The Navier–Stokes Millennium Prize Problem
20 days ago
- OpenAI used an unreleased model to solve the Navier–Stokes Millennium Prize problem, a $1,000,000 challenge.
- The discovery faces accusations of skulduggery from NYU professor Tristan Buckmaster, who collaborated with Anthropic employee Levent Alpöge on a similar solution using Claude and Codex.
- Buckmaster claims OpenAI may have accessed their work after hearing rumors, though OpenAI denies accessing user data directly.
- OpenAI's effort began on September 1 after hearing rumors, using 130 billion tokens for Navier–Stokes and 300 billion total across multiple problems.
- The situation parallels computer security: merely knowing an unpublished solution exists can trigger massive LLM spending to replicate it.
- The incident raises concerns about how AI labs use user data to improve models, potentially allowing competitors to benefit from private work.