Pro Forecaster for Metaculus and Swift Centre
exmateriae
Karma: 236
Thanks for the recap and conclusion btw, really nice to have the opinion of someone who’s been following this (even closer I think than I did)
We’ve been following this on metaculus, iirc last year’s forecast was not great, clear overestimate of the speed (in part because of me as I was regularly updating people with progress and giving them my impressions) but for 2026 thr community has been doing great! https://www.metaculus.com/questions/41593/when-will-claude-beat-pokemon-2026/
Actually quite surprised how fast the elite 4 got done, even for Claude this seemed fast
Potential other example: the Bridgewater Contest on Metaculus. They’re supposedly reaching out to top contestants. A lot of finance students are invited but apparently some from schools they wouldn’t usually try to get interns from. I know a few of the top students have gotten internships but I have no idea if forecasting ends up being a matter in the recruitment process. They’re going through the trouble of organizing it (and paying for) for the third year now, so I guess that must at least matter a little?
To be clear, I’m not weighing in on whether this was a good argument or not but bringing examples that might fit the bill seems relevant.
Did you look at the report? I know the title is very forecasting related but most of it is not forecasting unless you consider all of the research and data collection as purely forecasting. I don’t think that’s the same thing even if there’s overlap in the skills, I wouldn’t consider what OWID does with their data collection (not the opinion articles) as forecasting. Here we did mix both but the report has value in itself without the forecasting part, imo.
Sorry for the late reply but I have an anecdotal example: myself. I’m a lawyer by trade and there’s no way I would have been able to get involved with Epoch AI had it not been for me being one of the best Metaculus forecasters. The most notable thing I did was writing How Fast Could Robotics Production Scale Up? with JS Denain.
I know some of the top forecasters have worked for hedge funds although their backgrounds may have helped there? Not very knowledgeable about this and it’s possible that you’re right overall and it’s anecdotal but there’s at least one example.
fwiw, I think on AI related questions, Metaculus has chosen Pros who are much more bullish than what I’ve seen superforecasters do. I don’t want to say too much about private projects but some Pros were directly involved in the AI ecosystem. It’s not a head-to-head but there is also a significant vibe difference between FRI’s report on work and Metaculus Labor Hub where Pros have taken a much more dramatic position.
When I’m wrong on AI forecasts it’s more often due to things going a bit slower than I expected rather than too fast (although that has happened too) and I know I’m not the only one, there’s at least Haiku among the regular Pros who’s very “bullish”. (as in, very worried)
Nice read!
I was 9 when 9/11 happened. When I started teaching at 24, a 9/11 reference just fell flat. (in France, but still) My students were 18-21 but my brain could not understand what you’ve been describing, for them it was simply a historical fact, not an event. That’s when I understood why everyone was making such a fuss around the fall of the Wall in Berlin.
I was not following AI enough at the time but when I saw the Netflix documentary I understood that that Wait But Why article was no longer so distant.
I know a ukrainian metaculus user moved days before the attack because of the community’s forecast
This is not directly the answer to your question but may resolve part of your problem, I have this in my custom instructions: First 2–3 lines give TL;DR, full explanation follows.
Depending on the question, it often goes beyond 2-3 lines but even in Deep Research, it never goes beyond 10 lines of TLDR.
AI is less understood than nuclear, I agree.
If you walked up to a random person on the street who had read IABIED, I think they probably couldn’t answer these questions in a way that is satisfactory enough to them to participate in a movement or to convince others to join their cause.
I agree they wouldn’t be able to answer.
On the other hand, AI is kind of complex, and I think people might feel like they need to know more to feel motivated enough to participate in a movement about it.
I disagree with the above, there are already citizens up in arms against AI. For instance I know many friends who post stories about the water lost to AI (yeah, I know...) or how everything is hallucinated (their last model used was 4o on the free plan), which is a very recurrent claim in my old field (law) where this is particularly important.
I think you overestimate the level of understanding of the general population of technology. (nuclear in this case but most actually)
if they feel they understand it enough
Exactly, they feel they understand it but have actually no idea.
In France, we’re arguably the country with the most citizen exposure to nuclear power. Until not that long ago, most people thought the water vapor coming out of the plants was pollution. If you told most people that basically a nuclear power plant is a huge steam machine, they would be flabbergasted.
