I'm Kevin Roose, a tech columnist at the New York Times.
I'm Casey Newton from Platformer, and this is Hard Fork.
This week, SpaceX, Anthropic, and OpenAI are all heading to the public markets, but what do their IPOs mean? Then, author Kevin Hartnett is here to talk about why some mathematicians are sounding the alarm about the use of AI in their field. And finally, some ChatGPT.
Well, Casey, the big news this week is that the AI IPO race is heating up. It's hot IPO summer.
It really does seem like it is going to be a hot IPO summer, Kevin. I am told that we are on track to see what might be the three biggest IPOs of all time. Yes. So, SpaceX is getting ready to go public as soon as maybe next week. And then just this week, Anthropic filed a confidential S-1 with the SEC, noting that it intends to go public. That is sort of the first step in the process. And there are reports that OpenAI is going to file their S-1 soon as well. This is obviously long awaited. People have been wondering when these giant private companies were going to go public and now it seems like they're all racing to do it as quickly as they can and potentially beat each other to market.
Yeah. And the consequences are really important and we're going to get into them. But before we do that, truly there has never been a better or more important time to do our disclosures. Yes. I work for the New York Times, which is suing OpenAI, Microsoft, and Perplexity, and my fiance works at Anthropic. Okay, so those are our extra special AI disclosures this week. Now, Casey, let's talk about these IPOs.
Let's do it. Maybe we should start with SpaceX. What is going on with the SpaceX IPO? What are we expecting and what does it mean?
Yeah, so you know, as you noted, Kevin, SpaceX is just the furthest along right now. They seem like they're getting very close to the finish line and they just have some staggeringly ambitious plans. They plan to sell their shares at $135 a piece which would raise $75 billion. That would make it the largest IPO of all time. It would also value the company at between $1.75 and $2 trillion which would sort of instantly make it among the very biggest companies in the world.
Yeah, those are crazy numbers like that. Those are just numbers that we have not seen before in the history of capitalism. And we should also just remind people that when we say SpaceX, we are talking about sort of the combined Frankenstein.
Yeah. Remind us what is actually in SpaceX.
So they make rockets, they make Starlink, they also as of fairly recently own xAI and X the social network. And so that is all sort of part of this Elon Musk conglomerate that is going public.
Hobby Lobby. Is Hobby Lobby part of it as well?
Not yet, but don't give their biz dev team any ideas.
We're going to do Hobby Lobby in space. So this giant conglomerate is being sort of positioned in the market as a way for people to invest in AI. Obviously, they do have an AI company inside of SpaceX, but I would say investors are also, I would say, more excited about the space part of it, which is the most developed, and they make stuff that people actually use.
Yeah. I mean, look, there are two great businesses in here, right? One is a reusable rocket business that delivers satellites into space. It's very hard to build that kind of company, right? So, like SpaceX just has an incredible moat. There aren't that many competitors to it. We saw Blue Origin, one of its biggest competitors, lose a rocket on the launchpad, you know, just over the past week or so. So that is what makes that an incredible business. And then Starlink is just on fire, right? Like they're using their ability to deliver satellites into space to also create a really powerful global internet access system that is just growing like wildfire. So there are two amazing businesses in there and then they also have two terrible businesses called xAI and X.com. And it'll be really interesting to see what the interplay of the good businesses and the bad businesses are, you know, in the months to come.
Yeah. Sometimes you just got to take the good with the bad and especially when they're all packaged in the same stock ticker.
What's interesting is that you didn't have to take the good with the bad. You could have just had the good. Like up until very recently, SpaceX was just SpaceX and Starlink. Like it was just a pure good business. But then it seemed like Elon Musk decided he needed to kind of, you know, hide his losses somewhere. And so they inherited the two worst companies he owns.
