OpenAI claims AI model disproves long-standing Exponential Time Hypothesis
OpenAI announced that its AI system has produced a formal disproof of the Exponential Time Hypothesis and also reported breakthroughs in mathematics and physics, pending independent verification.
OpenAI said its artificial-intelligence model generated a formal disproof of the Exponential Time Hypothesis (ETH), a conjecture that underlies many complexity-theory results in computer science. The company posted a write-up and a Lean-verified proof on GitHub after the model produced the result on 16 October and the proof was checked the following day.
A team of researchers led by Russell Impagliazzo and Ramamohan Paturi of UC San Diego and Ryan Williams of MIT will review the claim. In an open letter they clarified that their involvement does not constitute endorsement of the result. > “Our involvement should not be taken as certification of the claim. A result of this significance requires careful, independent scrutiny,” the three wrote.
Williams added on X that the reviewers have no non-disclosure agreements and will not be compensated. He said OpenAI agreed to release all raw data related to the ETH finding and to fund a conference for the research community.
In parallel, OpenAI reported that its models produced proofs for several longstanding problems in mathematics, including the local smoothing conjecture in all dimensions, which would also resolve the Kakeya and Stein restriction conjectures. The company also claimed a rigorous proof that certain four-dimensional quantum field theories, such as quantum chromodynamics and pure Yang-Mills theory, exist. Nobel laureate David Gross will lead the physics verification team, which has opted to write up the findings themselves rather than release the AI-generated material immediately.
OpenAI stated it does not seek academic credit for the discoveries and has allocated $5 million to the Simons Foundation to support workshops on AI-produced results. The computer-science researchers warned that credit allocation could favor those who publish first or have the most powerful tools, echoing concerns raised in the recent Leiden declaration.