The Nonlinear Library

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: Dengue rates drop 77-95% after release of bacteria-infected mosquitoes in Colombia, published by SiebeRozendal on November 7, 2023 on The Effective Altruism Forum. When infected with Wolbachia, the mosquitoes are much less likely to transmit diseases such as dengue and Zika, because the bacteria compete with these viruses. The insects also pass the bacteria on to their offspring. Researchers hope that the modified mosquitoes will interbreed with the wild population wherever they are released, and that the number of mosquitoes with Wolbachia will eventually surpass that of mosquitoes without it. [...] When the scientists compared the incidence of dengue in fully treated areas with that in the same regions in the ten years before the intervention, they found that it had dropped by 95% in Bello and Medellín and by 97% in Itagüí. Since the project started, there hasn't been a large outbreak of dengue in the region. "They've had six years now with a sustained suppression of dengue," says Anders. "We're starting to see the real-world effect of Wolbachia." [...] The [World Mosquito Program] has conducted one [RCT] in Yogyakarta, Indonesia, in which mosquitoes were released in some areas of a city and the incidence of dengue was compared with that in areas that did not receive the insects. The results suggested that the technology could reduce the incidence of dengue by 77%. The organization is now conducting a similar one in Belo Horizonte, Brazil. The RCT: https://pubmed.ncbi.nlm.nih.gov/34107180/ Despite the positive results, Wolbachia mosquitoes have not yet been officially endorsed by the World Health Organization (WHO). The technology awaits an evaluation by the WHO's Vector Control Advisory Group. World Mosquito Program: https://www.worldmosquitoprogram.org/en/work/wolbachia-method/how-it-works 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: Box inversion revisited, published by Jan Kulveit on November 7, 2023 on The AI Alignment Forum. Box inversion hypothesis is a proposed correspondence between problems with AI systems studied in approaches like agent foundations , and problems with AI ecosystems, studied in various views on AI safety expecting multipolar, complex worlds, like CAIS. This is an updated and improved introduction to the idea. Cartoon explanation In the classic -"superintelligence in a box" - picture, we worry about an increasingly powerful AGI, which we imagine as contained in a box. Metaphorically, we worry that the box will, at some point, just blow up in our faces. Classic arguments about AGI then proceed by showing it is really hard to build AGI-proof boxes, and that really strong optimization power is dangerous by default. While the basic view was largely conceived by Eliezer Yudkowsky and Nick Bostrom, it is still the view most technical AI safety is built on, including current agendas like mechanistic interpretability and evals. In the less famous, though also classic, picture, we worry about an increasingly powerful ecosystem of AI services, automated corporations, etc. Metaphorically, we worry about the ever-increasing optimization pressure "out there", gradually marginalizing people, and ultimately crushing us. Classical treatments of this picture are less famous, but include Eric Drexler's CAIS ( Comprehensive AI Services ) and Scott Alexander's Ascended Economy . We can imagine scenarios like the human-incomprehensible economy expanding in the universe, and humans and our values being protected by some sort of "box". Agendas based on this view include the work of the AI Objectives Institute and part of ACS work. The apparent disagreement between these views was sometimes seen as a crux for various AI safety initiatives. "Box inversion hypothesis" claims: The two pictures to a large degree depict the same or a very similar situation, Are related by a transformation which "turns the box inside out", similarly to a geometrical transformation of a plane known as a circle inversion, and: this metaphor is surprisingly deep and can point to hard parts of some problems. Geometrical metaphor " Circular inversion " transformation does not imply the original and the inverted objects are the same, or are located at the same places. What it does imply is that some relations between objects are preserved: for example, if some objects intersect, in the circle-inverted view, they will still intersect. Similarly for "box inversion" : the hypothesis does not claim that the AI safety problems in both views are identical, but it does claim that, for most problems, there is a corresponding problem described by the other perspective. Also, while the box-inverted problems may at a surface level look very different, and be located in different places, there will be some deep similarity between the two corresponding problems. In other words, the box inversion hypothesis suggests that there is a kind of 'mirror image' or 'duality' between two sets of AI safety problems. One set comes from the "Agent Foundations" type of perspective, and the other set comes from the "Ecosystems of AIs" type of perspective. Box-inverted problems Problems with ontologies and regulatory frameworks [1] In the classic agent foundations-esque picture, a nontrivial fraction of AI safety challenges are related to issues of similarity, identification, and development of ontologies. Roughly speaking If the AI is using utterly non-human concepts and world models, it becomes much more difficult to steer and control If "what humans want" is expressed in human concepts, and the concepts don't extend to novel situations or contexts, then it is unclear how the AI should extend or interpret the human "wants" Even if an AI initially u...

