OpenAI Claims Breakthrough in Navier-Stokes Problem

OpenAI Claims Breakthrough on 90-Year-Old Navier-Stokes Equation, But Scrutiny Mounts

OpenAI, the artificial intelligence research lab behind the widely-used ChatGPT chatbot, has announced a potential major advancement: solving the notoriously difficult 90-year-old Navier-Stokes mathematics problem. The company claims its novel AI system, employing a coordination of approximately 10,000 “agents,” achieved this feat in just 88 hours.

The Navier-Stokes equations are fundamental to fluid dynamics, describing the motion of viscous fluid substances. Their complexity has baffled mathematicians for decades, making them one of the seven Millennium Prize Problems, each carrying a $1 million reward for a verifiable solution from the Clay Mathematics Institute.

OpenAI detailed its approach in a recent release, stating that its AI system leveraged a network of “coordinating agents” powered by an internal AI model. These agents were equipped with tools such as internet caching capabilities and the ability to execute code. The problem-solving process involved agents being subdivided into communicating groups, with the ultimate resolution of the Navier-Stokes equations emerging from a collective effort of around 10,000 concurrent agents. The company indicated that the final solution was reached on September 5th, approximately 88 hours after the initial deployment of the agents.

However, OpenAI’s ambitious claim has quickly encountered skepticism from the academic community. Tristan Buckmaster, a mathematics professor at New York University, has voiced concerns regarding the methodology and potential data sources used by OpenAI. In a public statement, Buckmaster revealed that he and Levent Alpöge, a mathematician at OpenAI competitor Anthropic, had been independently working on mathematical challenges, including the Navier-Stokes problem.

Buckmaster explained that he received information suggesting that OpenAI might have been privy to their research progress. He noted a perceived similarity between OpenAI’s claimed approach and their own ongoing work, stating that the path to a solution, as described by OpenAI, does not align with the typical iterative process of tackling such complex problems. Furthermore, Buckmaster raised questions about whether OpenAI’s models had access to or were trained on his and Alpöge’s collaborative sessions within OpenAI’s Codex platform, given their use of large language models in their own research.

“I would like to be clear about what I am not claiming. I have not seen OpenAI’s proof. I do not know what their model did, or how. I do not know whether our data was used,” Buckmaster emphasized.

In response to these concerns, OpenAI acknowledged that its Navier-Stokes initiative commenced on September 1st after “hearing a rumor” about progress on the problem, which they later understood to be related to Alpöge and Buckmaster’s work. The company stated that its researchers and agents did not access any of the mathematicians’ work prior to its public release. While acknowledging that it cannot entirely rule out the possibility that de-identified data derived from their product usage might have contributed to model improvements, OpenAI maintained that no specific user data was accessed for the purpose of solving this problem.

The Clay Mathematics Institute, which established the Millennium Prize Problems in 2000, has yet to issue a formal statement on OpenAI’s purported solution. The initiative aims to highlight the ongoing frontiers of mathematical research and recognize significant contributions. Other unsolved Millennium Prize Problems include the Birch and Swinnerton-Dyer Conjecture and the Riemann Hypothesis, both representing profound challenges in number theory and algebra. The scientific community now awaits a more thorough review of OpenAI’s methodology and the potential validation of its groundbreaking claim.