OpenAI Says Its AI Solved One of Mathematics' Seven Great Unsolved Problems
The company reports that 10,000 AI agents working for 88 hours produced a proof about the equations describing how fluids move — a problem open since the 1930s and worth a million-dollar prize. Mathematicians are checking it, and a dispute over credit has already started.
OpenAI published a claim today that an internal system of its AI agents has produced a proof for the Navier–Stokes problem, one of the seven Millennium Prize Problems — a list of the most important unsolved questions in mathematics, each carrying a million-dollar reward. The company says the work took about 88 hours and involved roughly 10,000 agents working in parallel.
The problem concerns the equations that describe how fluids move — water in a pipe, air over a wing, blood in an artery. The equations are used constantly in engineering, but nobody has been able to prove that their solutions always behave sensibly, or whether they can "blow up": reach an infinite value in a finite time, which would mean the mathematics breaks down even though the real fluid obviously does not. OpenAI's proof reportedly shows that such a blow-up can occur.
Nature covered the claim, which is the right level of seriousness, and so is the caution that accompanies it. A proof of this kind is not true because a company announces it. It becomes accepted when mathematicians who specialize in the area work through it line by line and agree, a process that takes months and occasionally ends with a flaw being found. That verification has just started.
A separate dispute has already overshadowed the result. Questions have been raised about credit — about what human mathematicians contributed, and about the position researchers are in when they use a frontier lab's tools to work on unpublished results. If you run your half-finished discovery through a company's model, whose discovery is it? That question has no settled answer, and it now has a very high-profile test case.
Assume for a moment the proof holds. What it would demonstrate is not that AI is smarter than mathematicians, but that a very large number of capable systems, working in parallel and checking each other for days, can get somewhere that individual humans have not in ninety years. That is a genuinely new way of doing mathematics, and it is worth watching whether it survives review.