Science
OpenAI Claims to Have Solved Decades-Old Math Problem in 88 Hours
BBC · 2 weeks ago · Read the Full Story on BBC
AI SUMMARY
OpenAI claims to have partially solved a 90-year-old math problem related to fluid dynamics using 10,000 AI agents in 88 hours, amid allegations of using leaked competitor research.
- OpenAI claims it solved a decades-old math problem—part of the Navier-Stokes equations regarding fluid motion—in 88 hours using roughly 10,000 autonomous AI agents.
- The proposed solution addresses the Navier–Stokes existence and smoothness problem, which is central to understanding fluid turbulence.
- The solution resolved two of the four statements required for the Clay Mathematics Institute's $1 million Millennium Prize, though it has yet to be independently verified.
- NYU math professor Tristan Buckmaster accused OpenAI of initiating its Navier-Stokes work only after receiving leaked information about his and Anthropic mathematician Levent Alpöge's progress on September 3.
- OpenAI denied accessing Buckmaster's work or user data prior to its public release, stating that their final proofs and precise results differ significantly.
- To reach the solution, the AI agents exchanged nearly 3 million messages and consumed 130 billion output tokens on the Navier–Stokes problem alone.
- The work was powered by a new, highly capable internal AI model that was trained in late August and has not been publicly released.