Frontier AI labs say their models refuted the Exponential Time Hypothesis. What does that mean?
techcrunch ·
OpenAI has released solutions to three major mathematical problems, including a surprising refutation of the Exponential Time Hypothesis (ETH), a long-standing assumption in computer science regarding computational complexity. Anthropic subsequently released its own independent refutation of ETH, suggesting that AI models can solve certain logic puzzles in sub-exponential time, which challenges previous beliefs about algorithmic efficiency. The scientific community responded with mixed reactions, leading OpenAI to structure its release process differently for mathematicians, computer scientists, and physicists. MIT Professor Ryan Williams, who previously criticized OpenAI's 'proof dump' tactics, agreed to collaborate on interpreting the findings and organizing a community response. This development raises significant questions about the P vs NP problem, potentially altering how researchers view the tractability of hard computational problems. While the new algorithms do not immediately break existing cryptographic systems, they may necessitate revisiting numerous scientific papers that relied on the validity of ETH. Experts are now focused on understanding the implications for future research and raising the bar for human scientific inquiry in an era where AI handles increasingly complex tasks.
techcrunch