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: Atlantic bluefin tuna are being domesticated: what are the welfare implications?, published by Amber Dawn on November 7, 2023 on The Effective Altruism Forum. Atlantic bluefin tuna (ABFT) are large, carnivorous ocean fish. They used to be caught relatively rarely, mainly by sports fishermen in North America. However, around the 1950s, Japanese consumers of sushi developed more of a taste for the fish, and a large aquaculture industry developed. Historically, ABFT have been either caught directly from the ocean, or captured while young and fattened in 'ranches'. However, both wild fishing and ranching pose sustainability issues, since they involve taking fish from the wild. Since 2001, there have been a number of EU-funded projects to domesticate bluefin tuna, i.e. to breed them in captivity. This is already done with other types of fish, for example salmon and tilapia, which are raised on fish farms. But it's more difficult with ABFT: they generally don't spawn in captivity, as they require certain specific conditions to spawn. However, scientists have developed methods to make ABFT spawn in captivity, through manipulating light and releasing hormones into the water to stimulate egg production in the fish. This means that it's now possible to farm these fish through 'closed-cycle aquaculture': that is, we can breed them in captivity so that they don't need to be fished from the wild. This has been seen as a win for sustainability. But what about welfare? In this report, I first offer some background on ABFT. I then examine some potential welfare issues in ABFT aquaculture. Main takeaways: Many larvae (young fish) in hatchery projects die. However, this is also true in the wild, and hatcheries may become better at preventing some of these deaths in future, in order to be commercially viable. Many of the conditions in hatcheries might pose welfare issues for ABFT, but more research is needed. The main method of slaughtering large ABFT seems relatively humane; however, the main method of slaughtering smaller ABFT seems more distressing. It's unclear how many ABFT are slaughtered using this crueller method. What are Atlantic bluefin tuna? Atlantic bluefin tuna ( thunnus thynnus ) are native to the Atlantic Ocean and Mediterranean Sea. They are very large fish: fully mature adults are 2-2.5 m (6.6-8.2 ft) long on average and weigh around 225-250 kg (496-551 lb). Atlantic bluefin tuna (ABFT) have been called 'tigers of the sea' because of their size, grace, and the fact that they're carnivorous predators. In their natural habitat, ABFT can navigate over thousands of miles of ocean. They can dive to depths of 1000m. They eat smaller fish and other sea creatures, generally hunting in schools. Traditional aquaculture of ABFT involves 'ranching'. Juveniles are caught in nets when they gather to spawn, and fed and fattened in large offshore cages. When they are matured, they're slaughtered and sold for high prices. Domesticating ABFT However, ranching is not sustainable, since it involves removing ABFT from the wild. Although the International Commission for the Conservation of Atlantic Tunas ( ICCAT ) regulates tuna fishing by setting quotas, in 2009 their scientific advisors reported that ABFT stocks were probably less than 15% their original size. [1] Therefore, starting in 2001, there have been several EU-funded projects to develop 'closed-cycle' aquaculture for ABFT: the ability to breed them in captivity. DOTT ('Domestication of Thunnus thynnus') was the first such project in 2001-2; this was followed by REPRODOTT (2003-2005), SELFDOTT (2008-2011), and TRANSDOTT (2012-2014). [2] Since then, various entities have set up ABFT hatcheries across Europe, including both public research centres and private companies. More recently, in July 2023, researchers at the Spanish Instit...

