Mathematicians launch First Proof, a first-of-its-kind math exam for AI

A team of leading academics has launched "First Proof," a novel exam consisting of unsolved mathematical problems relevant to current research, designed to challenge and rigorously test the capabilities of Artificial Intelligence systems in pure mathematics. This initiative stems from frustration with the AI industry's unsubstantiated claims regarding mathematical proofs and a desire for transparent, controlled experiments, unlike previous successes on Olympiad problems or existing literature searches. Eleven mathematical experts, including a Fields Medal winner, contributed encrypted proofs for these problems, which are described as "lemmas"—small theorems crucial to larger research but not usually published alone. The goal is not to solve major open problems but to demonstrate AI's near-term potential as a tool to accelerate the tedious aspects of daily mathematical research.

(Source：Scientific American)