Author Archives: 4gravitons

Yang-Mills Exists: What Does That Mean, and What Is It Good For?

Leave a reply

Last week, OpenAI announced they tackled a longstanding problem in mathematical physics: the existence of quantum Yang-Mills theory. Since one of this blog’s first posts was an explainer about what Yang-Mills theory is, I figured I should probably comment.

In 1954, Yang and Mills wrote down a type of theory that would turn out to describe three of the four fundamental force of nature. Electromagnetism uses the simplest version of the theory, while the two nuclear forces are more complicated. Gravity is the only force that isn’t described by Yang-Mills theory.

But while Yang and Mills wrote down a theory that any physicist would be happy with, that’s not enough for mathematicians. Physicists take shortcuts and make assumptions in order to get a theory they can readily use, and for quantum theories those shortcuts and assumptions can paper over a lot of subtleties. Normally, that doesn’t matter much: when physicists want to use Yang-Mills theories to understand the real world, they use well-tested approximations and simulation techniques. But in a controversial-enough corner case, where you might want ironclad mathematical proof to convince skeptics, physicists wouldn’t have that option. The true theory, without any approximation tricks, wasn’t defined in terms of axioms that were proven to make sense, not with a mathematician’s carefulness. In short, despite its successes, Yang-Mills wasn’t proven to exist.

Over the years, people tried to fix this. Not many people, mind you: most physicists didn’t know enough math, most mathematicians didn’t know enough physics. But mathematical physicists had managed to find axioms and proved they worked for other theories, simpler than Yang-Mills. And a few people made progress towards more complicated theories. It’s a problem that has a lot of prestige associated with it: it’s the first step in one of the million-dollar Millennium Prize Problems in mathematics (though to be clear, just the first step: put a pin in that). But that prestige didn’t draw many people to work on it: it was viewed as too difficult and not particularly urgent, not the kind of productive direction people could build careers on.

It definitely wasn’t a type of research I understood. Hearing about it occasionally from the outside, it seemed not only mysterious but messy. It felt like there was a zoo of different assumptions and objections to those assumptions, theorems and loopholes to those theorems. It felt like the kind of research area where someone out there would be convinced that they had already solved the problem, while someone else was equally convinced that that solution was totally unsatisfactory. So even after OpenAI announced they had tackled a different Millennium problem, I didn’t expect Yang-Mills to be next on the list. Not because it was hard: it was clear to me by then that AI could solve hard problems. Rather, because it seemed like not just a mathematical problem, but a social one, of convincing people that some specific kind of approach actually did the job.

So, what happened?

I’m probably wrong about that, though we don’t quite know yet.

AI has made prestigious problems a lot more tempting. If you don’t need to dedicate your career to becoming the right kind of expert, then why not give it a shot?

Not every prestigious problem can be solved, of course. And trying to solve one you can’t solve yet is more likely to just make the AI “cheat”, like how an impossible hacking challenge got an AI to hack the website HuggingFace.

But at this point, some people with the right AI tools and the right expertise had been working on this particular problem. Some of them were Ethan Sussman and Levent Alpöge (the Anthropic employee who was involved in the controversy around OpenAI’s solution of the Navier-Stokes Millennium problem). Others, apparently, were working at OpenAI.

Much like with Navier-Stokes, OpenAI got there first. They haven’t said much about how, besides a rough timeline, which puts their result before Sussman and Alpöge posted their in-progress work online.

Instead of posting their own results online, OpenAI reached out to David Gross. David Gross isn’t known for this kind of mathematical work, but he is a Nobel prizewinner who made an important discovery about Yang-Mills. He’s also one of the founding figures of string theory.

Reaching out to David Gross was pretty clearly a way for OpenAI to try to get approval from someone considered a pillar of the theoretical physics community (at least, by the people who don’t think of him as an arrogant jerk). David Gross in turn put together a team of people to check the result. Of them, Arthur Jaffe is a major figure in making these types of theories mathematically consistent, while Michael Douglas is someone who has been trying to get people more involved in these kinds of prestigious or mathematically deep problems, and has been keeping track of progress on this problem specifically for the last twenty years. The other two, Igor Klebanov and Nikita Nekrasov, are more generally known as smart, mathematically savvy physicists and don’t have a direct connection to the topic that I’m aware of.