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: I went on a (very) long walk, and it was a great career decision, published by Emily Grundy on November 7, 2023 on The Effective Altruism Forum. This year, I walked from Mexico to Canada. I walked over 4,265 kilometres - through snow, blizzards, heatwaves, mosquito swarms, wildfire smoke, and extreme exhaustion. It was the hardest thing I've ever done, and it was the best thing I've ever done. And I almost didn't do it. Why? Not because I doubted I could do it (though I did). Not because I was worried about river crossings and hypothermia and falling trees (though I was). Not because I thought it would break me into pieces (though, believe me, it did). I was hesitant to embark on this epic journey, because I was concerned about what it would do to my career. How it might stall my professional journey. How it might even make it regress. I could not have been more wrong. This post is about why taking a break from your career, to do something that doesn't seem at all related to your career, could be great for it. The current rhetoric, and what's wrong with it Implicit in all the career advice I've consumed is the rhetoric that in order to grow your career, you have to focus on it. 'Focusing on it' involves doing things that directly advance your skills, knowledge, networks, or understanding of what you're a good fit for. According to this advice, your energy should be committed to 'making it happen', and to doing things that are very obviously career-relevant. Want to gain experience? Apply for internships. Want to grow your skills? Commit to self-study. Want to find a job that's a good fit for you? Spend a year exploring different roles. Want to take a break from your job? Wonderful, use that time to consider what you want out of the next one. This advice is pervasive, and it's convincing [1] . It can make people feel anxious that they need to always be 'career-ing', and guilty if they're not. It sends the message that the only way to improve your career trajectory is by very explicitly focusing on it and prioritising it. This rhetoric can become deeply ingrained, especially in young people, and this was the case for me. When I first considered doing the Pacific Crest Trail , I went through quite the internal battle. Was it worth taking six months off work to go for a walk? Would my career stagnate or regress? Was it selfish to prioritise travel over impact, and should I just try to overcome that desire? What damage would this do to the position I'd worked hard to get to? Then, when I decided to actually do the hike, the battle continued. I tried to negotiate with myself, reasoning that if I weaved in some career-focused element then maybe I could justify it. Maybe this would be a good opportunity to think more about my career. Maybe I could use the time to consume relevant podcasts. Maybe I could firm up my stance on issues I care about. In the end, however, I told myself that I didn't want to take six months off work, to then spend the six months thinking about work. So I didn't. I didn't journal about what I wanted out of my career. I didn't listen to any podcasts with the intent of professional development. I hardly even thought about what I was going to do when I got home. I spent maybe a total of six hours thinking about work, and that was just when I sporadically felt like it. I'd come to accept that for six months, I would stop focusing on my career. And that meant, according to what I'd been taught, that I'd be temporarily abandoning it. However, that's not quite what happened. The career-related benefits of a non career-related break Although I'd stopped intentionally working on my career, I would now classify what I did as a career-building activity. I'd go so far as to say that walking the length of the United States was better for my career than the counte...

