OpenAI claims research advances across math, computing and physics

OpenAI says its AI models have produced potentially significant results in mathematics, theoretical computer science and quantum field theory. The company is treating the manuscripts as raw research material requiring independent examination, rather than established scientific breakthroughs.

In mathematics, OpenAI claims a proof of the local smoothing conjecture in every dimension, which it says would also resolve the Kakeya conjecture and Stein’s restriction conjecture across all dimensions. The natural-language argument and its formalization in Lean, a software-based proof-checking system, are publicly available on GitHub. In computer science, the company claims a disproof of the Exponential Time Hypothesis, an assumption underpinning many conditional limits on algorithm efficiency. Researchers including Russell Impagliazzo, Ramamohan Paturi and Ryan Williams are investigating. The result would not automatically settle P versus NP, and practical benefits remain uncertain.

The physics claims involve existence proofs for interacting quantum field theories, including quantum chromodynamics and pure Yang-Mills theory. An independent group led by David Gross will examine the material before deciding how to release it. The Yang-Mills claim does not establish a solution to the complete Millennium Prize problem, which also requires a mass gap.

Academic scrutiny must establish whether the formal statements capture the intended scientific claims and assess their significance. Participating researchers will remain independent and free to publish their conclusions.