OpenAI cracks 1946 math conjecture — and mathematicians fear obsolescence
openai disproved the unit distance conjecture and its new astra model solved ten frontiermath problems, signaling a shift where ai matches phd-level capability and threatens the traditional role of human mathematicians.
openai recently disproved the unit distance conjecture from 1946 using ai, and shortly after, human researchers adapted the technique to solve another conjecture. epoch ai announced a second solution to its frontiermath benchmark, while openai's astra model demonstrated the ability to solve ten complex unsolved problems. experts like abhishek saha note that frontier models perform as well as phd students on routine tasks, turning mathematicians into 'conductors.' while some view this as a golden age of productivity, tim gowers predicts the rise of ai theorem provers will completely revolutionize and potentially displace pure mathematical research.