Animal Welfare Society ($36,000): One year costs to conduct a survey on Farmed Fish Welfare in Tanzania which aims to influence the Tanzanian fisheries policy.

Arba ($25,000): Supporting research on meat consumption and two cage-free campaigning positions in Peru.

Faunalytics ($21,000): Supporting a meta-analysis looking for substitution effects during meat reduction efforts, especially for increases in chicken and fish consumption.

George Stiffman ($20,000): Providing 3 months salary to pilot restaurant partnerships in order to help evaluate “cuisine expansion” as a possible new intervention.

Sociedade Vegetariana Brasileira ($20,000): Supporting expanding meat reduction efforts in public services of two Brazilian cities.

Che Green ($7,500): Research and writing on methods to reduce suffering of urban rodent populations.

Nicolas Delon ($6,800): Research support and summer salary to review neglected and tractable issues concerning urban wild animal welfare.

Ano...

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: Polymarket Covid-19 1/17/2022, published by Zvi on January 17, 2022 on LessWrong.

Polymarket’s prediction markets are back and they are more plentiful than ever, both on Covid-19 and otherwise. Americans are not eligible to trade, but plenty of my readers reside elsewhere. And even if you can’t trade, the information component is still free.

Liquidity in any given market is not yet as high as I’d like, but it is still high enough to very much count as real money, provide insights on real problems and allow the astute reader who is eligible to participate to perhaps pick up some value.

To celebrate having all these new markets, and to take advantage of the information and the intuition pumps, I’ll go over what’s available, and say how far I’d be willing to trade a given contract or set of contracts on the assumption that I had access to the markets plus an adequate bankroll.

I’m going to limit this post to Covid-19 markets, because there’s enough value out there that I don’t want to postpone posting while I go through the other stuff.

One problem with the old Buy/Sell/Hold is that it is a conservative system. If something is trading at 50% and I say I’d buy it to 60%, I’m doing that because even after adjusting for the market’s opinion I still think the probability is substantially above 60%. Sometimes I’m uncertain for some reason and I’m avoiding making big mistakes, but my gut says it is much higher. Other times I’m not uncertain, so I’m looking for a smaller edge. This makes it tricky when the time comes to assign a Briar score or otherwise evaluate predictions mathematically.

My solution to that is that I’ll be including my fair price explicitly, in addition to what I would do, while keeping in mind that they are not always well-considered. Then one can either evaluate market performance (and stop where I say to stop) or go full Green Knight and see if my full estimates are superior.

And of course, note all of these carefully:

Markets tend to improve in accuracy over time.

That means lack of market movement over time is (some) evidence of accuracy.

Others are also reading this, so if you don’t get to it fast, that’s adverse selection for you.

If liquidity has gone up, that’s also evidence of accuracy, more adverse selection.

Many of these markets are time-sensitive, and their fair value changes over time rapidly even if nothing happens.

None of this is investment advice!

Without further ado, let’s see what we’ve got. Keep in mind, this was intentionally written and posted quickly, so mistakes do happen.

USA Case Counts

Market / fair:

Will the 7-day USA case count on February 1, 2022 be above.

200k: 96% / 98%

400k: 86% / 86%

600k: 28% / 30%

800k: 6% / 4%

1mm: 3% / 1.5%

The thing that jumps out at me right away is that this is placing the majority of the probability in a narrow range between 400k and 600k. With a full two weeks to go, this seems superficially seems overconfident. Yet there’s good arguments that it’s not, and on reflection I mostly agree.

How low could this plausibly go?

The Northeast already peaked last week, and likely so did some areas in the South. Let’s say the overall American peak could easily have been mid-month. One should expect a slower fall than rise, but not obviously by that much depending on distribution. The issue is that if different regions peak at different times, then that will decrease the slope of the downward trend until everyone has peaked.

On January 1, the case average was 416k, just above the 400k threshold. That was the week of Christmas and New Year’s, and we see a big jump on January 3 as the information back-fills. The jumps per day are pretty dramatic, so moving by even two days makes a big difference here.

Thus, we can consider going under 400k to require a symmetrical fall that matches the peak that requires a relat...