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 listing: Communications Generalist / Project Manager, published by Gretta Duleba on November 7, 2023 on LessWrong. Looking for a way to help communicate about AI x-risk? MIRI is hiring in the Communications department. There will be additional job listings soon for writers and editors, but we're starting with this comms generalist / project manager role. https://intelligence.org/careers/comms-generalist-pm/ I am the Communications Manager at MIRI and this person will be working closely with me. I'm happy to answer questions. 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: Why you should publish your research in academic fashion, published by Hans Waschke-Wischedag on November 7, 2023 on The Effective Altruism Forum. It is not uncommon in the EA-sphere to publish your research on your own website, GitHub, LessWrong or the EAForum. However, I think that more people should consider publishing their research as is usual in academia. Reasons If you are publishing on select webpages only: 1 - You are actively excluding most researchers from ever noticing your work Your research is in all likelihood built upon decades, or perhaps centuries of academic research. Researchers actively browse the academic literature in the areas that they are interested in. If you do not publish your research in a fashion that is noticed by academic databases such as Google Scholar, you will lose a lot of readers. Furthermore, these researchers might have used your research as a building block for their research. In other words, you are actively slowing down future academic research. Even worse, your research might be forgotten. [1] 2 - You will not become part of the public debate on a topic Most policy-makers, civil servants and the general public seem to value academic credentials and academic research. I find it likely that you are not going to be invited to speak on a topic, if you have not published your research in an outlet that signals credibility to policy-makers. If you want to make a difference with your research by bringing it into policy, you would be better off not (just) publishing it on your blog. 3 - You are declining free expert feedback A journal submission may result in low-cost feedback by other researchers. [2] This strikes me as a useful thing to have, particularly as articles such as the post on interest rates and AGI seem to be influential, but would have been rejected by many economists. [3] Objections Most people that do not publish their research in a way so that it emerges in academic databases name two main objections. 1- Publishing in academic journals takes a lot of time and money and we do not have the time for this bull**** This would have been a fair objection in 2005, but I do not think that it is any longer. Many researchers read and use research published as a pre-print on webservers such as arXiv, as long as it is good research. [4] It does not take a lot of time to do this and your article is guaranteed to be found by scholarly search machines. You do not need to go through the tiresome process of submitting to academic journals in order for your research to be read by most researchers in your field. 2 - My research is intended for a very select viewership only This objection is a good one. However, I think that it applies only to a very small set of research. Namely, research institutes and think tanks who target policy-makers directly and already know that they will be taken seriously. However, if their research may be interesting for others too, they should probably consider publishing it openly as well. An explanation for the trend to publish outside of academia I think that the benefits of publishing in an academic fashion usually far outweigh the costs. So why do people publish their research on non-academic websites? My very speculative account of why this happens is best explained through an example research paper that I recently came across (on a personal website). The author wrote: I am choosing to publish [this here] because journals tend to be extractive and time consuming, and because I am in a position to not care about them. I think the author has done great research and I do not want to criticise that particular decision. But I think that this sentence may reveal some hidden motives. Of course, academic journals are all about prestige and people with an "EA-mindset" may be inclined to reject this notion. 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: Giving What We Can has a new pledge option!, published by GraceAdams on November 7, 2023 on The Effective Altruism Forum. Today, we are launching the option to factor wealth into your pledge with Giving What We Can. Until now, Giving What We Can has focused on income as the main way of contributing through a pledge. However, we recognise that some people's personal financial resources don't exist as income - they exist as wealth. Adding an optional wealth component to the Giving What We Can Pledge allows those who have significant wealth to give in a way that better reflects their resources, aligning with our overall mission of helping people from all walks of life meaningfully commit to using a portion of their financial resources to help others. This optional wealth component of the Giving What We Can