technology

Alpöge of Anthropic sets argument in Lean proof language

Anthropic mathematician Alpöge says he used the company's AI to prove the Riemann hypothesis, but Terrance Tao says verification is still some way off.

anthropic, riemann, an anthropic, alpöge, named, maths, problems, problems anthropic

Key facts

  • Alpöge, an Anthropic employee, set out the argument in Lean, a computer language used to confirm mathematical proofs. The complete textual proof is…
  • The claim has not been verified. Prominent mathematician Terrance Tao said on Friday that checking the proof is still some way off, the BBC reported.…
  • Named for the mathematician Bernhard Riemann, the hypothesis proposes that prime numbers follow a determined order along the number line.

Alpöge, an Anthropic employee, set out the argument in Lean, a computer language used to confirm mathematical proofs. The complete textual proof is more than 250,000 pages long.

A mathematician at Anthropic said he had used the company's artificial intelligence chatbot to prove the Riemann hypothesis, a nearly 170-year-old problem widely considered the single most famous in mathematics.

The claim has not been verified. Prominent mathematician Terrance Tao said on Friday that checking the proof is still some way off, the BBC reported. Experts are expected to need months, if not longer, to test whether it holds.

Named for the mathematician Bernhard Riemann, the hypothesis proposes that prime numbers follow a determined order along the number line.

Prime numbers are a foundation of higher mathematics. A confirmed breakthrough related to them would have broad impact.

The claim comes as AI firms, including OpenAI, have in recent months announced scores of solutions to major problems in higher mathematics that have eluded scholars for decades.

OpenAI said in September that employees using a non-public version of its tools had solved part of a notoriously difficult problem in 88 hours. The problem was part of the Navier-Stokes equations, which concern how fluids move.

The proof OpenAI released on Navier-Stokes has yet to be confirmed as accurate. It set off criticism in academic and higher-mathematics communities about the use of AI to solve mathematical problems.

A month later, OpenAI released what it said were solutions, proofs or major progress on 372 further outstanding problems in higher mathematics.

More than 10,000 AI agents, computer-programmed bots that operate somewhat autonomously, were put to the task as a training exercise.

Anthropic has also claimed other progress and solutions in mathematics and physics.

Alpöge said the moment chosen to go public was poor."The precise timing is horrible,"he said, referring to Anthropic's public stock listing.

A group of mathematics scholars said AI companies boasting of major solutions to show how sophisticated their commercial products had become was"detrimental to the science of mathematics".

No human was obviously part of major mathematical proofs made earlier this year using AI. Those claims, like the Riemann argument, still await expert review.

Dig deeper

Who is Terrance Tao?

The name refers to Terence Tao, a UCLA mathematician and Fields Medal winner whose early caution signals that the claim is still far from accepted.

What is Anthropic?

Anthropic is a U.S. artificial intelligence company, known for the Claude chatbot, and the employer behind this claimed proof.

What is Lean?

Lean is a proof assistant: software that accepts only arguments written so every step can be checked mechanically, which is why a full formal text can run to hundreds of thousands of pages.

What is the Riemann hypothesis?

Stated by Bernhard Riemann in 1859, it says the non-trivial zeros of the zeta function lie on one critical line, a claim about how primes are spaced, and it is a Clay Millennium Prize problem.

Why does this matter?

A verified proof would be a landmark in number theory, but an unchecked AI-assisted formalization does not yet change the problem's unsolved status.

What happens next?

Experts are expected to spend months or longer testing whether the Lean argument actually proves the hypothesis.

Lire en français