Well, and what's interesting about that is that one of those companies, xAI, appears to be pivoting. So they are now renting out compute that they originally built for themselves to Anthropic, another one of these companies that's going to IPO this year. So they are positioning themselves as something like a space company with a kind of AI neocloud business attached to it and a social network that is going to become the everything app. All of it is a little mysterious but basically this is the long awaited time when all of Elon Musk's sort of Voltron-like companies are going to sort of take a stab at going public together. Okay. So that is the SpaceX IPO. Now let's talk about Anthropic. They have filed their confidential S-1. They are also expected to go public at something over a trillion dollars. That is not what...
Oh, I'm just shaking my head at the insanity of that considering what their ARR was one year ago today.
Yes. I mean, this one is just wild to me. Like I was thinking the other day about the first time I ever visited this company, which was in 2023. So three years ago. And when I tell you that this company was not only ambivalent about making money, but seemed to actively resist the idea of making money. They were like a small group of like very earnest AI safety obsessed people who were like tinkering with and building models for unspecified purposes.
I remember your story about it and it was just kind of about how glum and strange the office was. Which you might expect for a bunch of very, you know, safety focused people who hadn't even decided whether they really wanted to start making a product yet. But boy howdy did they make a product and decide that they actually liked making money and wanted to make a lot of it. And now they are going to be one of the largest IPOs of all time. Just 3 years after I was hanging out with them in their little like Jackson Square walk-up office.
No. And like you know, like in January of 2025 this company has, you know, an annualized revenue run rate of about a billion dollars. And you know, recently they've said it's 50. You know, who knows what it's going to be by the time they IPO, but yeah, I mean, that just is unprecedented growth in Silicon Valley.
Yeah. OpenAI, we don't know anything about their upcoming IPO except that they have said that they plan to do it. They may file as soon as this week their paperwork with the SEC to start that process. People also expect this to be a big gigantic IPO. And so, all of this taken together, I think there are a few threads to pull on here. One of which is what is this going to do to San Francisco. Here we have two, call it two and a half companies because SpaceX has some headquarters and offices in other Texas and Southern California and other places. So two and a half companies going public in the same year based in San Francisco minting, you know, hundreds if not thousands of new millionaires, decamillionaires and centamillionaires and what that will do to the city's tech scene, to the local real estate market, etc. etc. What are your thoughts on that?
I mean my fear Kevin is that we are about to see a massive increase in inequality in a town that already had really significant inequality, right? And I just worry that it's going to sort of feel even worse. And where I'm already starting to notice it is when I talk to my friends who have like really good jobs paying like maybe even mid-six figures, they're looking at what they're reading about the folks who got in early at an OpenAI or an Anthropic and man, the comparison is not feeling good. And they're starting to wonder, what does this mean for me? Am I going to be able to lead the life that I wanted? And it just had me thinking a lot about how when I got to San Francisco in 2010, there was a sense of abundance here. There was a sense of anyone can do a startup and anyone can sort of have the life that they want. And it was true for many, many tens of thousands of people. And I feel like we're almost swinging back to this sort of scarcity mentality here, which is like, well, if you didn't make it in one of these two companies, you know, your future is in doubt. I don't know how true that is, but I can tell you that that is the anxiety that I'm hearing.
Totally. And it's obviously there's some, you know, it's hard to feel much sympathy for these people who are very well-paid engineers and tech workers who are looking at their slightly richer peers and thinking, I got to get some of that. But I think this is a real like status anxiety moment in San Francisco and Silicon Valley where even the people who sort of thought they had sort of made it in the world of tech are now looking at these people who have joined these like, you know, insanely fast growing companies that are sort of wanting to get in on that somehow but also like thinking it might be too late. Like people who I don't think were feeling precarious a year or two ago are now. And I think that's just a really interesting sort of social marker.
Yeah. And like you know, just to say again, like I think what you want in a society is for opportunity to be spread broadly and for everyone to feel like they have a chance to lead the life that they want. And so when you move into a world where there is what essentially amounts to a handful of lottery winners and those are the only people that truly get to live the lives that they want, like that just causes massive, you know, social instability and all sorts of other problems. So yeah, I have like a knot in the pit of my stomach about this.