The team has decided not to release the result until they’ve checked it and written a decent explanation. This has annoyed a lot of people, including the Advisory Group on Mathematics and Artificial Intelligence, who think the result should be released immediately, so everyone has a fair chance to try to understand and use the result. I can appreciate some of the logic behind waiting: the AI-written papers OpenAI has released before have been pretty poorly written, the kind of thing that would be heavily edited before being published if they were written by a human. And even if the math works, the important question is whether it actually represents Yang-Mills theory, which is something you need physics expertise to answer. Still, it’s hard to argue against more transparency. My goofy recommendation is that they don’t release the paper, but do release the version of the result in the mathematical programming language Lean, and people can work to understand that.

Until they do make something public, or release a decision, we shouldn’t count on the result actually doing what it was supposed to, and proving Yang-Mills exists. It still probably does, these AI systems are increasingly reliable. But there could well be a loophole that wasn’t considered.

What is it good for?

I’ve seen physicists shrug at this result. But I think some of the intuitions there are outdated.

We already know Yang-Mills is fine. Not in a mathematically rigorous way, but through experiments and tests and more or less consistent procedures. Nobody was doubting that Yang-Mills existed in the everyday sense of the word.

Now, mathematical rigor can still help. With a nice functional set of axioms, it can become possible to prove other things. The rest of the Yang-Mills Milennium problem is proving one of those things, namely that the massless gluon particles in Yang-Mills theories group up into massive glueballs. That’s still not really in doubt, it’s been checked in computer simulations pretty extensively. But the idea is, if you can prove that, maybe you can prove other interesting things.

This hasn’t been high priority, because math is slow. Often, by the time mathematicians have carefully nailed down a result, physicists have raced ahead and done something reckless. So while some physicists will pay lip service to the idea of getting mathematicians interested, most are pessimistic about it helping.

But I’ve seen a few people put two and two together here, and notice that math is no longer slow.

AI is good at math, better than it is at most other things. That isn’t because it’s some ultra-logical beep boop sci-fi robot. It’s because the best way to get an AI good at something is to train it on a huge number of examples of that thing, and you can generate a lot more examples automatically than you can by hand. So an AI trained on math can use the Lean programming language during training to check it got the answer right, rather than waiting for human graders. That makes it a lot easier to train AI to be good at math.

So when a part of physics becomes expressible in Lean code, it gets a lot easier for AI to get good at it.

What happens when an AI gets good at Yang-Mills, or similar physics theories?

First, the rest of the Millennium problem. If it can’t do that, it probably can’t do much else.

But after that?

Physics is an experimental science. You can’t prove a theory is correct with math alone, you have to test it.

But many of the most controversial questions in physics aren’t actually questions about experiments. They’re questions about theory.

You can’t use mathematics alone to prove which theory of quantum gravity is correct, you eventually need experiments. But as it happens, every theory of quantum gravity is controversial for mathematical reasons as well. String theorists believe there is one unique theory that unites the different types of string theory, others think they’re full of hot air. Loop quantum gravity researchers think their theory can describe a world like ours, others think the theory forces space and time into a crumpled up wad of nonsense. Asymptotic safety researchers assemble evidence that gravity has a special property that keeps it safe from the problems that make quantum gravity seem hard, others think the evidence doesn’t go far enough.

If any of those theories had axioms, maybe someone could write a proof to settle one of these controversies. If AI gets good enough at this, it might be able to.

Now, all of this still might not matter very much. Right now, AI contributions to math and physics seem to be a mix of public relations and advertising, benchmarks (where AI companies test how powerful their setup is by seeing how many math problems it can solve) and individual employees playing around with the system in their free time. That mix makes it hard to predict what will be tackled next. If making a system able to resolve big longstanding physics questions is an expensive detour and doesn’t look like it will generate good PR, then it might not make sense for AI companies to pursue it, and if the technology stays expensive it may be out of reach of users. But if it’s cheap, or for some reason they think it’s a high priority, then it might happen.

As always these days, who knows?

An AI Apocalypse, but Just for You

1 Reply

For Halloween, I like to talk about spooky scenarios. And these days, the most trendy one is an AI apocalypse.

You can run into a wild range of ideas for how AI could go wrong, of varying plausibility. They go from hacks shutting down key infrastructure or engineered viruses, to taking over nuclear arsenals and sending out armies of killer robots, all the way to nanotech that devours the world and flawlessly manipulative plans that end up filling space with computers simulating hell at the speed of light.

