- Published on
Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof
- Authors
- Name
- DeepHorus
In early September, OpenAI presented a 166-page potential solution to the Navier-Stokes problem. It claimed the proof followed Spanish mathematicians Diego Córdoba and Luis Martínez-Zoroa. Several expert groups now disagree, Scientific American reports. Gonzalo Cao-Labora of EPFL says the similarities are only surface-level.
"When you go to the details, it is a different approach," Cao-Labora said. The proof is written in Lean, a formal verification language that checks every logical step. It does not make the argument readable to humans on its own. The Clay Mathematics Institute has reviewed it since release.
Before publication, NYU mathematician Tristan Buckmaster accused OpenAI of building on work by him and Anthropic's Levent Alpöge. OpenAI denied using Buckmaster's data. Camillo De Lellis of the Institute for Advanced Study says a definitive answer is probably impossible because language models are black boxes.
In early October, OpenAI released roughly 400 more results the math community can barely keep up with. Publishing them as a loose GitHub collection angered many mathematicians. They felt it cheapened their field.
OpenAI deployed about 10,000 model instances for roughly 88 hours and 130 billion tokens. OpenAI researcher Noam Brown says nobody knows what that massive agent swarm contributed. Behavior at that scale has not been studied because comparison tests cost too much.
No one has tested how long a single agent would take to solve the same problem. It is nearly impossible to tell what concrete advantage 10,000 agents offer over 1,000. Brown puts less than 10 percent of the success down to the multi-agent setup.
These systems are new and flashy, Brown says, so they get disproportionate credit. What matters is the base model, which can grind through tasks over long periods and run in parallel. The Clay Mathematics Institute’s review continues, and no definitive answer may be possible.
