Astral Codex Ten Podcast

https://astralcodexten.substack.com/p/movie-review-dont-look-up I. Don't Look Up is primarily a movie about existential risk, and many great people have already reviewed it as such. I'm going to be less virtuous and use it as a springboard to talk about politics. But first, the plot in a nutshell: Male Scientist and Female Scientist discover a comet will hit Earth in six months. They contact the relevant authorities, Black Scientist and Asian Scientist, and go to meet the President (who, despite being a woman, is Donald Trump). The President says scientists are always doomsaying, if people get too panicked she'll lose the midterm election, and she'll get around to dealing with this later. (the Earth, at this point, has five months and however many days left)

https://astralcodexten.substack.com/p/lewis-carroll-invented-retroactive Retroactive public goods funding is one of those ideas that's so great people can't stop reinventing it. I know of at least five independent inventions under five different names: "social impact bonds" by a New Zealand economist in 1988, "certificates of impact" by Paul Christiano in 2014, "retroactive public goods funding" by Vitalik Buterin a few years ago, "EA loans" by a blogger who prefers to remain anonymous, and "venture grants" by Mako Yass. These aren't all exactly the same idea. Some are slightly better framed than others and probably I'm being terribly disrespectful to the better ones by saying they're the same as the worse ones. But I think they all share a basic core: some structure that lets profit-seeking venture capitalist types invest in altruistic causes, in the hopes that altruists will pay them back later once they've been shown to work. Upon re-reading some old SSC comments, I found a gem I'd missed the first time around: Julie K says that the actual first person to invent this idea was Lewis Carroll (aka author of Alice in Wonderland) back in 1894. She quotes from his book Sylvie and Bruno:

https://astralcodexten.substack.com/p/links-for-december [Remember, I haven't independently verified each link. On average, commenters will end up spotting evidence that around two or three of the links in each links post are wrong or misleading. I correct these as I see them, and will highlight important corrections later, but I can't guarantee I will have caught them all by the time you read this.] 1: List Of Games That Buddha Would Not Play. 2: Claim via NPR: When Brazil had high inflation in the 1990s, some economists developed a plan: price everything in inflation-adjusted units, so that people felt like things were "stable", then declare that the Inflation Adjusted Unit was the new currency. How Fake Money Saved Brazil. Also interesting: they tried it because the new finance minister knew no economics, recognized his ignorance, and was willing to call up random economists and listen to their hare-brained plans. 3: In the 19th century, a group of Tibeto-Burman-speaking former headhunters along the India/Burma border declared themselves the descendants of Manasseh (one of the Ten Lost Tribes) and converted en masse to Judaism. In 2005, the Chief Rabbinate of Israel accepted their claim and expedited immigration paperwork for several thousand of them.

https://astralcodexten.substack.com/p/acx-grants-results Thanks to everyone who participated in ACX Grants, whether as an applicant, an evaluator, or a funder. Before I announce awardees, a caveat: this was hard in lots of ways I didn't expect. I got 656 applications addressing different problems and requiring different skills to judge. I'll write a long post on it later, but the part I want to emphasize now is: if I didn't grant you money, it doesn't mean I didn't like your project. Sometimes it meant I couldn't find someone qualified to evaluate it. Other times a reviewer was concerned that if you were successful, your work might be used by terrorists / dictators / AI capabilities researchers / Republicans and cause damage in ways you couldn't foresee. Other times it meant it was a better match for some other grant organization and I handed it off to them. Still other times, my grant reviewers tied themselves up in knots with 4D chess logic like "if they're smart enough to attempt this project, they're smart enough to know about XYZ Grants which is better suited for them, which means they're mostly banking on XYZ funding and using you as a backup, but if XYZ doesn't fund these people then that's strong evidence that they shouldn't be funded, so even though everything about them looks amazing, please reject them." I have no idea if things really work this way, but I needed some experienced grant reviewers on board and they were all like this. I took these considerations seriously and in some marginal cases they prevented funding.

https://astralcodexten.substack.com/p/mantic-monday-dogs-in-wizard-hats I found this YouTube explainer about prediction markets on the subreddit. It's pretty good! My small nitpicks are that it overestimates their accuracy relative to traditional forecasters (it focuses on markets beating forecasters in 2008, but I don't think this is consistent) and underestimates their resilience against bad actors trying to skew the probabilities. Still, this will be my go-to source when someone wants a short explanation of what these are and why I'm so excited. Futuur Soon Last week I mentioned a new prediction market called Futuur . Today we'll look at it in more depth. Futuur sends non-Americans to their real money markets and Americans to their play money markets (because of the US' unique anti-prediction-market regulations). Their play money markets are awful:

https://astralcodexten.substack.com/p/highlights-from-the-comments-on-diseasonality The main highlight was an email I got from a reader who prefers to remain anonymous, linking me to Projecting The Transmission Dynamics Of SARS-CoV2. This paper is head and shoulders above anything I found during my own literature review and just comes out and says everything painfully tried to piece together. Either my research skills suck, the epidemiology literature is a bunch of disparate subthreads with wildly differing levels of competence, or both. The authors (including Marc Lipsitch who some of you might know from Twitter) are writing in May 2020, trying to predict the future course of COVID. To that end, they investigate the past course of two other coronaviruses called OC43 and HKU1, which cause mild colds. These show a seasonal pattern. Why?

