OpenAI and Anthropic say their algorithms challenge the Exponential Time Hypothesis, prompting computer scientists to organize a review of the claims.
OpenAI and Anthropic have each presented algorithms they say challenge the Exponential Time Hypothesis, a foundational assumption about how quickly certain difficult problems can be solved. OpenAI released solutions addressing three major mathematical and scientific problems; Anthropic later published its own proposed refutation of the hypothesis. The claims remain subjects for scrutiny, not an established change to computer science.
The Exponential Time Hypothesis, or ETH, concerns a class of logic problems for which the fastest possible algorithms are generally expected to require exponentially more time as the number of variables grows. The new work argues that these problems can instead be solved in sub-exponential time. If the proofs hold, researchers may need to revisit results that rely on ETH. That would not, by itself, prove that every problem whose answer can be checked quickly can also be solved quickly—the broader P-versus-NP question remains unresolved.
MIT computer scientist Ryan Williams is helping coordinate a response from the field, alongside UC San Diego researchers Russell Impagliazzo and Ramamohan Paturi, who first published ETH in 1999. A working group is seeking public feedback and discussion of standards for AI contributions. Williams told TechCrunch that parts of the work seemed unusual enough to merit wider examination, while emphasizing that further analysis is needed.
There is also a security question: stronger algorithms could matter to cryptographic systems if they made certain hard problems easier to solve. Williams said OpenAI science lead Dan Roberts told him the lab had checked for such effects and was reasonably confident the result did not imply that any systems were broken.
The claims could affect published research that assumes ETH, but computer scientists first need to evaluate the proofs and understand their implications. The field’s organized review is intended to help determine what the results establish and what, if anything, needs to change.