Astral Codex Ten Podcast

Connor Leahy discusses the idea of an 'AGI Fire Alarm' and argues GPT-3 might be the last such warning we'll receive before it's too late to act.

Suppose a lot of that stuff about bravery debates is right. That lots of the advice people give is useful for some people, but that the opposite advice is useful for other people. For example, "You need to stop being so hard on yourself, remember you are your own worst critic" versus "Stop making excuses for yourself, you will never be able to change until you admit you've hit bottom." Or "You need to remember that the government can't solve all problems and that some regulations are counterproductive" versus "You need to remember that the free market can't solve all problems and that some regulations are necessary." Or "You need to pay more attention to your diet or you'll end up very unhealthy" versus "You need to pay less attention to your weight or you'll end up in a spiral of shame and self-loathing and at risk of eating disorders." Or "Follow your dreams, you don't want to be working forever at a job you hate", versus "Your dream of becoming a professional cosplayer may not be the best way to ensure a secure future for your family, go into petroleum engineering instead."

It's been two and a half months since I deleted the blog, so I owe all of you an update on recent events. I haven't heard anything from the New York Times one way or the other. Since nothing has been published, I'd assume they dropped the article, except that they approached an acquaintance for another interview last month. Overall I'm confused. But they definitely haven't given me any explicit reassurance that they won't reveal my private information. And now that I've publicly admitted privacy is important to me – something I tried to avoid coming on too strong about before, for exactly this reason – some people have taken it upon themselves to post my real name all over Twitter in order to harass me. I probably inadvertently Streisand-Effect-ed myself with all this; I still think it was the right thing to do. At this point I think maintaining anonymity is a losing battle. So I am gradually reworking my life to be compatible with the sort of publicity that circumstances seem to be forcing on me. I had a talk with my employer and we came to a mutual agreement that I would gradually transition away from working there. At some point, I may start my own private practice, where I'm my own boss and where I can focus on medication management – and not the kinds of psychotherapy that I'm most worried are ethically incompatible with being a public figure. I'm trying to do all of this maximally slowly and carefully and in a way that won't cause undue burden to any of my patients, and it's taking a long time to figure out.

Trigger warning: deliberately provoking horror about graduates' real-​world post-​college prospects. Epistemic status: intended as persuasive speech, may somewhat overstate case. Ladies and gentlemen, I am honored to have been invited to speak here at the great University of [mumble]. Go Wildcats, Spartans, or Eagles, as the case may be! I apologize if what I have to say to you sounds a little unpolished. I was called in on very short notice after your original choice for graduation speaker, Mr. Steven L. Carter, had his invitation to speak rescinded due to his offensive and quite honestly outrageous opinions. Let me say in no uncertain terms that I totally condemn him and everything he stands for, and that I am glad to see the University of [mumble] taking a strong stand against this sort of thing.

[Related: Tyler Cowen on rationalists, Noah Smith on rationalists, Will Wilkinson on rationalists, etc] If I were an actor in an improv show, and my prompt was "annoying person who's never read any economics, criticizing economists", I think I could nail it. I'd say something like: Economists think that they can figure out everything by sitting in their armchairs and coming up with 'models' based on ideas like 'the only motivation is greed' or 'everyone behaves perfectly rationally'. But they didn't predict the housing bubble, they didn't predict the subprime mortgage crisis, and they didn't predict Lehman Brothers. All they ever do is talk about how capitalism is perfect and government regulation never works, then act shocked when the real world doesn't conform to their theories. This criticism's very clichedness should make it suspect. It would be very strange if there were a standard set of criticisms of economists, which practically everyone knew about and agreed with, and the only people who hadn't gotten the message yet were economists themselves. If any moron on a street corner could correctly point out the errors being made by bigshot PhDs, why would the PhDs never consider changing?

