The mystery of prime numbers is revealed: mathematicians have confirmed the proof of the Riemann hypothesis found by AI Claude

Источник изображения: Erwan Hesry/Unsplash

Anthropic recently announced that its AI model of the Claude family helped solve one of the most famous unsolved problems in mathematics, which had not been proven since 1859, the Riemann hypothesis. Independent mathematicians checked the proof obtained and practically eliminated the possibility of error.

  Image Source: Erwan Hesry/Unsplash

Image Source: Erwan Hesry/Unsplash

The breakthrough was reported by Anthropic mathematician Levent Alpöge, who worked on the hypothesis with Claude. It was formulated by the German mathematician Bernhard Riemann in 1859 and concerns the patterns of distribution of prime numbers. Riemann proposed that all non-trivial zeros of the zeta function describing the distribution of prime numbers lie on the same vertical line of the complex plane. Confirmation of this hypothesis is considered one of the most important problems of modern mathematics, for the solution of which the Clay Institute of Mathematics awarded a $1 million prize.

According to Alpöge, Claude played an important role in finding the proof and was particularly helpful in the final stages of the work. The mathematician admitted that he expected to approach this problem much later, but the capabilities of AI made it possible to speed up the research. The result turned out to be so complex that instead of the usual scientific work with a detailed presentation of mathematical calculations, the researcher presented it in the form of program code in the Lean language, which allows checking mathematical proofs.

Independent mathematicians, including MIT specialist Andrew Sutherland, checked the proof presented and formally confirmed its correctness by independently compiling Alpöge’s result into Lean. Sutherland said the likelihood of finding anything questionable in the work is virtually zero.

“The Lean proof was compiled and independently verified by the Lean core using only the three standard axioms. Frankly, the likelihood that there is anything wrong here is almost zero,” Sutherland said.

Additional testing was carried out by members of the Lean and Mathlib communities, which are involved in the development of the language of the same name for the formal verification of mathematical proofs and its library. In a joint statement, they said that they checked not only the correctness of the software implementation, but also the consistency of the statement in the code with the original formulation of the Riemann hypothesis. This eliminated the possibility that the program was only proving a similar but mathematically different statement.

“We have full confidence in the correctness of Alpöge’s proof of the Riemann hypothesis and congratulate him on this achievement,” Lean and Mathlib representatives said in a statement. They separately confirmed the correctness of both the formulation of the theorem itself and its proof.

Thus, the verification was successful: independent researchers were able to reproduce the compilation of the proof, and additional tools confirmed that the code does not replace the original mathematical statements or use workarounds.

If you notice an error, select it with the mouse and press CTRL+ENTER.