All news

Anthropic: Claude finds an algorithm that refutes the Exponential Time Hypothesis for 3-SAT

Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .

Two labs have almost simultaneously claimed a result in the foundations of computing theory that was found by SI; now humans have to check it.

Anthropic published the paper "A 2^o(n) Time Algorithm for 3-SAT" with a Lean 4 formalization in its formal-math repository. According to the company, a deterministic algorithm solves 3-SAT in subexponential time, which means the Exponential Time Hypothesis (ETH, the conjecture that nothing faster than exponential time is possible) is false. The algorithm also solves the harder problem of counting satisfying assignments (#3-SAT); the README says the result is purely asymptotic and says nothing about solving formulas in practice. The paper lists Claude as its author and the result was found by an internal research model; according to retellings of the paper, the algorithm was obtained on October 16 with 14 million output tokens and no human input. Anthropic writes that 1,050 Lean modules were written by models and reviewed by other instances of the same models. The publication came shortly after a similar claim by OpenAI. Theorist Lance Fortnow called the approach "promising" but said he cannot yet vouch for the correctness of the work.

History

OpenAI models solve open math problems

8 events
  1. Anthropic: Claude finds an algorithm that refutes the Exponential Time Hypothesis for 3-SATOpen now
  2. OpenAI announces three results from its models: disproof of ETH, local smoothing and Yang–Mills
  3. Mathematicians dispute that OpenAI understands its own potential proof of the Navier–Stokes problem
  4. Experts debate whether SI that solves old math problems could undermine Bitcoin and Ethereum cryptography
  5. Mathematicians doubt OpenAI solved "the" Navier-Stokes problem: the proof rests on an artificially chosen external force
  6. OpenAI model solves more than a hundred long-standing open math problems; release of results to be coordinated by an independent group of nine mathematicians
  7. The Information: OpenAI expects to solve the Hodge conjecture soon, its second "Millennium Prize Problem" in a month
  8. OpenAI's SI agents solve one of the "Millennium Problems" — the $1 million Navier–Stokes problem
Follow on Telegram