Claude assists mathematicians in solving long-standing curve research record

Mathematicians have broken a long standing record in number theory with the help of artificial intelligence. As reported by Scientific American, Anthropic researcher Levent Alpöge and cryptographer Ava Howell used an internal version of the Claude language model to discover elliptic curves with higher complexity ranks than any previously known. While advancing to a rank 29 curve took human researchers 18 years, the AI assisted effort unlocked rank 30 and 31 curves in a matter of days.

Elliptic curves are fundamental mathematical structures that play a crucial role in digital security and cryptography. For decades, theoretical mathematicians have struggled to determine the upper limits of these equations. Demonstrating that language models can swiftly surface complex curves suggests AI could become a powerful partner in abstract mathematical research, replacing cumbersome manual calculations.

It remains uncertain whether there is an ultimate limit to elliptic curve ranks, a core question in mathematics that this breakthrough does not answer. Furthermore, because the researchers used an unreleased model paired with specialized prompts, it is unclear how easily other scientists can replicate these results. This raises an open question of whether AI is demonstrating genuine mathematical reasoning or simply acting as an exceptionally fast search engine for human experts.