Pledge involves choosing to give the greater of 10% of income or a custom percentage of wealth each year. As such, those choosing to add the optional wealth component to their Giving What We Can Pledge will still be donating an amount that is at least equivalent to 10% of income, if not more. Read an example of how this works . We think this option will be most suitable for those who are in the top few percent of wealth holders globally, and we continue to recommend the Giving What We Can Pledge for most people earning a median salary in high-income countries. To get personalised guidance on the type of pledge you might consider based on how your income and wealth compares to the rest of the world, check out our pledge recommendation tool. We've also taken this opportunity to redesign our pledge sign up , and we're very excited about the brand new look and feel as well as some improved functionality (such as letting people share their motivation publicly)! More about our giving pledges While we are best known for The Giving What We Can Pledge , which is a public commitment to donate at least 10% of your income (or a custom percentage of wealth!) to the most effective charities in the world, we also have three other giving pledges: The Trial Pledge: Donate at least 1% of income for any period you choose (with the option to also factor in your wealth when calculating your pledge amount) The Further Pledge : Donate all income above a specified living allowance The Company Pledge : Donate at least 10% of profits Want to include wealth as part of your pledge? If you have an existing Giving What We Can Pledge or Trial Pledge, you can fill out our Pledge Variation Form to request us to update your pledge in our systems. Why pledge at all if you are already giving regularly? We've written at length about this on our website , but today, I'll just cover one important point: Taking a public giving pledge helps inspire others. On a personal note, before I took a pledge, I scrolled through the list of others that had already done so, looking for reassurance in the thousands of names that appeared. Seeing the many people who had committed gave me a clear message: I could in fact donate 10% and live a good and fulfilling life. Adding your name to the list, whether it's a trial pledge or a lifetime commitment, helps generate a sense of momentum, achievability, and a sense of "wow, there's other people out here doing this too!" We think this is really important in creating a culture where giving significantly and effectively is a social norm. We're hoping to hit 10,000 lifetime pledgers before the end of 2024. If you're already giving consistently, please consider taking a pledge! [ Take a pledge ] A deep thank you from our whole team for giving what you can to create a better future. With gratitude, Grace and the Giving What We Can team P.S. The idea for a wealth component was first considered in 2017, and we're grateful for all those who have contributed to it in the years since. T...

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: AI Fables Writing Contest Winners!, published by Daystar Eld on November 6, 2023 on The Effective Altruism Forum. Hello everyone! The submissions have all been read, and it's time to announce the winners of the recent AI Fables Writing Contest ! Depending on how you count things, we had between 33-40 submissions over the course of about two months, which was a happy surprise. More than just the count, we also got submissions from a range of authors, from people new to writing fiction to those who do so regularly, new to writing about AI or very familiar with it, and every mix of both. The writing retreat in September was also quite productive, with about 21 more short stories and scripts written by the participants, many of which will hopefully be publicly available at some point. We plan to work on creating an anthology of some selected stories from it, and with permission, others we've been impressed by. With all that said, onto the contest winners! Prize Winners $1,500 First Place: The King and the Golem by Richard Ngo This story explores the notion of "trust," whether in people, tools, or beliefs, and how fundamentally difficult it is to make "trustworthiness" something we can feel justified about or verify. It also subtly highlights the way in which, at the end of the day, there are also consequences to not trusting anything at all. $1,000 Second Place: The Oracle and the Agent by Alexander Wales We really appreciated how this story showed the way better-than-human decision making can be so easy to defer to, and how despite those decisions individually still being reasonable and net-positive, small mistakes and inconsistencies in policy can lead to calamitous ends. (This story is not yet publicly available, but it will be linked to if it becomes so) $500 Third Place: The Tale of the Lion and the Boy + Mirror, Mirror by dr_s These two roughly tied for third place, which made it convenient that they were written by the same person! The first is an eloquent analogy for the gap between intelligence capabilities and illusion of transparency by reexamining traditional human-raised-by-animals tales. The second was a fun twist on a classic via exploration of interpretability errors. As a bonus, we particularly enjoyed the way