AI Is an Intelligence Multiplier, Not an Independent Mathematician

Summary

John Cook argues that artificial intelligence acts primarily as an intelligence multiplier: it can greatly increase the productivity of highly skilled programmers and help top mathematicians pursue difficult conjectures at extraordinary speed. The benefit is uneven, however, because effective use depends on knowing which ideas to pursue and how to assess the result. Cook says he regularly receives supposed proofs of open problems from amateurs who used ChatGPT, but these proofs are generally invalid. He contrasts those failures with the recent Navier–Stokes proof, arguing that AI did not originate the work or guide it independently; professional mathematicians directed the system using their understanding of the problem and current mathematical approaches. His broader conclusion is that AI amplifies the quality of its users’ input, making “mediocrity in, mediocrity out” a useful update of “garbage in, garbage out.”