MIT professor leads review of OpenAI Exponential Time claim
MIT’s Ryan Williams is helping interpret OpenAI’s claimed rejection of the Exponential Time Hypothesis, as labs and researchers debate what sub-exponential algorithms would mean.
OpenAI released claimed solutions to three long-standing math problems, including a rejection of the Exponential Time Hypothesis (ETH).
Hours later, Anthropic released its own claimed ETH refutation after one of its mathematicians questioned whether OpenAI’s work drew on earlier research.
MIT’s Ryan Williams is organising the computer-science response with ETH’s 1999 authors, Russell Impagliazzo and Ramamohan Paturi.
The labs claim algorithms that solve certain hard logic puzzles in sub-exponential time—faster than ETH assumed was possible.
OpenAI’s science lead told Williams the lab had checked cryptography implications and was reasonably confident no systems were broken.