And of course, on the other end, the possibility that banks suddenly become convinced that the investments underpinning AI companies’ trillion dollar evaluations aren’t going to pay, collapsing the economy.

So, lots of things to be scared of this year!

But in all of those scenarios, at least we suffer together. If civilization cracks like an egg, everyone who survives has to adapt and figure out a new path forward. And if everyone’s dead, well, everyone’s dead. There’s something oddly comforting about that.

That’s not guaranteed, though.

I occasionally see quotes from “100+ reactions to 100+ solutions”, a compilation of thoughts by mathematicians about OpenAI’s recent claim to have solved hundreds of difficult open problems in mathematics. For the most part, these people aren’t writing about an AI apocalypse that sweeps civilization into a new form. They’re writing about what AI seems to be capable of in mathematics, right now, and wondering how they personally will withstand it.

Picture a world where you’ve defined yourself around something. Maybe you’re competitive, someone who wants to be the best in the world. Maybe you just want something of your own, accomplishments that came from you, not anyone else. Either way, it’s something you love, something you chose a strange and difficult path to pursue.

Then, that identity falls apart. For the competitive, you’re not the best anymore, not at the kind of thing you were doing, you’ll never be the best at that again. For the rest, your sense of ownership is less secure. Someone could hear about that problem you’re tinkering with, decide to tackle it on a whim, and jump so far ahead that you’d never be able to catch up on your own merits.

And if there isn’t a wider collapse, you’re facing this on your own. You might have to find a new job, though in a world where universities still need math classes that seems unlikely. But you will have to find a new purpose. You’d have to become a different kind of mathematician, at minimum, one more dedicated to something the AI can’t do and less dedicated to the things that happened to define your style of mathematics. Your friends in other fields will never quite understand. Your friends outside of academia will understand even less.

If we thread the needle between the apocalypses of powerful AI and an economic bubble that pops so hard that companies can’t host the models anymore, we end up with a series of quieter, more personal AI apocalypses, people who feel like their own world is ending, and have to figure out how to go on. For now, that’s the possibility many mathematicians are considering. We’ll see how many others consider it in the years to come.

The Weird Incentives of Word Count

2 Replies

Many freelancers get paid per hour, or per day. Some have standard rates for typical tasks.

Journalism is rarely typical. And usually, we’re paid per word.

I’ve had many friends whose first thought is how easy that must be to exploit. Often they remember a story about a writer like Charles Dickens or Victor Hugo padding their books with digressions to up their pay. Those stories probably didn’t happen, and these days, that method is even less likely to work. When I pitch a piece, the editor and I agree beforehand on a target word count. The only way I get paid more than expected is if the editor agrees that it makes sense.

That said, there definitely are ways to exploit being paid by the word, if you wanted to. They’re just more subtle.

The source of any such exploit is to notice that writing is not actually the hard part of journalism. Research is time-consuming: the more people you talk to, the longer it takes. Writing, once you’re used to it, can go quite quickly. So the length of a piece itself doesn’t determine how much work it takes to write: research does.

In principle, longer pieces are supposed to take more research. Editors will have rough expectations for the number of sources you’ll interview, so this isn’t something you can skimp on, per se. And publications known for their high standards for research usually pay larger rates per-word.

But no matter how much you’re paid, research can go out of hand in the other direction. Journalism is about finding out the truth, and sometimes the truth is more complicated than expected. Sometimes you need more information, and need to fit in one more interview, then another, or run around chasing after missing info. Then you need to figure out how to fit it all into the same word count you started with.

So suppose you’re a freelancer who aspires to consistently pay their bills. You want to earn a certain amount per hour worked, but some projects can take much more work than others. So what do you do?

You look for the simple. You try to find stories that are unlikely to be complicated, that won’t have clashes of opinion that need to be represented or experts with different sides of the story. If you don’t find such a story, you simplify: you write a piece that makes the story sound cohesive, that papers over doubt, even if you’ve run into doubts yourself while writing it. If you’re consistent about it, you can do what your friends with salaries do, and earn a stable living.

Not the kind of exploit you had in mind, I suspect. And not, despite the sound of it, something you can do reliably. I find I’m always tempted by complexity, always drawn to investigate a story that’s a bit too big for the wordcount it’s supposed to fit.

But it is an ever-present incentive, one I suspect people with a larger roster of leads start to feel. If you’re paid by the word, then the best story tends to be a tidy one.

Showing Me Your Amateur Breakthrough Is Evidence That It’s Wrong

1 Reply

Ok, people who read my post on the Anthropic science blog: we need to talk.