Yeah, I do too. And one thing that is making it slightly better is that I did see this coming. And I do get to crow about my correct prediction I made on our prediction episode back at the end of last year, which was that 2026 I predicted was the last year to buy a house in San Francisco. I'll make one more quick AI bubble prediction, which is that 2026 is the last year to buy a house in San Francisco because I think this year, this coming year, we will see the IPOs of at least one major AI company. Anthropic is reportedly planning an IPO for next year. We may also see OpenAI go public. SpaceX may go public. Stripe may go public. There's sort of these other sort of very highly valued private companies that may go public. When that happens, their employees get rich and they go buy houses, which now appears to be true. The real estate market is going nuts. Houses are going for, you know, many multiples of their asking price. There was a story this week. Did you see this one about the San Francisco homes that are on sale asking for Anthropic or OpenAI stock instead of cash? I mean, look, I think those people are smart. Like there's a very real chance that that stock will appreciate in value even faster than your house. So I say shoot or shoot.
Yes. So there are two other things about these IPOs that I want to discuss with you. One of them is the effect it's going to have on philanthropy because one of the strange characteristics of these particular companies is that many of their employees, and I would say this especially applies to Anthropic but there's also a piece of this at OpenAI, are committed to effective altruism and other sort of similar philanthropic movements that basically teach you that you should, if you want to make a maximum impact on the world, you should make a bunch of money and then give it away. And so nonprofits, philanthropies, donor advisory networks in and around San Francisco are now starting to ask like what if we just have this insurgence of new philanthropic capital coming from these IPOs. Our mutual friend Nan Ranshoff wrote a great post about this the other day calling it the third wave of philanthropy, basically where you have tens of billions or possibly hundreds of billions of dollars sort of flooding into these charitable movements and causes. How does that affect like what gets funded? Are there going to be new institutions that need to be built to sort of absorb all of this philanthropic capital? I think this is something that people outside San Francisco don't quite understand is like how much money is going to be flowing into these philanthropies over the next couple of years.
Yeah. I also really love Nan's post and had a chance to talk with her about it in person recently and it was just so fascinating to hear about the sheer volume of philanthropic capital that we're expecting to emerge as a result of these IPOs and how little infrastructure there is to absorb it. Something that I think some people might not know about Anthropic is that from their start they have told people as they come on board if you pledge like a percentage of your equity to philanthropy we will actually match it and so this is just a program that is, you know, dramatically amplifying the amount of philanthropic capital that is about to become available.
So well it was not just matching. So there are two things that Anthropic did that sort of speak to their, I don't know, historical ties to effective altruism and that sort of world of philanthropy. One is that all of the co-founders pledged to give at least 80% of their wealth to charity. So right there off the top we are talking about hundreds of billions of dollars potentially that are being earmarked for charity just by the eight co-founders of Anthropic. Then you have the stock matching program which as you alluded to not only offer to match employees who pledged a certain percentage of their stock to charity share for share but match them three to one in the case of some early employees and that just is creating like when you actually lay out the numbers as Nan did in her post like it is just staggering the amount of money like we are going to see something bigger than the Gates Foundation every year potentially for the next few years.
And like it's going to go to some stuff that will seem to the outside world fairly weird, right? Like what...
The joke going around the sort of AI circles is this is going to be a great year for shrimp welfare. Because shrimp welfare, for arcane reasons that are probably not worth going into here, has become like a sort of half-joking pet cause of the effective altruists, but it's a great year to be a shrimp. It's probably also a great year to be working on global health and pandemic prevention, AI safety, all these other sort of cause areas that are very closely affiliated with effective altruism.
Yeah. I mean, it makes me glad that we shredded the social safety net and did all that like sort of reductions in pandemic preparedness. So now the San Francisco billionaires can step in and rebuild it hand by hand.
