OpenAI and Anthropic refute Exponential Time Hypothesis

Based on reporting by TechCrunch.

What happened

OpenAI released solutions to three math problems, including a refutation of the Exponential Time Hypothesis.

Anthropic also released its own refutation of the Exponential Time Hypothesis.

MIT Professor Ryan Williams is organizing the computer‑science community’s response to the new finding.

OpenAI’s models claim to have algorithms that solve ETH‑related puzzles in sub‑exponential time.

OpenAI checked whether the new algorithms affect cryptographic systems and was reasonably confident they do not break any.

AIBeat analysis

This could prompt researchers to revisit complexity classifications that rely on the Exponential Time Hypothesis and suggests that, despite the new algorithms, existing cryptographic protocols may remain secure, though further peer review will be needed to confirm the assessment.