https://astralcodexten.substack.com/p/addendum-to-no-evidence-post The day after I wrote The Phrase "No Evidence" Is A Red Flag For Bad Science Communication, FT published this article: Like many uses of "no evidence", they meant that one particular study of this complicated question had failed to reject the null hypothesis. Here's what happened to Metaculus' prediction tournament when the same study came out: The consensus prediction dropped from 72% chance that it was less lethal, to 63% chance. But it quickly recovered, and is now up to 80%. This is an unusually clear example of the difference between classical and Bayesian ways of thinking.

https://astralcodexten.substack.com/p/addendum-to-luvox-post In my post yesterday, I quoted a Vox article describing work by Dr. Ed Mills and others to get the FDA to approve Luvox for COVID. As of that point, the FDA didn't know how to process an application without a sponsoring drug company: [Professor Ed] Mills, who thinks that fluvoxamine and budesonide are both appropriate to prescribe to patients sick with Covid-19, compares public messaging on fluvoxamine to communications about Merck's drug molnupiravir. The evidence for molnupiravir is in many ways weaker than the evidence for fluvoxamine, but molnupiravir was produced by a major pharmaceutical company that can shepherd it through the process of becoming a recommended drug. On a call last week, Mills said, the FDA told him "they don't know how to deal with submissions where there isn't someone to be responsible for it." But it looks like just as I published, he and his colleagues found a way around the problem:

https://astralcodexten.substack.com/p/the-fda-has-punted-decisions-about I. Here's my pitch for fluvoxamine (Luvox) for COVID. In the midst of all the hype about ivermectin and hydroxychloroquine, scientists put together the giant 4,000-person TOGETHER trial, intended to test all these exciting COVID early treatments. You know what happened next: ivermectin and hydroxychloroquine crashed and burned. But a different drug, the SSRI antidepressant fluvoxamine, actually did really well! It decreased COVID hospitalizations by about 30% - not the perfect cure rate the rumors attributed to ivermectin, but a substantial decrease. Given the size and professionalism of this study, and another smaller one that also got positive results, I and many others take Luvox pretty seriously. At this point I'd give it 60-40 it works. Can you prescribe a medication when you're only 60% confident in it? There's some thorny philosophical issues around this, but I think in the end you have to compare risks and benefits. What are the risks? Like every medication, including Tylenol, aspirin, etc, Luvox has some common minor side effects and some rare major ones. But let's step back a second. Fluvoxamine is a bog-standard SSRI. Its side effects are generic SSRI side effects. We give SSRIs to 30 million people a year, or about 10% of all Americans. As a psychiatrist, I'm not supposed to say flippant things like "we give SSRIs out like candy". We do careful risk-benefit analysis and when appropriate we screen patients for various risk factors. But after we do all that stuff, we give them to 10% of Americans, compared to 12% of Americans who got candy last Halloween. So you can draw your own conclusion about how severe we think the risks are. For some reason the same experts who don't mind prescribing SSRIs when people have mild depression freak out about prescribing them when they're the only evidence-based oral medication for a deadly global pandemic. "What about SSRI withdrawal?", they ask. After a ten day course? On 100 mg imipramine-equivalent dose? Minimal. "What about long QT syndrome?" The VA system took 35,000 high-risk older patients off of an unusually-likely-to-cause-QT-syndrome SSRI in 2011, and were unable to find any evidence that this prevented even a single case of the syndrome, let alone any negative outcome!

https://astralcodexten.substack.com/p/mantic-monday-let-me-google-that Let Me Google That For You New from Google this month: Creating A Prediction Market On Google Cloud. Google announces that they've been running an internal prediction market for the past year, with "over 175,000 predictions from over 10,000 Google employees". 1 Predictive analytics.jpg Most of it's classified because they're predicting stuff about Google's corporate secrets, but some friendly Googlers were at least willing to walk me through the article and clarify pieces I didn't understand. The market, called Gleangen, is actually the second prediction market Google's tried. The first, in 2007, was called Prophit - the team included occasional ACX commenter Patri Friedman, who's since moved into the charter city space. (source) Prophit wound down because the founders left and nobody really knew what to do with; you can read about some of their findings here. In 2020, with all the uncertainty around coronavirus, some Googlers decided to try again. Gleangen is the result. Unlike most prediction markets, anybody can create a question on Gleangen. This usually goes badly: most people are terrible at writing questions with objective resolutions. Google manages by having a dedicated team of moderators who go over everything and amend it when needed. The market pays out in play money and the right to be on a leaderboard. So far it's not doing much else. The Googlers I talked to saw no evidence that company executives were paying much attention to it when making decisions. Why not? Hal Varian, Google's chief economist, said in a Conversation with Tyler Cowen: COWEN: Why doesn't business use more prediction markets? They would seem to make sense, right? Bet on ideas. Aggregate information. We've all read Hayek. VARIAN: Right. And we had a prediction market [referring to Prophit in 2007]. I'll tell you the problem with it. The problem is, the things that we really wanted to get a probability assessment on were things that were so sensitive that we thought we would violate the SEC rules on insider knowledge because, if a small group of people knows about some acquisition or something like that, there is a secret among this small group.