Exactly. I want to bring up one other thing about these IPOs and ask for your opinion about it, which is that the thing that makes me nervous about these IPOs is not that, you know, they could go sideways or people could lose money or these companies are very speculative. All of that is true. What worries me is the safety angle here because I am of the belief that these AI systems are getting more powerful, that those increased capabilities also bring increased risks. And I just know that many of these AI companies, OpenAI and Anthropic specifically, were started by people who were worried about safety and specifically worried about the ability of a for-profit corporation to develop AI safely. At OpenAI, you've had this sort of governance struggle that resulted in, among other things, Sam Altman being fired and rehired. At Anthropic they have sort of made themselves a public benefit corporation to try to kind of lessen the influence of sort of shareholder capital and fiduciary duty on their ability to make decisions related to safety. But I think all of that just gets much harder in a world where these are publicly traded companies that, you know, big investors, index fund holders that, you know, retirees are invested in. It was already going to be hard to sort of slow down or maybe refuse to release something that was dangerous because of the, you know, enormous sums of money that these companies have raised but it's going to be a lot harder when the public markets are also pressuring these companies to race and go as fast as they can. Does it matter that both OpenAI and Anthropic are structured as public benefit corporations and like sort of have like are allowed to make social commitments that a sort of, you know, traditional like SpaceX for example will not be beholden to?
I'm of a couple minds on this. I think it's probably better that they are public benefit corporations than not because public benefit corporation is sort of this, you know, legal designation that allows you to take into account things like are we being socially responsible? It doesn't, you know, investors can't sue you as easily for sort of breaching your fiduciary duty if you do something that is counter to their interests as shareholders. But they are still corporations and they still exist at the pleasure of shareholders. And there are certain concessions they can make to social issues and impact. But when it comes down to it, like when the rubber meets the road and one of these companies develops a model that is truly dangerous, they are going to now need to weigh not just what do we think the right thing to do is or what do our private investors think we should do or what do our employees think we should do, but they're also going to have the public markets breathing down their neck. They're going to have activist investors and things like that. So, I'm just nervous about the structure that is now going to grow up around these companies and just push them in the direction of acceleration.
I think that's fair and I may be coping, but when I think about the possibility of a lab developing a really dangerous model and saying, well, due to, you know, shareholder pressure, we're just going to put it out there. You have to remember that like the shareholder pressure can work in the other way, too. Because if you put out a model that can create a new bioweapon, you're probably going to get sued for securities fraud by a shareholder that said, I trusted you to only release safe models. So, I do think that there are going to be some positive pressures here that hopefully, you know, keep them from doing anything too silly.
Yeah. How do you think this will impact the average person who's not an investor or a shareholder in these companies who may just, you know, want this stuff to be developed well and safely?
I mean I think on balance you're right that like because these are corporations we do have to worry that like capitalist pressures will lead them to cutting corners and doing things that are unsafe. So that is I think the right concern to have and to keep your eye on it. On the other hand, like I could also see an argument that when your company goes public, you are introducing more democratic oversight and governance into it, right? These folks will now have to like report their earnings. They will have to give us information about their financials. They will have to make certain disclosures as you know their products come out and as their company changes. And you know, shareholders will be able to maybe vote on certain things. So, you know, I think these are all good things because right now we have almost no levers whatsoever that people can pull other than trying to prevent a data center from being built in their backyard. So maybe these give folks some new ones.
Yeah. And I think there's been a lot of hand-wringing over these indexes and these exchanges that have changed their rules like the NASDAQ 100 and the S&P have already or are considering loosening their rules around these so-called seasoning periods. Basically, it used to be if you were a brand new public company that had just IPOed, you could not be included on these major stock indexes because they sort of wanted to see like whether you were stable enough to like become part of the basket of blue chips that people invest in when they buy an index fund. Now partially because of these looming IPOs, those rules have been relaxed so that these companies are not going to have to wait 3 months or 6 months or a year anymore to be included on these exchanges. Some people have said, well, that sounds bad and we're exposing retail investors to these like volatile and risky stocks. I'm not that worried about it. I think like investors want exposure to these companies. I know several people in San Francisco who have been sort of like devising these crazy like hairbrained schemes to get pre-IPO stock in one of these companies. And I think that letting the public benefit from the upside of the AI boom I think is going to do more help than harm. But I could be wrong if all of this goes up in a conflagration and people lose their shirts.
