A breakthrough in the Navier-Stokes existence and smoothness problem has been shadowed by accusations of unfair tactics as OpenAI reportedly used confidential information and extensive AI compute to publish a full proof shortly after independent mathematicians revealed preliminary results.
- OpenAI used info from NYU’s Buckmaster and Anthropic’s Alpöge to accelerate their proof.
- Effort consumed around $22.5 million worth of AI compute over one week.
- Dispute includes alleged pressure to remove collaborator’s credit and threats to reputation.
What happened
NYU mathematics professor Tristan Buckmaster and Anthropic’s Levent Alpöge collaborated on three new proofs addressing the Navier-Stokes existence and smoothness problem, a famously unresolved Millennium Prize problem with a $1 million bounty. Their work, assisted by AI models including OpenAI's Codex and Anthropic's Claude, marked a significant theoretical advance. However, shortly after announcing their preliminary results, OpenAI published a full proof obtained through an unreleased AI model reportedly designed specifically to tackle difficult unsolved problems.
The publication by OpenAI followed a week-long project that generated approximately 300 billion output tokens, corresponding to an estimated $22.5 million in compute costs. Buckmaster claims that OpenAI’s rapid progress was partly because they gained information about his and Alpöge’s unpublished findings, which influenced their approach. OpenAI confirmed commencing their focused effort after rumors of recent breakthroughs in Millennium Problems surfaced but did not clearly explain the extent of human versus AI input in the process.
Why it matters
The Navier-Stokes existence and smoothness problem is central in fluid mechanics and theoretical physics, with solutions likely to drive foundational advances. The swift, AI-powered proof attempts by OpenAI demonstrate the transformative impact of large-scale AI models on high-level mathematical research, while also raising concerns about research ethics and collaboration practices in academia.
The controversy highlights challenges in transparency and intellectual credit when powerful AI tools and private labs compete in traditionally academic domains. Allegations that OpenAI sought to exclude Alpöge’s credit and discouraged publicizing the dispute underscore the tensions between competitive secrecy and open scientific progress. The episode may influence future AI collaborations and set precedents for how AI-driven discoveries are shared and credited.
What to watch next
The scholarly community and Clay Mathematics Institute will closely scrutinize OpenAI’s proof for validity and originality to determine if it qualifies for the prize. Further clarification is expected on the role of AI in producing the proof and whether the rapid approach following Buckmaster and Alpöge’s findings involved any inappropriate use of confidential data.
Additionally, the evolving dynamics between AI research labs like OpenAI and Anthropic, especially concerning transparency, collaboration, and intellectual property, will be pivotal. How the mathematical and AI communities respond to this case could shape future policies governing AI-assisted research and disputes over credit in groundbreaking work.