AI Keeps Pissing Off Mathematicians

Summary

The video examines OpenAI’s claimed progress on difficult mathematical problems and the controversy surrounding its reported work on the Navier–Stokes Millennium Prize Problem.

Why the problem matters

The Navier–Stokes equations describe how fluids and gases move, and they are used in fields such as engineering and weather prediction. The unresolved mathematical question is whether smooth initial flow must remain smooth, or whether the equations can produce a singularity—such as infinite velocity—in finite time.

The Clay Mathematics Institute offers a $1 million prize for solving each of seven Millennium Problems. The video notes that the Navier–Stokes problem remains officially listed as unsolved.

The mathematicians’ work

NYU mathematician Tristan Buckmaster and his collaborator Levent Alpoge had been using several large language models to develop a proof strategy based on earlier work by Spanish mathematicians Diego Córdoba and Luis Martínez-Zorrilla. Buckmaster says their draft proof was confirmed by August 22, though it did not by itself settle the Millennium Problem.

OpenAI’s reported result

OpenAI says it independently worked on a related problem after hearing rumors of a breakthrough. According to the account presented, its internal system used thousands of AI agents over several days to generate a long proof, which another model then formalized.

The claimed result concerns the possibility of finite-time singularities under a forcing condition. The video stresses that the proof still requires careful review and that OpenAI said it would not claim the prize.

The dispute

Buckmaster argues that OpenAI’s team followed the same unusual approach as his unpublished work, and questions whether his Codex chat sessions may have influenced the company’s result. He also alleges that OpenAI offered a publication arrangement that would exclude Alpoge and made a threatening remark.

OpenAI disputes his account, and mathematician Sebastian Bubeck offered a different description of the discussions. OpenAI’s public explanations about possible data influence reportedly changed over the course of the controversy. The company ultimately said that data submitted after a cutoff date could not have affected the system. The video does not establish that OpenAI used Buckmaster’s chats.

Broader concerns

The story raises questions about privacy, the use of unpublished research, authorship, and whether AI-generated proofs can advance mathematics without preserving the understanding and insight involved. Mathematicians including Terence Tao warned against turning research problems into marketing demonstrations. The Clay Mathematics Institute said it would review the matter.

The narrator argues that AI can be genuinely useful in areas such as programming and higher mathematics while still raising serious concerns about industry practices and trust.

Educational product feature

The video includes a sponsor segment for Brilliant, an interactive math and programming learning platform. Its AI tutor, Koji, is presented as guiding learners with questions and step-by-step hints rather than simply giving answers. The narrator demonstrates using it while working through a Python course. This is a product promotion, not an independent review.

Main speakers and sources

• Dagogo Altraide, narrator and creator of ColdFusion
• Tristan Buckmaster, whose account of the research and dispute is discussed, including through an interview with Brady on Numberphile
• OpenAI and Sebastian Bubeck, represented through public statements and reported comments
• Other cited voices include Terence Tao, the Clay Mathematics Institute, and Melissa Lee of Monash University