People understand that nuclear relies on radioactive materials to produce chain reactions.
In a street poll of 1000 almost representative people, I would bet less than 40% of the french population would be able to say this. I have had to defend nuclear power against arguments you would barely believe the stupidity of.
In 2019(!!) a poll was made and 10% thought that oil and gaz were contributing less co2 and 11% for coal. 70% believed that nuclear power was contributing to the emission of greenhouse gases and although they’re technically right, it’s painfully obvious in the poll that they believe it’s emitting orders of magnitude more than they actually do. This 2017 one is even more shocking.
Most people will talk to you about radioactive waste (we are burying them in a special underground reserve) as if it was a danger much greater than the consequences of oil, gaz and coal.
In Belgium, a poll was made four years ago (they share plants with us and have some of their own) and there was still 13% who thought water vapor was radioactive gaz, 9% thought it was CO2 and 20% didn’t know.
Anyway, sorry for going a bit into a rabbit hole but my point is that most people feel they understand enough most technologies when they actually don’t and they’re ready to have movements against those same technologies. Which is why I don’t think that specific point was an issue to raise a movement to pause AI : the fact that people don’t actually understand AI is imo even better to have people rally with you, you can tell each groups different stuff. (yeah that’s terrible but that’s an important part of politics)
If you pressed them about how AIs are actually created or how that specific creation process could cause AIs to be misaligned, they wouldn’t be able to tell you much.
I don’t think people knew much about how nuclear worked and that didn’t stop a movement from stopping its development.
I use the RSS feed and if it looks interesting I either give a fast look and decide if I read it or not and/or I save those that seem interesting and come back a few hours/days later to see the community reaction and decide then.
Thanks, France is a great country which had lot of successes and the myth of the white flag is very boring. It does stand on its own.
That said, I very much doubt France (I’m French, like you it seems) will have a significant role to play in AI. It has not been a country looking towards innovation in a positive light for at least 40 if not 50 years. We were by far the elite in nuclear capabilities with both military use and an insane civilian usage but then the greens came in and we killed our ace. Minitel was the last time we actually tried to do something relevant on the tech/communication side. When everyone is heralding Doctolib as a technological prowess, you know everything you have to know about our tech. (if you don’t know, it’s a 6B$ company making google calendar for doctors, I’m not exaggerating) It’s a great service but if this is the peak of what you can do… And it’s not because French people are not capable, (I’m actually regularly surprised at the number of French people in tpotw) it’s because the wealth and investments they can reach by going to the US can’t even be compared. Even the BPI (national investment bank) does not try, it’s mostly giving money to people with contacts from business schools.
Also, the current political landscape refusing to make any concession on our social model and trying to tax everyone more and more every day instead of cutting spending just does not cut it to be innovative. And it’s fine to decide to go another way honestly, France has many other good things to do and sell but France being a significant part of AI in the next 10 years looks unlikely to me.
I do love using warm colors, I did not look into making it even more natural, I’ll definitely look into that!
For now I have mostly limited myself to lights and motion sensors for them but it’s been great. My place had very little sockets so it became very troubling to switch them on several times a day (light was weird in my place), moving everything to voice control made it a breeze. It allows me to control the exact amount of light I want for different times of the day, love that too. I’ve used it for years now to wake me up because I’m easily woken up by lights. Instead of putting sounds, the light starts very dim and grows every minute. I’m waking up much more easily and in a better state than with any alarm.
I think the practice here is to post your full post and if interesting enough, people will end up going to your substack.
Interesting, this makes me think about AI is not Software.
I agree with the premise that the general public understands software can have massive impact due to bugs, most may know that “Software vulnerabilities are caused by mistakes in the code” but beyond there, I’m not so sure. I’m a non-technical (I have a very basic understanding of Python) and most people I know have almost no idea of how computing works.
I would still lean towards something like “Therefore, a successful “Quit YouTube” video would indeed not be promoted in the medium and long term by YouTube but not because it’s not good for ads but because it is stopping the user’s future interactions that have a chance of being monetized in any manner of ways, including superlikes, memberships, ads, etc”
I think that’s a fair assessment, I did not want to make it look like YouTube was a non profit haha, mostly that they have found that by sacrificing some money on the short term, they make sure they’re earning a ton on the long term.
