OpenAI released a statement on Tuesday claiming that an internal, non‑public artificial‑intelligence model has solved the Navier‑Stokes problem, one of the seven Millennium Problems that have challenged mathematicians for more than two centuries. The company reported that the effort required roughly 88 hours of compute time and the coordinated work of as many as 10,000 AI agents operating concurrently.
How the breakthrough was achieved
According to OpenAI, the model accessed a cached version of the internet, ran code, and generated a Lean proof that demonstrates a solution to the fluid‑dynamics equations at the heart of the Navier‑Stokes challenge. The company emphasized that the result is a snapshot of progress rather than a final, definitive proof, and that further peer review will be necessary.
Response from the mathematics community
Prominent mathematicians have expressed both curiosity and caution. Professor Terence Tao of UCLA likened the situation to watching a movie by skipping straight to the ending – the technical result may be correct, but the deeper insight and learning could be lost. Tao warned that rapid AI‑driven competition might turn mathematical research into a “production‑quota” exercise with limited long‑term benefit.
New York University professor Tristan Buckmaster, who was working on the same problem with Anthropic researcher Levent Alpöge, said he had not seen OpenAI’s proof and could not verify the methods used. Buckmaster stressed that he was not accusing anyone of wrongdoing, but noted the timing of OpenAI’s announcement coincided with his own team’s efforts.
OpenAI’s stance on data use and collaboration
OpenAI acknowledged that its work began on September 1 after hearing a rumor about another Millennium Prize problem being solved. The company said it did not access any specific user data from Buckmaster’s or Alpöge’s projects, though it could not entirely rule out that de‑identified data from public usage of its products may have contributed to model improvements.
OpenAI reached out to Buckmaster and Alpöge offering a concurrent release of results and full visibility into the prompts used by its agents. The company also published the Lean proof alongside its announcement, inviting independent verification.
Historical context of the Millennium Problems
The Clay Mathematics Institute identified the seven Millennium Problems in 2000, offering a $1 million prize for a correct solution to each. To date, only the Poincaré Conjecture has been officially solved. The Navier‑Stokes problem, originating from 19th‑century work by Claude‑Louis Navier and George Gabriel Stokes, concerns whether smooth solutions exist for all time under certain fluid‑dynamic conditions.
OpenAI clarified that it does not intend to claim the official Millennium Prize for the result, leaving the prize decision to the Institute after rigorous peer review.
Implications for AI and scientific research
If validated, the breakthrough could signal a new era in which AI systems assist or even lead in solving complex mathematical and scientific challenges. Proponents argue that such tools can accelerate discovery, while critics caution that reliance on opaque AI methods may diminish the educational value of the problem‑solving process.
OpenAI’s announcement adds to a growing list of AI‑driven achievements in fields ranging from protein folding to climate modeling, underscoring the rapid expansion of machine‑learning capabilities across scientific disciplines.
Original reporting: 40/29 / KHBS (NW Arkansas) — read the source article.
