OpenAI Claims Navier-Stokes Solution Amid Data Dispute — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI announced a solution to the Navier-Stokes existence problem—one of seven Millennium Prize Problems—using an unreleased model, per the company's blog post and coverage from CNBC, the New York Times, and The Verge.
- The company acknowledged it "cannot rule out that de-identified data derived" from mathematicians Levent Alpöge's and Tristan Buckmaster's use of its products helped improve its models—a caveat OpenAI framed as unlikely.
- Levent Alpöge and Tristan Buckmaster publicly released their own Navier-Stokes work, with OpenAI congratulating them and stating no specific user data was accessed during its own effort to solve the problem.
- An NYU mathematician (per TechCrunch) accused OpenAI of "fighting dirty" on the career-making math problem, while the Clay Institute (per Implicator.ai) declined to call the problem "solved" after OpenAI's proof claim.
- CNBC reported OpenAI claimed to solve the 90-year-old problem in 88 hours, and an OpenAI tweet acknowledged that—like every competitor—the company uses de-identified user feedback and data to improve models holistically.
- Mark Chen of OpenAI clarified on X that no human or agent looked at user data specifically for the Navier-Stokes effort, but confirmed de-identified data is used across ChatGPT and Codex training in general.
Why it matters: OpenAI's claim to crack a Millennium Prize Problem is undercut by its own admission that de-identified data from the mathematicians who solved it first may have helped train its models—an acknowledgment that gives critics a foothold. The Clay Institute's refusal to validate the result and an NYU mathematician's "fought dirty" charge raise the stakes for how OpenAI documents provenance in any landmark research release, especially as it competes with Anthropic on headline-grabbing AI achievements.
Ask SkimNews