I mean, yes, absolutely. You know, Kevin and I are not financial advisers. You know...
I am actually a certified financial adviser.
I am not a financial adviser. But I do think that if you know, you're a retail investor and you believe in this stuff, you should have the ability to like make that bet because we live in a country where you can bet on Bitcoin in an exchange fund. You can go on a prediction market as a member of the military and bet on an operation that you're a part of. So in a world where those are our restrictions on, you know, sort of financial gambling, if you want to buy a share of OpenAI, I say Godspeed.
Yeah. I think part of the icky feeling that people are having about the AI industry now is that so much wealth is being concentrated in so few private companies and so few hands. And so in an optimistic scenario where these IPOs go off without a hitch and these companies, you know, keep growing at hyperscale, I think maybe having the benefits shared a little more broadly through things like index funds could be good for people's feelings like, oh, there's something for me in this. I'm going to go further and saying it's not just good. I'm going to say it is necessary. Like we cannot have a very small handful of companies that are growing this quickly that are concentrating wealth and power that much into so few hands. It has to be shared more broadly than that. And while an IPO is a very small step in that direction, I do think it is a necessary one. Okay. Well, that is enough about the IPOs for this week. We will continue to cover these IPOs and everything that comes out of them, including possible space data centers. I don't know. I'm excited to learn more about those.
Yeah, I will say when it comes to Starlink, I was not a believer. And then I went on my first Starlink-equipped airplane last week and Casey, this is going to be the biggest company in the world.
When people get a taste of 200-plus megabits in the air on a plane, you're never going back.
Yeah. You can like actually watch YouTube if you start like on your plane.
You can watch so many YouTube videos. And the deal that they struck, I'm told, is they struck the deal where at least with United they're like we'll put Starlink on your plane and you know we will charge United for that but you can't charge your passengers. So everyone's experience of Starlink is it is a free miracle that's being delivered to me in my airplane seat which is not a bad marketing strategy.
Not a bad marketing strategy.
Well, Casey, get out your TI-83 graphing calculator because today we're going to talk about math.
Can I play Snake on it or do we actually have to talk?
No, we have to talk because today we are going to talk about what is going on with AI and math. Now, this is a subject that we have talked about before on this show, but there's actually been a lot happening just over the past couple of weeks. So two weeks ago on May 20th, OpenAI announced that one of their models had reached this big mathematical milestone. Basically it had disproved this long-standing geometry conjecture by identifying a new way of thinking about this famous math problem, one of these Erdős problems that basically no human mathematician had considered before. That was considered a very big deal in the world of mathematics. And at the same time, there's also this backlash brewing in mathematics to the use of AI. Just this week, a group of mathematicians have been passing around and signing something called the Leiden Declaration, which is basically an open letter about the use of AI in mathematics from people who are concerned that maybe they're sort of eroding the human foundations of this sort of academic mathematical discipline.
Yeah. And when I asked a mathematician why, they said y = mx + b.
Okay, very good. That of course is the classic slope-intercept formula for a straight line. Which model did you use to look that up?
So, we just thought it was a really good idea to check in on the state of AI and math. And to help us make sense of what is going on right now, we have turned to one of the best guests I can imagine for this subject. Kevin Hartnett is a journalist. He has covered math and computer science for many publications including most recently as a senior reporter for Quanta magazine. He's also the author of a book that comes out next week called The Proof in the Code which is sort of about this formal math language called Lean and how it's transforming math and AI. Today he works as the editorial lead at Cursor, the AI coding platform and I just thought he would have a really good view of the situation. Also, we just like to bring on people named Kevin because they tend to be really smart. Okay, let's bring in Kevin Hartnett. Kevin Hartnett, welcome to Hard Fork.
