OpenAI fought dirty on career-making math problem, says NYU mathematician
TechCrunch Russell Brandom ● Covered by 7 sources
An NYU professor says OpenAI raced his math team to a proof after learning about their progress. The fight is now over who started first — and whether AI labs are playing clean.
Based on reporting by TechCrunch, Russell Brandom — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
NYU mathematician Tristan Buckmaster says he and collaborator Levent Alpöge have found three proofs tied to a major unsolved problem in theoretical math, and that part alone would have been news. But the louder story is the one hanging off the side of it: Buckmaster says OpenAI learned about their progress, then showed up claiming a full proof of the same problem.
The dispute is over the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize problems. It carries a $1 million prize from the Clay Mathematics Institute, and a solution would be a real step forward in mathematical physics. The equations are used constantly in fluid mechanics, but the theory behind them still has a giant hole in it.
Buckmaster says the tactic he and Alpöge used was not the obvious one. In his account, almost nobody else was working that route, which is why he found it suspicious that OpenAI landed there too. He says his team was still finalizing its results when information about their progress reached OpenAI. After that, he says, OpenAI told them it had already found a full proof, but got evasive when asked when the work began and how much human help was involved.
Then the story gets sharper. Buckmaster says he was told an entire OpenAI team had been working on the problem and that a huge amount of compute had been used. He also says the first prompt was eventually agreed to have been sent only in the last few days, after OpenAI had learned about his work. If that is right, the ugly implication is obvious: a lab with serious compute may have used speed and scale to beat researchers to a result it only chased after seeing someone else’s path.
Sébastien Bubeck, who leads OpenAI’s mathematical research, rejected Buckmaster’s version and called it “false and inflammatory.” He said he followed academic norms and promised a fuller statement later. Buckmaster, meanwhile, says he’s not accusing anyone of anything; he says he’s trying to keep the public record from being rewritten by a chain of announcements that, in his view, would tell the wrong story.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI research that actually matters: not shiny demos, but who gets to claim credit when a model and a lab show up late with more compute. The industry loves preaching openness when it’s convenient and suddenly discovers discretion when a rival’s name is on the paper. If AI is going to help with real math, then the process needs more daylight, not more hallway diplomacy and vague heroics.
Read more about this at: TechCrunch