It’s understandable. You saw me write about AI being used to tackle a serious, publishable physics problem. Naturally, some of you want to share that you accomplished something even more impressive with AI, something the scientific journals and arXiv moderators won’t acknowledge. Maybe you’re hoping I’ll write about it, maybe you want me to check your math. Maybe you just think it’s neat.

But here’s the thing: most of the time, I’m not the right person to send those ideas to. And the fact that you didn’t figure that out is a bad sign.

Outside of the movies, scientists are specialists. Not just “physics” or “particle physics” or “theoretical particle physics”, but so specific that they often focus on particular methods, or are geared around particular goals. For any idea you propose, if it’s good, there will be a small list of people who will be excited to see it and able to evaluate it. They’ll be people who are trying to accomplish the same thing, or people who used the methods you’re using for other things. Those people aren’t hard to find, if your idea is in a field you’re actually interested in: they’re the people who write papers on the subject. If you find the topic interesting, you’ll have read their papers, and know who to reach out to.

So even back when I was active as a physicist, there was only a small list of ideas where I’d be the right person to talk to. Now, though, as I mentioned in the Anthropic post, I’m not a scientist anymore. I haven’t been following the field in enough detail to actually verify anything in detail, even on topics I used to research. Part of the punchline of the post was that I don’t have a clear idea of what’s possible in the field right now. So even if you’ve figured out something in scattering amplitudes, you should reach out to someone who’s still researching the subject, a few of whom were even mentioned in the post!

Now, it might be you’re not interested in someone checking your work, and instead you’re writing to me in my capacity as a science journalist, because you feel like you did something newsworthy. But as a science journalist, I’m not supposed to evaluate your work by myself. I’m supposed to reach out to an independent expert, and if I can’t find any who will speak positively about your result, then the editors won’t publish it. So if you think you did something big enough to merit news coverage, the first thing to do is once again, reach out to a scientist who actually has the expertise to evaluate your work, and get them excited about it. Then you two can contact me together!

If you don’t know who to contact, that’s a sign you don’t read papers in the field. It makes it more likely you managed to convince yourself of something nonsensical, because you don’t have the expertise to evaluate it. It’s a pretty clear sign that I shouldn’t trust you or your AI methods about the result.

It’s also just plain bad for you. If you produce a paper without ever wanting to read any, it suggests your priorities are in a dangerous place: that you’re trying to do something that sounds cool and important over doing something you actually care about.

I’m coming from personal experience here.

As an academic, I spent years telling myself that not enjoying reading papers or going to talks was just part of the job. Then it really cost me something, and I realized the job wasn’t worth my superficial desire to be a physicist. So I left.

Then, I spent a year trying to get a job in industry, and was told the same thing. They wanted to hire people who cared about the job. The finance guys wanted to hire people who read books about finance. The data scientists wanted to hire people who did projects in data science. They wanted, if not passion, then at least grace. Someone who could take to the job easily, and naturally.

Now, what I do for a living is something I was doing once a week, for no more incentive than a vague sense of personal obligation, for over a decade. And I’m doing a lot better.

You want to do something impressive with AI? You think you’re a whiz with Claude Code, that your custom workflow is immune to AI hallucinations?

Then use it for something else. Pick something you actually enjoy reading about, not just the pop news, but the real technical literature. Something you don’t need the AI to explain to you. Something that matters to you, in a way that actually counts.

At the very least, it’ll work out better than writing to me.

Congratulations to Francis Halzen!

Leave a reply

The 2026 Physics Nobel Prize was awarded this week, going to Francis Halzen for the work that led to the IceCube Neutrino Observatory.

Does that image look weird, like something is missing? It should. This is the first time in over thirty years that the physics Nobel was awarded to just a single person. Most years, it goes to three people, even if they have to awkwardly jam loosely related achievements together to make it work. Some years, it goes to two. The last time one person received the whole prize was 1992.

So why didn’t Halzen have to share the prize like Penrose did in 2020 and Parisi did in 2021? As the saying goes, your guess is as good as mine.

I can say that I was pleasantly surprised to read the announcement and see a name I recognized. I probably couldn’t name the main figures behind most big experiments. If the prize had gone to another founding figure, I’d be out of luck. But I’d heard of Halzen. And specifically, I’d heard about him in the context of the long road it took to get IceCube working.