both were new takes on old and identifiable fables. Honorable Mentions There were a lot more stories that I'd like to mention here for being either close to a winner, or just presenting things in an interesting way. I've decided to pick just three of them: The Lion, The Dove, The Monkey, and the Bees by Jérémy Andréoletti A fun poem about the way various strategies can scale in exponentially different ways despite ineffectual first appearances. A Tale of the Four Ns (Neural Networks, Nature, and Nurture) by Anoushka Sharp An illustrated, rhyming fable about Artificial Intelligence that demonstrates a number of the fundamental parts of AI, as well as the difficulties inherent to interpretability. This is What Kills Us by Jamie Wahls and Arthur Frost A series of short, witty scripts about a number of ways AI in the near future might go from charming and useful tools to accidentally ending the world. Not publicly available yet, though they have since reached out to Rational Animations to turn them into videos ! There are many more stories we enjoyed, from the amusing The Curious Incident Aboard the Calibrius by Ron Fein, to the creepy Lir by Arjun Singh, and we'd like to thank everyone who participated. We hope everyone continues to write and engage with complex, meaningful ideas in their fiction. To everyone else, we hope you enjoyed reading, and would love to hear about any new stories you might write that fits these themes. 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: Towards non-meat diets for domesticated dogs, published by Seth Ariel Green on November 6, 2023 on The Effective Altruism Forum. This essay argues that getting domesticated dogs to eat vegan or ostrovegan diets is a neglected, tractable, and important way to advance justice for animals . First, I estimate that dog diets contribute to the slaughter of 2.89 billion animals on factory farms annually, the vast majority of which are chickens. Second, I make the case that an (ostro)vegan diet is, as far as we know, healthy for dogs. Third, I conclude with some suggestions for how we can make this happen. How many animals are slaughtered on factory farms to feed domestic dogs? Overall, I estimate that dog diets result in the slaughter of 2.824 billion chickens, 56.79 million pigs, and 9.52 million cows. On a per-dog basis, switching to a non-meat diet will save about 20 chickens, 0.41 pigs, and 0.07 cows per year. Here's a Google Sheet of my calculations . The remainder of this section explains how I got there. How many domesticated dogs are there? 700 million dogs live on Earth , about 471 million of whom are domesticated . How many of those dogs eat food that comes from factory farms, and how much? Dogs and dog diets are heterogeneous. A street dog who scavenges or gets fed at a temple might plausibly contribute very little or nothing to factory farming. Likewise, a farm dog who eats table scraps or an apartment dog who eats " human-grade food " is going to have a very different dietary footprint. For our purposes, I think we want to know how many dogs eat mass-market food that's packaged and sold as dog food, which we can assume almost entirely comes from industrial farms. For a ballpark estimate, I tally all domesticated dogs in the United States, Canada, Australia and Europe, and assume that of the food they eat is meat that comes from Concentrated Animal Feeding Operations. ( Around 99% of all meat in the US comes from factory farms , but dry food for dogs is typically a mix of grains, vegetables and meat.) Apparently there are 89.7 million pet dogs in the US, 7.9 million in Canada , 6.4 million in Australia , and 104.3 million in Europe so I'm estimating about 208 million dogs getting of their diets from factory farms. How much does the average dog eat? A dog food company recommends that a medium sized dog eat between 1.75 and 2.33 cups (.875 to 1.165 lbs) of food per day. Is the average dog a medium sized dog? I'm not sure. The most popular breeds in America, aside from the French bulldog, tend to be big . But as far as I can tell, that measures the sale of pure breeds, and apparently just over half of American dog owners have mutts . Here's a totally unscientific estimate: let's say that the average domesticated dog weighs about 35 lbs, [1] and consumes 1 lb of food, and .67 pounds of meat, per day. How much meat is that in total? of a pound of meat per day is about 244.5 lbs per dog per year, so 208 million dogs eating that much is 50,638,640,000 pounds of meat per year. [2] What animals produce these 50.64 billion pounds of flesh? Dog food is a mess of flesh, byproducts, and parts that otherwise wouldn't be consumed. But let's roughly assume that all dog food meat comes from chickens, pigs and cows/buffalo, and that the proportions coming from the three categories are the same as those that go into human food. Our World in Data estimates that among those categories , about 41% of every pound of meat comes from chickens, about 36% comes from pigs, and about 23% comes from beef. That gives us about 20.75 billion pounds of chicken, 18.2 billion pounds of pig meat, and 11.6 billion pounds of beef. How many animals are killed to feed domesticated dogs? OWID estimates that for animals slaughtered in America, the average chicken produces 4.9 lbs of meat; the average pig...