Thanks, Kevin. Great to be here.
Well, Kevin, we've brought you here today to talk to you about AI and math. You just wrote a book called The Proof in the Code. Which is, I'll just say it, the most interesting book I've ever read about math.
I won't ask the following question. It narrowly edged out One Fish, Two Fish, Red Fish, Blue Fish.
So when we last checked on the field of mathematics and AI, it was last summer and three of the big AI labs, Google DeepMind, OpenAI, and Harmonic had all reported that their math models had achieved a gold medal score at the International Math Olympiad. That was something that, you know, people had been saying for years would be impossible for or would take many many decades for computers to be able to do but their AI models did it last summer. What has been happening in the field of AI and math since then?
Yeah. I mean virtually everything. So like the IMO had been a benchmark for a long time. In my book there's a whole chapter about the previous year's IMO, the 2024 IMO where Google DeepMind got a silver medal score and that was kind of considered a small watershed. And then as you said last year, three labs got this gold medal level score which had really been the kind of the benchmark that had been set out. At that point AI was still just doing essentially high school math, the hardest high school math in the world, but still just high school math. And I think for people who never went beyond high school math, it's hard to like really appreciate how far that is from the frontier of research math. Like forever far. That's like barely even wading into the field. It's like 0% of the way to the frontier. So it was a proof of concept maybe, but it certainly didn't mean much in terms of can these models actually do research.
That's very hard for me to hear as someone who was not that good at high school math. I have to say like I'm feeling a little defensive, but I get what you're saying. I believe you. It's just making me defensive. Why were the labs so focused on the IMO and on math in general? Was it because that was just like a very hard challenge that they liked or was it because they thought that being able to do math at a high level would enable their models to do other important things?
Well, it's definitely both. I think the challenge, this IMO Grand Challenge which was the name that a researcher at Microsoft Research gave to it, was that was really about like can we just like create models that can do amazing math research. It was just really kind of research for research's sake. At a certain point, the labs and these startups you mentioned adopted that challenge themselves and their motivations were a little different. And there's very much this belief that if you can teach a model to reason about math problems and solve math problems, it will be much better at other things. And I always think about this statement like my high school math teacher would give when people would ask like why are we learning this? What's the point of this? It's like they would either say so you can balance your checkbook or to teach you how to think, right? It was like one or the other. And the to teach you how to think is really the point here. It's like if you can reason through a math problem, think logically, then you can apply that type of skill to like all other parts of your life. And I think the labs believe that if you can teach a model to reason through math problems, it's going to be able to do all these other things that are much more probably like commercially valuable as well.
So I'm having a flashback to when we first started to talk about AI and math and the knock on these models was that they were actually quite terrible at it, right? And that if you would try to get them to do basic addition or multiplication, they would utterly fail. So, Kevin, sketch out for us a little bit what the labs did to sort of navigate through that problem and get to a place where they could kind of credibly try to advance the frontier of the science.
Right. Right. When ChatGPT came out in November 2022, mathematicians were like passing around all these like, haha, look at this stupid model, telling me that there are only finitely many primes when we all know there are infinitely many primes. And like 2 + 2 is 5. Right, basically that kind of thing. I mean I think essentially the models got better. There's definitely an element of reinforcement learning on math problems that make these models better like RL on math but I think it is like just like the general improvement in these models that we all experience kind of in a lot of ways we use them has led to these kind of incredible reasoning tasks they become capable at.
So, let's talk about one of the areas where it seems like we've seen some creativity in math with AI lately, which are these Erdős problems. Kevin, can you tell us who this Erdős was and why he left us with so many problems?