The IceCube Neutrino Observatory is on the face of it a crazy idea. Neutrinos are almost undetectable, particles that can whiz through a light-year of lead without noticing a thing. To detect them, other experiments had build massive underground tanks of heavy water, shielded from outside interference and covered in cameras, so that on the extremely rare occasions when a neutrino happens to nudge a water molecule, it can be recorded and measured.

IceCube does this on another level, in wild surroundings, by sinking a grid of cameras into a kilometer of Antarctic ice.

So when I saw someone else propose a wild particle detection idea, I talked to my contacts at IceCube, to get an idea of how those wild ideas get fleshed out. And they told me about Halzen.

What they told me, what resonated again and again in Halzen’s writing, and what keeps coming up in quotes from him today, is that it took both luck and persistence. Nature needed to cooperate, there was no guarantee that IceCube would work. But also, the experiment couldn’t have existed without predecessors. He built off plans to put detectors on the bottom of the ocean, and tried out smaller experiments in Greenland.

Those experiments intrigue me, because they filled two roles. On the one hand, they were in a sense mere prototypes, and theorists had strong arguments that they wouldn’t be able to detect anything of interest. On the other hand, those arguments could have been wrong, and the experiments wouldn’t have been made if they could only be justified as prototypes: they were made because there was a chance they could actually see interesting neutrinos.

That combination strikes me as fundamental. We never know, in science, if an experiment will work. So every experiment, deep down, serves this combined purpose. It has to have a chance, even if low and poorly understood, of seeing something new and interesting. And it has to serve as a prototype, to teach us how to make the next experiment. Experiments are almost never one or the other, they are almost always both.

And eventually, as for Halzen, we get lucky, and nature cooperates.

Beginnings Don’t Matter as Much as You’d Think

Leave a reply

I’ve met people who think they’re uncomfortable with evolution, but when you actually talk to them, they’re worried about something very different. They ask, “How could something come from nothing? How could life come from non-life?”

Biologists aren’t sure of the origin of life. They’re working on it, there’s a whole research field on what the experts call abiogenesis. But it’s hard going, because there’s so little direct evidence.

There is a lot more evidence for evolution. Evolution isn’t a theory of where life comes from, it was never meant to be one. It’s a theory that explains the copious evidence we have from the history of life, from the fossil record to genetic comparisons to leftover features in related animals. It’s a theory that tells you what to expect when you already have some life, and wait a few generations.

Evolution and abiogenesis are different ideas, with different domains. You can have one without the other. And when you already have life, and want to understand it, evolution is quite a bit more useful than abiogenesis.

There’s a similar story with the big bang.

Some people get mystified at the idea that the big bang made something out of nothing. But that’s not really how the big bang works.

It’s not just that the big bang wasn’t the explosion you imagine, a single point expanding in an empty space. Rather, it was a time when the entire universe, everywhere that anything could be, was in a hot dense state, not empty but brim-full.

It’s that the most important part of the big bang is what happened next. As the universe cooled, things spread away from each other, like the residents of Hilbert’s Hotel all moving one room apart. Clumps formed, electrons slowing down and getting caught by protons to make atoms, dust pulling on dust until it collapsed into stars.

The patterns of a universe that cooled this way are distributed over the sky. They tell us the kinds of atoms we expect to see in spectra of stars, the distribution of galaxies, the ripples in the afterglow of the first light to avoid getting absorbed by hot plasma. Those patterns are the core of physical cosmology, the bulk of the field’s research and its strongest argument.

In contrast, the question of where that hot dense state came from is murkier. There are a variety of pictures, different attempts to approximate a context where gravity’s quantum properties should be especially relevant by people who still don’t have a clear idea of how to think about quantum gravity.

But while the question is murky, it’s also just not very useful. Wherever that state came from, knowing it exists is usually enough to chart the rest of the universe’s history. In fact, it may even be impossible for us to know what happened before a certain time: we have only one sky’s worth of evidence, and eventually that evidence runs out.

People are used to created things: food, clothes, cars. Created things don’t change in complicated ways over time, for the most part, just break down. To understand a created thing, understanding its origin matters.

But science is full of stranger things, things with a life, or dynamics, of their own. Natural phenomena change, swiftly becoming something other than what they began as. Understanding them means understanding those changes, much more than it does understanding how they began.

It Got to My Field

18 Replies

For the folks who found this blog from Anthropic’s site, welcome!

For everyone else, I’d better give some context.

