Claude AI Failed 650 Times…Then Beat The Human Record

An internal, unreleased version of the Claude AI was prompted to tackle the long-standing Riemann hypothesis. After 650 initial failed attempts, the AI was able to successfully improve a related bound regarding the nontrivial zeros of the Riemann zeta function, surpassing the long-held human record. The experiment demonstrates the growing potential for AI in performing complex mathematical research beyond simply solving provided problems. The researchers note that while this is a significant leap, it does not constitute a proof of the hypothesis itself, but rather highlights how AI can navigate complex problem-solving spaces.