OpenAI refutes Exponential Time Hypothesis, sparks debate in computational complexity

OpenAI released solutions to three mathematics problems, one of which refuted the Exponential Time Hypothesis (ETH), a long-standing assumption in computer science. Anthropic subsequently released its own refutation of ETH. OpenAI structured its release across three disciplines: mathematicians received proofs on GitHub, computer scientists formed a working group to request public sharing and a conference, and physicists formed a working group to determine release norms. MIT Professor Ryan Williams, who has previously criticized OpenAI's proof methods, agreed to collaborate to help interpret the solutions.