Terence Tao, a Fields Medalist, critiques claims that AI models like GPT-4 can solve International Mathematical Olympiad (IMO) problems, highlighting that these models operate under altered conditions, such as rewrites and teamwork, which differ significantly from the actual test environment. He warns against confusing AI’s lab success with true human problem-solving, stressing that the comparison lacks fairness and misrepresents the intrinsic abilities of both AI and students.