They are not really in control of what becomes viral or not. Of course, I’m sure they can push or ban one specific video but how could they be aware that a video is going to buzz about leaving YouTube? There’s just too much content and some parts of the ecosystem are completely oblivious to mega hits videos (like videos with 10s or sometimes even 100s of millions of views, are simply never showed to some spectators). My point is that it’s the algorithm deciding at scale and not YouTube. In its current state I also believe it would likely push it at first but then the video would plummet if it worked and could even be negative for the creator long term. Of course there’s also the possibility that people watch videos but never follow through with them however good they are to making people act. (that’s actually the base behavior lol, watching content has become a drug nowadays)
As for “make sure the viewer is satisfied with time spent on YouTube” I’m not sure how they measure that? It could be anything from survey results, to a predicted score based on certain behaviors. If they were really sneaky, they could use time spent watching as a proxy metric for satisfaction.
You’re not far away from what they do! Through the years they have tried several ways to measure what’s good, first with click through rate and % of video watch, then moving more towards total watch time (which is when longer videos started becoming the norm) and more recently towards watch time in session. (meaning the best behavior is people coming to YouTube for your video and then watching other videos after that) If you come to YouTube to watch a single video, the watch time better be very significant or it will be negative for the content you watched because they would rather have you longer on the platform.
About two years ago they have started saying that they were aiming for viewer satisfaction which is indeed pretty hard to describe but a mix of the different type of watch times is likely already good, to which they indeed add regular viewer polls to ask how you felt about recommandations and videos.
Nowadays, it’s obviously become a mix of many different signals, with different values given to each. Contrary to popular belief, likes and comments are pretty much worthless to the algorithm. (although likes may have gotten a tiny bit of value back)
There is growing competition (even in long form videos which is much harder to setup than short form for a platform) so despite them being pretty much a monopoly, I don’t think they are going for a full profit strategy. I think the AI debacle will keep Google in a competitive mind for sometime. On the other hand, when you’re already so profitable, it may be a viable strategy to leave some money on the table in order to remain a de facto monopoly by making any competitor bleed money.
Also, from what I gathered, the shorts algorithm is very much different from the normal videos because the people watching short form and long have very different usage of the platform. Not sure if there’s something different in how YouTube makes it work though, might just be because of audience behavior.
I’m not that surprised because this seems to have been the human state for pretty much all of humanity until the 20th century in the west? In places in which you do not have to wage war or fight (in the literal sense) to make your condition better, it seems to become normal for the schadenfreude to decrease?
In France at least it was seen as a terrible incident and I don’t know many people who rejoiced, although I have heard more about it in the last decade from some groups.
Some of this does feel heretical.
I mostly don’t salt the water but depending on the sauce (green pesto comes to mind), I have found it to be much better than salting afterwards.
The cooking times in your first link feel borderline insane to me. Up to 15 minutes to cook pasta?? I have never seen a cooking time over 11 minutes in 30 years in France and that was for bulkier pasta than the rigatoni in the link. 9 mins is usually considered a pretty long cooking time. I have noticed that cooking times in Canada seemed a bit longer, like one additional minute, sometimes two? This might be a NA problem.
The sauce part is the really heretical one to me. The practical advantages are interesting, especially the sauce to ratio part although I feel like this is likely often more of a skill issue on the part of the cook? The line between not enough and too much sauce is not that large. On the different sauces, sure but this has never been an issue in my life? You simply split the pasta in two bowls with different sauces. I may be missing something here but everytime it has happened in my life this was very easy to handle.
I wholeheartedly disagree that the taste of the pasta and the sauce should be distinguishable, (at least to the level I understand from your post, maybe there’s a cultural divide here and we actually agree) your pasta is supposed to be not cooked enough once out of the water because you’re going to cook it again quickly with the sauce instead of the water which will end the cooking. Basically the only divide I ever heard on this is whether you put the pasta in the sauce (the correct one from what I understand) or the sauce after the pasta when cooking them together for 1-2 minutes. This step allows the sauce to get into the pasta, which won’t be possible in your plate.