OpenAI Publishes Three AI-Generated Basic Science Research Findings, Invites Academic Community to Verify

On October 26, OpenAI announced that its internal models had generated results in three areas of fundamental scientific research and attempted to subject them to academic review in different ways. In mathematics, the company stated that the model proved the local smoothing conjecture in arbitrary dimensions; if the argument holds, it would also involve the Kakeya conjecture and the Stein restriction conjecture. Both the natural-language draft and the Lean formalized proof of this result have been made public. In computer science, OpenAI stated that the model refuted the "exponential time hypothesis," which is widely used to derive computational hardness, and released supporting evidence. A team led by experts including Russell Impagliazzo, one of the hypothesis's proposers, will organize public verification and interpretation. This claim does not equate to resolving the "P versus NP" problem. In physics, the company stated that the model provided rigorous existence proofs for theories such as four-dimensional quantum chromodynamics, including an argument for the existence of pure Yang-Mills theory. A team led by David Gross will first study and organize the materials; currently, the AI-generated physics manuscripts have not been made public. OpenAI noted that these drafts should be regarded more as raw research materials awaiting transformation into scientific understanding. External participants are working independently and without compensation, and the company is not seeking academic credit for the related findings.