Last month, I had a blog post titled “It Only Counts When AI Gets to My Field”. The title was a joke, the content less so. I said that if some of the longstanding problems of my old research field got solved by AI, then I’d sit up and take notice.

The post turned out to be a bit of a self-fulfilling prophecy, after some folks at Anthropic read it and decided to tackle one of the outstanding problems I mentioned. They then invited me to write a post about it on their science blog.

The post is here. I recommend reading it, then coming back. The rest of this post will be a little Q&A.

Q: At the end of that post, it says Anthropic paid you for your time. Can we trust what you wrote?

A: When I agreed to write the post, I made it clear that I was going to give my own opinion, not write an ad for Anthropic. They could propose light edits, but that’s it. And I avoided signing anything with them, not even an NDA, so I could freely tell you if they pushed my boundaries.

They didn’t push my boundaries. They asked me to clarify a few things, and to give more detail on the science. I didn’t change the message or the takeaways.

The reason I asked them to pay me is that, as a freelance writer, I don’t have a salary to fall back on. Time I spend on a project like that post is time I’m not spending on journalistic projects, so if I worked on it for free I’d essentially be using my vacation time for it. The rate I’m charging them is roughly in the middle of what I’d have gotten paid if I used that time on journalism: a bit more than I would have made working for the lower-paying outlets, a bit less than I would have made with the higher-paying ones.

I’m also not expecting this to be the start of a longer-term business relationship or anything like that. So overall, I don’t think I’m incentivized to lie on their behalf. You can trust me.

Q: How confident are you that they did what they said they did?

A: I didn’t get the feeling they were lying. But I didn’t start out skeptical.

I haven’t seen the logs from the LLM, or anything like that. That’s the kind of thing I would dig into if I were more suspicious of their story. If there’s a reason to be suspicious, I’ll ask.

But so far, I don’t feel that I have much reason to be suspicious. They were able to look up details on the fly when I asked them, they didn’t seem to have a polished message they were trying to push past me. And more importantly, as I’ll mention in a bit, I don’t think anything they described is all that outlandish. It lines up, largely, with the capabilities I’d expect their Claude Science platform to have.

It’s also relevant that they didn’t use an internal model for this. It means that scientists will likely be trying out similar problems soon, so if it does turn out they exaggerated something, people are going to figure out quite quickly.

Q: So how big is this? We just saw AI solve a Millennium problem, after all.

A: This was a lot easier than a Millennium problem. But it was also a lot cheaper.

The problem they solved is one with a clear recipe, honed and explained over multiple papers. It’s something I expected to be hard to do without access to a lot of computer power, so I thought it would need to be approached with a novel technique, in order to avoid using that computer power.

In the end, it didn’t need that. As I mention in the post, another group got most of the result at around the same time, with a much smaller amount of AI assistance.

It’s not an easy recipe, to be clear. I think other people will be surprised that a reasonably affordable program like Claude Science can do this without a lot of guidance. I’m not surprised, mostly, because I’ve been paying enough attention to what people have been doing with these things, and carrying out this kind of recipe consistently is actually something the good science AIs can do right now. If you haven’t been following as closely, that’s going to be a lot more surprising.

So maybe the best way to answer this is: instead of paying $15 million to solve one of the most famous problems in mathematics, they paid $1,000 or so to make a significant next step in an ongoing research program in theoretical physics. This doesn’t tell you much about whether AI can achieve field-defining breakthroughs, but it tells you a lot about the kinds of things a theoretical physicist with $1000 spare budget can do right now.

(As an aside, it feels crazy to me that 90% or more of the compute cost was from the LLM, not the calculation itself. On the one hand, it feels nuts to essentially use ten times more computer power to do this than would have been needed if a human had done the coding. On the other hand, I’ve applied for grants that budgeted more than that per year for travel costs, and spending this kind of money on getting a result seems a lot more useful than spending it on airfare.)

Q: Isn’t it reckless to publicly post a problem for AI like that?

A: I posted my challenge before OpenAI announced their Navier-Stokes result. At that point, there had been a few awkward surprises, but for the most part AI companies had a pattern of talking to an academic first before trying to solve a problem. It’s what they did back in March.

I expected that was what they’d do this time, and got more than a little blindsided when they just solved the problem on their own and reached out to Lance afterwards. I’m lucky that Lance doesn’t seem mad at me about it, it would have been quite understandable if he was.

