“I believe that, in collaboration with Claude, I have found a proof for the Riemann Conjecture,” this statement appeared on Levent Alpöge’s personal website on Friday. He is a mathematician at Anthropic.
Currently, there are two things available for viewing: a formal proof written using the Lean framework, and a detailed document said to consist of over 250,000 pages. The paper in the traditional sense of the term has not yet been published. He also did not specify which version of Claude was used or how much computing power was required to generate it.
The timeline was indeed provided: starting work at the beginning of October and finishing on November 16th, a period of one and a half months. According to him, using the same method, he also managed to prove the Generalized Riemann Conjecture as a byproduct.
The Riemann Hypothesis was proposed in 1859, and its main assertion is that the real parts of all non-trivial zeros of the Riemann zeta function are equal to 1/2. This hypothesis explores to what extent the distribution of prime numbers can be described with precision. In number theory, a large number of mathematical theorems begin with the assumption that the Riemann Hypothesis is true; if this hypothesis is proven correct, all these theorems would no longer require this assumption to hold.
It’s still too early to draw any conclusions at this point.
No one is capable of reading through a document that consists of 250,000 pages in its entirety; therefore, the focus of this verification will be on the use of the Lean method. The Lean approach aims to ensure that “every step in the derivation process is logically valid, based on these definitions” – it does not examine whether these definitions actually correspond to the Riemann Conjecture itself. Whether a mathematical proposition has been correctly formalized can only be determined by manually checking the most fundamental statements at the beginning of the derivation. In the coming weeks, the mathematics community will likely focus on precisely this aspect.
The Clay Mathematics Institute offered a prize of one million dollars for each of the seven Millennium Problems. The rules were clearly stated: any solution had to be published in an established academic journal, wait at least two years after publication, and ultimately be widely accepted by the mathematical community. Among these seven problems, only the Poincaré Conjecture has been solved so far; Perelman, who provided the proof, refused to accept the prize money.
In September this year, OpenAI claimed that its internal model had solved another millennium-old problem: the existence and smoothness of solutions to the Navier-Stokes equations. However, that result is still under verification. Over the course of more than two months, both companies each attributed themselves the credit for solving this problem.
In his statement, Alpöge said that this issue was a “secret” from his student days. His plan at that time was to spend several decades learning about all the tools that might be useful in the future. This year marks the 200th anniversary of Riemann’s birth.