Full Breakdown
OpenAI Unveils 372 AI-Generated Math Results, Sparking Debate
By Drooid · · How we work
New Wave of AI-Driven Proofs
OpenAI announced the release of 372 results from its latest internal model, each addressing a major open question in mathematics or theoretical computer science. The results were posted to a public GitHub repository at 6 P.M. EDT. The claimed breakthroughs include a solution to the four-dimensional Kakeya conjecture, algorithmic improvements, and progress toward the Riemann hypothesis. The company says the bulk of the work emerged from a single prompt given to a single AI agent, contrasting with the earlier Navier-Stokes solution that relied on a 10,000-agent swarm costing millions of dollars in compute.
Background and Recent Advisory Move
The release follows OpenAI’s earlier high-profile resolution of a Navier-Stokes problem, which generated controversy over transparency. On September 21, the firm announced the formation of an independent advisory group of mathematicians tasked with recommending responsible publishing practices for AI-generated results. The group’s guidance calls for public disclosure of the model, exact prompt, and compute time for each proof.
Official Statements & Responses
It also emphasized that the flood of proofs will require months of community effort to assess novelty and significance.
Criticism & Opposition
Daniel Litt of the University of Toronto argued that withholding the proofs keeps them secret from the community and that openness would benefit mathematics. Renowned mathematician Terence Tao has publicly criticized the “insane” speed of AI-generated results, warning that even the labs themselves struggle to understand many of the new proofs.
Verbatim Quotes
- “Until and unless they release the model and people can replicate their results, I think you should treat any claims about one-shotting problems with a single agent as unverified,” — Andrew Sutherland, a mathematician at the Massachusetts Institute of Technology
- Daniel Litt, a mathematician at the University of Toronto, argued that the company should not keep the answers secret.