Yes. I mean, so Paul Erdős is a really colorful mathematician. He was essentially the Bob Dylan of math in that he like spent his life on the road. He died at age 83 actually at a math conference. He slept on mathematicians' couches his whole life. And as he went around, he compiled lists of problems that he thought were interesting. Either he would find them in the wild or he invented many of them. And he created this Erdős list. He endowed them with these like tiny little rewards like $20 for solving this problem, $500 for solving this problem. And that fund still exists after his death. And these things are paid out. Actually, I don't know if the LLMs have received the money or who gets the money when they do it. But anyway,
Don't give them the money. They don't need the money. Yeah.
When a human solves them, they get the money. So Paul Erdős collected...
Over a thousand problems, 1,200 problems that he thought were interesting, and he just kind of left them out there. The AI labs and these startups have looked for kind of benchmarks, kind of mountains they can climb, things they can do to prove that their models work well. The IMO was one of the most prominent ones. They got the gold medal there. They needed to move on. They moved on to the Putnam exam, which is the premier college math competition, and started to do quite well there. And then they just started looking for kind of like new targets, and these Erdős problems were sitting out there, 1,200 or so problems that for the most part mathematicians had never looked at or had received very little attention. And so they essentially set their models to work on, you know, all of them, like see what you can do on them, and it would cook up answers to them. And through the beginning part of this year, we would see like on Twitter one announcement after another: solved problem 737, solved Erdős problem 63. And I think mathematicians viewed those very kind of like hypey announcements differently than the energy behind the announcements themselves.
How did mathematicians view them?
They said something isn't adding up here.
Um, I'm gonna have to definitely like think of some puns. I'm going to get one pun out before we finish this podcast. That's my goal.
Yeah. So, mathematicians, um, there are just a lot of unsolved math problems in the world. And just because a math problem is unsolved and has been around for like decades and was dreamed up by a famous mathematician does not make it an important problem. Like an important problem is one the field kind of in its like collective wisdom determines either like the answer to the problem really will change how we view math or more significantly the methods we will need to develop to solve that problem are just going to remake the field. It's going to create important new math. These Erdős problems were just not viewed that way. They're kind of like sophisticated riddles in a way.
Uh, sophisticated arithmetic numerical riddles.
These are like the sudoku of math. It's like the wordle of math. It is like the wordle of math. I think that's a fair statement. And so anyway, mathematicians didn't spend a lot of time looking at them. I think the conventional wisdom was these Erdős problems are like toy problems, not serious problems. But like it's not true about all of them. About a month ago, OpenAI came out with this like big new result. They've solved what many people think of as one of the most important Erdős problems. This thing called the unit distance conjecture. What was important about the unit distance problem is it was a problem that a lot of people had looked at. You could just say no humans have really tried to solve this. Like people have looked at it. They hadn't solved it. The methods underneath it were very sophisticated and surprising. It was not just kind of a clever cobbling together of like obvious techniques. And the result itself was just like so good. Pretty unanimously people agreed this could be published in the Annals of Math, the top journal of math. In a way over the last year, there had been this kind of shifting of goalposts. AI did this but it can't do that. Oh, it did that. Nope, it still can't do this. There's still some shifting going on. There's still room to shift. They haven't solved the Millennium Prize problem. But this result, this proof of the unit distance conjecture really said AI could do absolutely top tier research.
Yeah, I've been following the story of AI and math in part through people like Terence Tao, who is widely considered the greatest living mathematician. Um, he has been sort of experimenting and writing and making, you know, videos about his experiments with these AI programs for use in sort of frontier math research for a number of years now. And when he started sort of working with them, he was like, oh, these aren't that helpful, or maybe they're like a, you know, a sort of mediocre grad student who you'd have assisting you. And more recently he seems to be actually saying like this is revolutionary for the kind of frontier math research that I and other professional mathematicians do. He recently made a video with OpenAI talking about how he can now just try a bunch of sort of crazy ideas and experiments because the sort of cognitive friction of using these models means that you can just sort of have an idea and give it to the model and say go test this a bunch of different ways and figure out if there's anything there. Is that a widely held view among mathematicians or is he just sort of on the extreme AI pill end of the field?