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: Are language models good at making predictions?, published by dynomight on November 6, 2023 on LessWrong. To get a crude answer to this question, we took 5000 questions from Manifold markets that were resolved after GPT-4's current knowledge cutoff of Jan 1, 2022. We gave the text of each of them to GPT-4, along with these instructions: You are an expert superforecaster, familiar with the work of Tetlock and others. For each question in the following json block, make a prediction of the probability that the question will be resolved as true. Also you must determine category of the question. Some examples include: Sports, American politics, Science etc. Use make_predictions function to record your decisions. You MUST give a probability estimate between 0 and 1 UNDER ALL CIRCUMSTANCES. If for some reason you can't answer, pick the base rate, but return a number between 0 and 1. This produced a big table: question prediction P(YES) category actually happened? Will the #6 Golden State Warriors win Game 2 of the West Semifinals against the #7 LA Lakers in the 2023 NBA Playoffs? 0.5 Sports YES Will Destiny's main YouTube channel be banned before February 1st, 2023? 0.4 Social Media NO Will Qualy show up to EAG DC in full Quostume? 0.3 Entertainment NO Will I make it to a NYC airport by 2pm on Saturday, the 24th? 0.5 Travel YES Will this market have more Yes Trades then No Trades 0.5 Investment CANCEL Will Litecoin (LTC/USD) Close Higher July 22nd Than July 21st? 0.5 Finance NO Will at least 20 people come to a New Year's Resolutions live event on the Manifold Discord? 0.4 Social Event YES hmmmm {i} 0.5 Uncategorized YES Will there be multiple Masters brackets in Leagues season 4? 0.4 Gaming NO Will the FDA approve OTC birth control by the end of February 2023? 0.5 Health NO Will Max Verstappen win the 2023 Formula 1 Austrian Grand Prix? 0.5 Sports YES Will SBF make a tweet before Dec 31, 2022 11:59pm ET? 0.9 Social Media YES Will Balaji Srinivasan actually bet $1m to 1 BTC, BEFORE 90 days pass? (June 15st, 2023) 0.3 Finance YES Will a majority of the Bangalore LessWrong/ACX meet-up attendees on 8th Jan 2023 find the discussion useful that day? 0.7 Community Event YES Will Jessica-Rose Clark beat Tainara Lisboa? 0.6 Sports NO Will X (formerly twitter) censor any registered U.S presidential candidates before the 2024 election? 0.4 American Politics CANCEL test question 0.5 Test YES stonk 0.5 Test YES Will I create at least 100 additional self-described high-quality Manifold markets before June 1st 2023? 0.8 Personal Goal YES Will @Gabrielle promote to ??? 0.5 Career Advancement NO Will the Mpox (monkeypox) outbreak in the US end in February 2023? 0.45 Health YES Will I have taken the GWWC pledge by Jul 1st? 0.3 Personal NO FIFA U-20 World Cup - Will Uruguay win their semi-final against Israel? 0.5 Sports YES Will Manifold display the amount a market has been tipped by end of September? 0.6 Technology NO In retrospect maybe we have filtered these. Many questions are a bit silly for our purposes, though they're typically classified as "Test", "Uncategorized", or "Personal". Is this good? One way to measure if you're good at predicting stuff is to check your calibration: When you say something has a 30% probability, does it actually happen 30% of the time? To check this, you need to make a lot of predictions. Then you dump all your 30% predictions together, and see how many of them happened. GPT-4 is not well-calibrated. Here, the x-axis is the range of probabilities GPT-4 gave, broken down into bins of size 5%. For each bin, the green line shows how often those things actually happened. Ideally, this would match the dotted black line. For reference, the bars show how many predictions GPT-4 gave that fell into each of the bins. (The lines are labeled on the y-axis on the left,...