GoKawiil25 Fields Medallists warn AI benchmark race is distorting mathematics research
A group of 25 Fields Medal-winning mathematicians has issued a joint declaration warning that AI companies' rush to use famous unsolved math problems as benchmarks for large language models is harming the discipline. They argue that while LLMs have rapidly gained the ability to crack major open problems, this benchmark-driven approach conflicts with how mathematical knowledge has traditionally been built and validated over generations.