Brief
Mathematicians weigh OpenAI’s flood of AI‑generated proofs
OpenAI’s recent release of over 370 AI‑generated proofs has sparked both excitement and criticism among mathematicians.
By Felo News Desk · Published
OpenAI dropped a repository of more than 370 AI‑generated results on 372 open math and computer‑science problems, prompting a mixed response from the research community, Scientific American reported on Oct. 8.
What happened
The company made the archive public on Oct. 6, saying a single prompt to an internal model produced the proofs. Within two days, mathematicians began reviewing the papers, noting that about 42 % are written in the Lean formal‑proof language, which gives confidence in their logical soundness despite often poor English explanations.
What the reports add
Harvard mathematician Michael Douglas called the moment “the day when a new era of math and mathematical physics began,” comparing it to the earlier Navier‑Stokes claim. Hector Pasten of the Pontifical Catholic University of Chile said the new papers improve citation practices, while Roland Bauerschmidt of New York University described many as “AI slop” that would be discarded if received from an unknown sender. Andrew Sutherland of MIT warned that three papers have already been withdrawn after serious errors were found, and he expects further mistakes among the unverified proofs.
What was said
“Yesterday will be the day when a new era of math and mathematical physics began,” Douglas told Scientific American. Pasten added, “I think they are doing a better job of giving credit.” Bauerschmidt said, “If I had received these by e‑mail from a nobody, I probably would have deleted it, it’s so badly written.” Sutherland noted, “While I expect most of these will ultimately be found to be correct, or at least correctable, we should expect some of the proofs to contain mistakes, possibly serious ones.”
How it came about
OpenAI’s release follows earlier controversy over its claim to solve the Navier‑Stokes problem, which drew criticism for lack of peer review. The company has previously announced collaborations with the Institute for Advanced Study to involve mathematicians in future work. This latest batch reflects the model’s ability to generate formal proofs, but the community stresses that months of verification will be required to assess novelty and correctness.
Key facts
- OpenAI released over 370 AI‑generated math and computer‑science results on Oct. 6, 2026. (scientificamerican.com)
- Around 42 % of the results are coded in the Lean formal‑proof language. (scientificamerican.com)
- Three papers have already been withdrawn after experts found significant errors. (scientificamerican.com)
- Harvard mathematician Michael Douglas called the release a new era for math and physics. (scientificamerican.com)
- Mathematicians report uneven quality, with some papers readable and others described as "AI slop." (scientificamerican.com)
Sources
- [1] scientificamerican.com — originally reported as “Mathematicians marvel, and grumble, at OpenAI’s trove of new results”
