OpenAI and Anthropic Claim to Disprove ETH — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI and Anthropic published GitHub claims, in natural language and machine-checkable Lean format, that ETH is false; researchers have not independently verified the proofs or had time to understand them.
- OpenAI asked the UC Santa Barbara scientists who proposed ETH and an MIT academic who refuted a version to rally researchers; the effort includes a Google volunteer form and a conference OpenAI has committed to fund.
- Peter Jonsson said his work due at the IEEE Symposium on Foundations of Computer Science in two weeks is “dead” after the claimed disproof of the 27-year-old hypothesis underpinning it, and expects many talks to begin with “ETH is false, half of my paper does not work.”
- Daniel Lokshtanov estimated that, if confirmed, the result would leave about 50 of his published papers “pointless” or in need of revision, calling it a “giant bomb dropped into the middle of the field.”
- Thore Husfeldt said most of this year’s work on one problem is now irrelevant, while weekly or biweekly AI announcements have left researchers depressed enough that his department considered hiring an occupational psychologist.
- OpenAI published 722 manuscripts on 400 problems on GitHub earlier this month, prompting complaints about volume and errors; academics said its roughly 60-page ETH paper was better written than previous batches.
- Victor Lagerkvist said he plans to return to problems he had given up under ETH if the claims hold; Daniel Lokshtanov said the result does not directly affect cryptography but should scare cryptosystem designers because web security assumes some mathematical problems take an impractically long time to solve.
Why it matters: If the unverified ETH claims hold, researchers face the revision or abandonment of dozens of papers, including about 50 by Daniel Lokshtanov. Lokshtanov said the result does not directly affect cryptography, but should scare cryptosystem designers, making independent review consequential for both academic theory and web security.
Ask SkimNews
