AI-rewritten: This is a summary of an article from New Scientist, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

OpenAI has released 722 mathematical papers proving and disproving various problems, a volume academics describe as both impressive and baffling. While AI models struggled with basic math exams in 2019, OpenAI recently solved complex fluid dynamics equations. The company did not name the specific model used, referring to it only as an "internal frontier model." Francis Johnson from University College London worked on one of these problems for 25 years and produced two books before giving up. He noted he was surprised by the AI’s speed but not its ability, finding the analysis sophisticated and clear.

However, Kevin Buzzard at Imperial College London urges caution regarding the validity of these results. He stated that while 30 papers were relevant to his field of number theory, only seven seemed impressive, with just one formally verified using Lean, a computer analysis tool. Buzzard explained that community acceptance will take time as experts must verify the work or wait for formal proofs. Ben Allanach from the University of Cambridge agreed that results are forthcoming but called the release confusing and disruptive. He questioned whether this output truly enters human knowledge, noting that mathematicians often value the process of solving puzzles over simply checking machine outputs.

The papers were published on GitHub rather than traditional scientific servers, a move Terence Tao from UCLA criticized as potentially harming the field by dumping discoveries without proper context. An independent Advisory Group on Mathematics and Artificial Intelligence was established to advise AI companies on best practices, recommending formalized proofs and transparency about failed attempts. Lindsay McCallum Rémy at OpenAI responded via email, stating they are exploring community-hosted alternatives that meet committee guidelines. The company committed to improving future paper quality through better citations and exposition while giving the mathematical community time to assess the work.