The short version
- OpenAI states that thousands of AI agents solved a portion of the Navier-Stokes existence and smoothness problem in 88 hours using an unreleased internal model.
- A New York University professor alleges that OpenAI may have accessed information about his concurrent research, prompting accusations regarding the integrity of the timeline.
- The company denies accessing user data for this specific task but acknowledges its models differ from those used by independent researchers.
OpenAI has announced what it describes as a significant milestone in artificial intelligence capabilities: the resolution of a longstanding mathematical problem concerning fluid dynamics. The company claims that by deploying approximately 10,000 autonomous AI agents, it identified a solution to aspects of the Navier-Stokes equations in just 88 hours. This achievement highlights the rapid evolution of machine learning tools, which are increasingly being applied to complex scientific challenges that have resisted human resolution for decades.
The Navier-Stokes existence and smoothness problem is one of seven Millennium Prize Problems established by the Clay Mathematics Institute. These problems represent some of the most difficult unsolved questions in mathematics, with a $1 million reward offered for each correct proof. For ninety years, critical components of these equations, which describe how fluids move and behave, have lacked rigorous mathematical proof. Turbulence, a phenomenon central to these equations, remains poorly understood despite its importance in physics and engineering.
According to OpenAI, the effort began in late August when researchers trained a new internal model that demonstrated exceptional proficiency in mathematics. This model has not been released to the public and is described as significantly more capable than the company's latest commercial offerings. On September 1, citing rumors that other Millennium Prize problems might have been resolved, OpenAI directed thousands of AI bots to attempt solutions for remaining challenges. By September 5, the system had produced a result addressing two of the four statements required by the prize criteria.
The computational cost of this endeavor was substantial. The AI agents exchanged nearly three million messages and generated 130 billion output tokens specifically for the Navier-Stokes task. Based on OpenAI's current pricing structures for its most advanced models, such an operation would have incurred costs of approximately $10 million. Despite the high expense and technical complexity, the company emphasized that its primary goal was to demonstrate progress in AI capabilities rather than to claim the monetary prize associated with the mathematical breakthrough.
The announcement has sparked immediate controversy within the academic community. Tristan Buckmaster, a mathematics professor at New York University, publicly challenged the narrative surrounding OpenAI's timeline. Buckmaster stated that he and Levent Alpöge, a mathematician at Anthropic, had been working independently on similar solutions using OpenAI's Codex tool. He alleged that information regarding their progress was shared with OpenAI prior to the company's public release of its findings.
Buckmaster expressed concern that OpenAI only began its intensive work on the Navier-Stokes equations after becoming aware of his team's efforts. He released excerpts from email exchanges with the company to support his claims, arguing that remaining silent would allow a misleading sequence of announcements to stand. His intervention highlights growing tensions between proprietary AI development and open academic research, particularly when both parties are pursuing the same high-stakes scientific goals.
OpenAI responded by congratulating Buckmaster and Alpöge on their concurrent work, describing it as remarkable. The company firmly denied accessing any user data related to their research for this specific project. However, OpenAI acknowledged that while unlikely, de-identified data from general product usage could have indirectly contributed to model improvements. The firm stressed that its proof differs significantly from the independent researchers' work, both in methodology and in the precise results achieved.
The solution provided by OpenAI has not yet been verified independently or accepted by the Clay Mathematics Institute. Until such validation occurs, the claim remains a preliminary assertion of capability rather than a confirmed mathematical truth. This episode underscores the dual nature of modern AI advancements: they offer unprecedented computational power to tackle ancient problems while simultaneously raising questions about transparency, data privacy, and the attribution of intellectual contributions in an era where human and machine intelligence are increasingly intertwined.
As the debate continues, the broader implications for scientific research become apparent. If AI systems can indeed solve complex mathematical proofs rapidly and accurately, it could accelerate discovery across multiple disciplines. However, the incident also serves as a cautionary tale about the need for clear boundaries between commercial entities and academic institutions. The coming weeks will likely see intense scrutiny of OpenAI's proof, as well as ongoing discussions about how to fairly credit contributions in collaborative yet competitive research environments.
Sources behind this briefing
Go to the original reporting
- BBC Business↗OpenAI says it cracked 90-year-old maths problem in 88 hours
