DE — heise online
Riemann Hypothesis: Anthropic Mathematician Says He Found a Proof with AI
AI1 min read
An AI researcher at Anthropic says he and its Claude AI model may have found a proof of the Riemann Hypothesis, one of the most famous unsolved problems in maths. A second, non-AI proof attempt is also being checked.
What happened
Levent Alpöge, who works at the AI company Anthropic, says he believes he found a possible proof of the Riemann Hypothesis together with Claude, Anthropic's AI model. He wrote the arguments in Lean, a software tool that checks maths proofs step by step for logical errors. So far there is mainly computer code, not a finished paper. A readable version is promised later.
Key facts
- The Riemann Hypothesis dates from 1859.
- Alpöge says his unpublished text, called the "GRH book", is over 250,000 pages long.
- Alpöge thought the text was too hard for people to read, so he did not publish it.
- Stanford mathematician Ravi Vakil, who is also president of the American Mathematical Society, reported a separate proof attempt by Alessandro Masullo.
- Masullo's work was made without AI, according to Vakil.
What the Riemann Hypothesis is about
Prime numbers, such as 2, 3, 5 and 7, seem to appear in an irregular way. Bernhard Riemann found a hidden order behind this pattern using the zeta function. The hypothesis says that all the important zeros of this function lie on one straight line. If it is true, mathematicians could describe the spread of prime numbers very precisely. Many maths results already depend on it being true.
Why it matters
A proof would make many existing maths results safe, not just guesses. It would also be a big test of whether AI can help solve hard research problems. Still, the claim is not yet accepted, and experts are checking it carefully.
What to watch
- Whether Alpöge publishes a readable explanation of his proof.
- Whether experts find errors in the Lean code.
- Brian Conrad's review of Masullo's work, which he says is being rewritten.
- Whether either proof is accepted by the wider maths community.
Glossary
- Anthropic
- An AI company that makes the Claude AI models.
- Claude
- An AI model made by Anthropic that can read and write text.
- Lean
- A software tool that checks maths proofs step by step to make sure they are logically correct.