I did tell them that if they wanted to tackle any of the other problems in that post, they should reach out to one of the scientists involved first, before attempting it. In addition to being polite, it’s a way to make sure that they understand the problem correctly and have a plan to verify the result. They happened to pick a challenge that was particularly easy to verify, but it didn’t have to go that way.

Q: Any thoughts about AI’s future impact on science?

A: If there’s a problem you can’t solve, but you think someone with expertise in another field or better programming skills could, then it’s probably solvable with AI.

A lot of problems don’t fall into that category. I’ve seen people talking online about AI solving quantum gravity, but solving quantum gravity is a question of which bullets you’re willing to bite, not a question of technical skill. That doesn’t mean AI will never be able to address it, but if so it will come from some sort of superpersuader aspect, not merely scientific capabilities.

And of course, some fields do require experiments. People are increasingly building AI-powered labs. I’ll leave it to people from actual experimental fields to think about the potential there.

Q: Can you say something explicit about the bigger risks from AI, now?

A: As I mention in the post, I didn’t learn that much about the bigger questions from this. I’m still not an expert.

But there’s one thing I think is worth emphasizing:

This technology is clearly getting more effective over time. I don’t think they could have done this six months ago. If you’re trying to predict what will happen next, you shouldn’t just assume this is the most powerful it will get, or the cheapest. If you think there’s a limit, you need to argue for it.

Edit: One more thing I should mention, now that I’ve confirmed she’s ok with it: the idea for my challenge came from some discussions at Lancefest, a conference in honor of Lance back in June, and particularly from Anastasia Volovich, whose talk declared the nine-loop calculation “a benchmark – or a ‘challenge’ – against which any “AI takeover” should be measured”.

Data Comes From Papers

Leave a reply

There’s a lovely resource that I suspect most non-physicists don’t know about. It’s called the Particle Data Group, and its interactive site PDGLive.

If you want to know the most up-to-date information on a subatomic particle, PDGLive has your back. From their home page, you can click on familiar particles like photons (\gamma), electrons (e) or gluons and find the best info experiments have provided on properties like their mass and charge. It’s a great place to get an authoritative number for any particle physics property you’re interested in.

Of course, scientists don’t just accept one authoritative answer for anything, even whether Pluto is a planet. That’s why each measurement in PDGLive comes with an arrow you can expand to see a table of past measurements, which you can compare to.

Each measurement comes with a link to a source. And each source is not an experiment webpage, or some entry in a database. It’s a document, a publication, a paper.

In fact, PDGLive itself is just a website version of a paper, the Particle Data Group’s “Review of Particle Physics”. Click enough times, and you’ll find sections of a pdf on the site, explaining their reasoning for picking this experiment over that, emphasizing this or the other thing.

People talk about papers as how academics persuade one another, or how they show off and establish credit. But papers are also just a way to organize data. Each time an experiment figures something out, all of their reasoning and procedures are summarized together with the numbers they got. And anyone who reports those numbers to you will include a link, so you can go back, and check where the number came from.

It probably feels a bit weird, all of these numbers and technical details bottoming out in an archaic practice of writing down words for other human beings. But it means that all of the richness of the process is there somewhere, linked together and collated by the same social forces that keep track of credit, all in one navigable whole.

So when you run into a number, spare some thought for where it came from. You can probably find out.

Everybody Who Isn’t ”Viewers Like You”

Leave a reply

Last week, I talked about how truthseekers get paid. But truth-tellers and truth-seekers are different things.

Consider educational kids’ shows on public television.

Nobody who works on Sesame Street is out there uncovering new letters and numbers. Bill Nye’s show wasn’t bringing analysis fresh from the lab.

The purpose of these shows is to educate. The purpose of education is to change minds.

So who pays for educational kids’ shows on public television?

If you’re from the US and watched PBS growing up, you remember one answer: “viewers like you!” US public television is supported by donations, ordinary people across the country who want it to keep on educating kids.

But you also might remember the lists of names that came before “viewers like you”. Some of those were things like “the Department of Education” or “a grant from the National Science Foundation”: government programs, in other words. Others were philanthropists and private foundations. Some were tied to companies, like the Intel Foundation, or Juicy Juice.

All of these groups, from government departments to donors, are trying to change kids’ minds. They support specific shows on specific topics, where they want kids to be better-informed. The same groups have the same kind of impact on schools. For example, I remember in elementary school we all learned to play a recorder, because a wealthy donor had given the school recorders out of the idea that music education was especially important.

