Dispute erupts over Navier-Stokes solution and AI-generated research volume

A dispute has emerged over a potential solution to the Navier-Stokes Millennium Problem, one of seven problems established by the Clay Mathematics Institute in 2000 with a prize of one million euros. Two mathematicians, one employed by Anthropic and another independent, made progress on the problem using OpenAI's 'Codecs' tool. After the independent mathematician inquired about credit usage, OpenAI reportedly deployed millions of compute instances to advance the specific area of Navier-Stokes where the pair had worked. OpenAI subsequently claimed a decisive step toward solving the problem, though the solution has not yet passed final peer review. The Clay Mathematics Institute is currently reviewing the work to determine the rightful author. Rumors suggest OpenAI may have accessed data collected by the mathematicians, though this remains unconfirmed.

This incident coincides with a broader trend of AI models solving complex mathematical problems. OpenAI's 'Astra' model helped Dr. Danny Hooker solve a theoretical computer science problem in two hours, a task he had worked on for years. OpenAI also reported publishing 700 mathematical works at once. Anthropic has used thousands of agents to formalize Andrew Wiles' proof of Fermat's Last Theorem in the Lean language, a process that took humans a decade to complete manually. The volume of AI-generated output is straining the peer-review system, leading to the use of AI filters for initial screening. Fields Medalist Terence Tao has called for a paradigm shift in mathematics, advocating for 'Mathematics 2.0' in a publication in Nature. Tao argues that the scientific community should place higher value on human understanding and the process of discovery rather than the mere count of published solutions.