So, I kind of deleted the blog. Sorry. Here's my explanation. Last week I talked to a New York Times technology reporter who was planning to write a story on Slate Star Codex. He told me it would be a mostly positive piece about how we were an interesting gathering place for people in tech, and how we were ahead of the curve on some aspects of the coronavirus situation. It probably would have been a very nice article. Unfortunately, he told me he had discovered my real name and would reveal it in the article, ie doxx me. "Scott Alexander" is my real first and middle name, but I've tried to keep my last name secret. I haven't always done great at this, but I've done better than "have it get printed in the New York Times".

https://slatestarcodex.com/2020/06/17/slightly-skew-systems-of-government/ [Related To: Legal Systems Very Different From Ours Because I Just Made Them Up, List Of Fictional Drugs Banned By The FDA] I. Clamzoria is an acausal democracy. The problem with democracy is that elections happen before the winning candidate takes office. If somebody's never been President, how are you supposed to judge how good a President they'd be? Clamzoria realized this was dumb, and moved elections to the last day of an official's term. When the outgoing President left office, the country would hold an election. It was run by approval voting: you could either approve or disapprove of the candidate who had just held power. The results were tabulated, announced, and then nobody ever thought about them again. Clamzoria chose its officials through a prediction market. The Central Bank released bonds for each candidate, which paid out X dollars at term's end, where X was the percent of voters who voted Approve. Traders could provisionally buy and sell these bonds. On the first day of the term, whichever candidate's bonds were trading at the highest value was inaugurated as the new President; everyone else's bonds were retroactively cancelled and their traders refunded. The President would spend a term in office, the election would be held, and the bondholders would be reimbursed the appropriate amount.

https://slatestarcodex.com/2020/06/16/open-thread-156-25/ Normally this would be a hidden thread, but I wanted to signal boost this request for help by Professor Steve Hsu, vice president of research at Michigan State University. Hsu is a friend of the blog and was a guest speaker at one of our recent online meetups – some of you might also have gotten a chance to meet him at a Berkeley meetup last year. He and his blog Information Processing have also been instrumental in helping me and thousands of other people better understand genetics and neuroscience. If you've met him, you know he is incredibly kind, patient, and willing to go to great lengths to help improve people's scientific understanding. Along with all the support he's given me personally, he's had an amazing career. He started as a theoretical physicist publishing work on black holes and quantum information. Then he transitioned into genetics, spent a while as scientific advisor to the Beijing Genomics Institute, and helped discover genetic prediction algorithms for gallstones, melanoma, heart attacks, and other conditions. Along with his academic work, he also sounded the alarm about the coronavirus early and has been helping shape the response. This week, some students at Michigan State are trying to cancel him. They point an interview he did on an alt-right podcast (he says he didn't know it was alt-right), to his allowing MSU to conduct research on police shootings (which concluded, like most such research, that they are generally not racially motivated), and to his occasional discussion of the genetics of race (basically just repeating the same "variance between vs. within clusters" distinction everyone else does, see eg here). You can read the case being made against him here, although keep in mind a lot of it is distorted and taken out of context, and you can read his response here.

https://slatestarcodex.com/2020/06/15/the-vision-of-vilazodone-and-vortioxetine/ I. One of psychiatry's many embarrassments is how many of our drugs get discovered by accident. They come from random plants or shiny rocks or stuff Alexander Shulgin invented to get high. But every so often, somebody tries to do things the proper way. Go over decades of research into what makes psychiatric drugs work and how they could work better. Figure out the hypothetical properties of the ideal psych drug. Figure out a molecule that matches those properties. Synthesize it and see what happens. This was the vision of vortioxetine and vilazodone, two antidepressants from the early 2010s. They were approved by the FDA, sent to market, and prescribed to millions of people. Now it's been enough time to look back and give them a fair evaluation. And… …and it's been a good reminder of why we don't usually do this. Enough data has come in to be pretty sure that vortioxetine and vilazodone, while effective antidepressants, are no better than the earlier medications they sought to replace. I want to try going over the science that led pharmaceutical companies to think these two drugs might be revolutionary, and then speculate on why they weren't. I'm limited in this by my total failure to understand several important pieces of the pathways involved, so I'll explain the parts I get, and list the parts I don't in the hopes that someone clears them up in the comments.

https://slatestarcodex.com/2020/06/11/wordy-wernickes/ There are two major brain areas involved in language. To oversimplify, Wernicke's area in the superior temporal gyrus handles meaning; Broca's area in the inferior frontal gyrus handles structure and flow. If a stroke or other brain injury damages Broca's area but leaves Wernicke's area intact, you get language which is meaningful, but not very structured or fluid. You sound like a caveman: "Want food!" If it damages Wernicke's area but leaves Broca's area intact, you get speech which has normal structure and flow, but is meaningless. I'd read about this pattern in books, but I still wasn't prepared the first time I saw a video of a Wernicke's aphasia patient (source):