For a truth-seeker like a journalist, accepting that kind of funding would be a problem. Grants for journalists tend to support things like travel, letting journalists learn more about specific topics, not pre-judging the conclusion. But children’s television is about truth-telling, not truth-seeking, so our standards are different. We trust the people making children’s television to care about whether they’re telling the truth. And because the topics aren’t new, we don’t usually worry about their judgement being biased.

All this is rather obvious. But now, consider science YouTube.

Some science YouTubers seem to have a mission much like children’s television. They’re there to teach, not to make independent judgements. They don’t search for truth on their own. And some of them are funded by educational grants, much like children’s television.

Others are a bit more like journalists, or even activists. People follow them for their opinions, to hear their assessment. They’re trying to be truth-seekers.

On YouTube, it’s not always obvious which is which.

There’s a particular group of philanthropists called Effective Altruists, and many of them are concerned about AI. So in between funding things like anti-malaria bed nets, some of them are giving grants to YouTubers to make educational content about AI-related risks.

Apparently, they reached out to Sabine Hossenfelder, which was a bad idea. Sabine Hossenfelder’s followers aren’t just looking for education on known facts. They’re looking for her judgements, her literal bullshit-rating on ideas. And so while she’s paid by “viewers like you”, she’s not really the type to get paid by that type of grant.

What I want to emphasize, and what looked like it was getting lost in the discussion, was that their pitch would have been totally reasonable for other YouTubers. Educators do occasionally get grants to educate on specific topics. This is in fact a totally normal thing. Some YouTubers are educators first and foremost, they aren’t there as truth-seekers, but truth-tellers, with a real difference in how careful they need to be about bias.

Some YouTubers are different from other YouTubers. News at 11.

Paying the Truthseekers

Leave a reply

Academics and journalists have a lot in common, at least in principle.

Whether you’re a reporter or a professor, your job is to go out into the world and figure out the truth. You’re supposed to be careful, to check and correct for how you might be wrong. And at the end of the day you’re supposed to communicate what you found.

The differences mostly come in how you’re paid.

You could imagine some sort of pure truthseeker, paid purely by how well they tell the truth. People would ask them to find out the truth about something, and pay them for the service. And the truthseekers with the best track record would get the most clients. But neither profession really works like this.

Journalism comes closest. Once upon a time, people bought newspapers in order to be the first to know when something important happened. While there’s still a little bit of that going on (I guess this is what Bloomberg Terminals are for?), it’s a lot less central because of the internet. Now, there are hundreds of ways to find out about things, from a multitude of news sites to social media. More and more, people expect to be able to get information for free.

In that environment, the news has to compete not on the facts themselves, but on how it presents them. People pay for news that’s curated well to match their interests, or news that feels more respectable. And more than either of those, they pay for news that’s entertaining. So while truthseeking skills pay, writing skills often end up mattering more. In a sense, it’s why it’s possible for me to do journalism at all. I was trained in the academic truthseeking tradition, not the journalistic one. I got into journalism by impressing editors with my writing, not my ability to suss out the truth.

That academic truthseeking tradition is quite different, in part because the rewards for it are much more indirect. Academics pay comes from two main sources: research grants, and student tuition. Students are mostly there to learn old facts, not new ones, so that source of money supports research only in so far that students believe that a successful researcher with time for research will also be a better teacher.

Research grants, in principle, pay for truthseeking. But they’re typically paid by governments, which often don’t have a clear idea of what they’d like to learn, since the more practical questions are already being researched by private companies. So the decision gets delegated out to other academics, who have a vague shared sense of what’s worth knowing and what’s not. Accuracy should have an impact: that is, it should be easier to get grants if you’re better at finding the truth. But in practice, unless someone does so badly they trigger a scandal, academics don’t usually get things all that wrong. So grants are mostly based on other factors.

Paying someone purely to deliver the truth, not to entertain or match a culture, seems tricky. You could imagine sci-fi scenarios. What if we could track the logic people used to make decisions, and demand payment if those decisions were based on facts we uncovered, like a journalist getting a percentage of every short made in response to bad news they dug up about a company? What if governments paid in proportion to how valuable academic ideas turned out to be, centuries after they were discovered, and modern-day academics sold shares in that future payout to fund themselves? What if prediction markets something something?

For the moment, academics and journalists are both in a weird middle space. They’re truthseekers, still, by culture and inclination and desire. But they’re paid for something else.