OpenAI announced that their model has solved the Navier-Stokes problem—just to remind you, this is one of the seven unsolved Millennium Prize Problems.
And there's a huge scandal brewing around this. In short, here's the gist:
• OpenAI claimed that their internal AI solved one of the 7 Millennium Problems—the Navier-Stokes problem.
• Then it turns out that mathematicians Tristan Buckmaster and Levent Alpoge (who works at Anthropic 🙂) have been working on this problem with the help of chatgpt and claude for a year, and are now directly approaching its solution.
• According to Buckmaster, he contacted OpenAI and asked—what's going on, guys? Did your model happen to train on our data? To which he didn't get a clear answer, and a representative from OpenAI literally said—okay, bro, you can have the solution, we don't care about that million dollars, we've got plenty of our own, but Levent from Anthropic needs to be excluded from the co-authors—it's not right for someone from our main competitor's team to solve a Millennium Problem. They even threatened damage to his career if he refused to exclude the co-author.
• Meanwhile, OpenAI insists they had no access to the drafts, and they obtained the result independently (yeah, right, at the same time and for the same problem 🙂).
If you happen to be close to solving another Millennium Problem right now—it's better not to upload your solution to OpenAI, or else Sam Altman might announce tomorrow that GPT has made yet another mathematical breakthrough.