Overview
- At a March 10 UCLA IPAM event, Fields Medalist Terence Tao and OpenAI executive Mark Chen discussed recent advances that elevate AI from weak assistant to credible collaborator in mathematics.
- Chen reported longer autonomous working time and sharply fewer hallucinations, enabling dependable code generation, literature search, and verification of lemmas that mathematicians often skip writing out.
- Tao said AI is shifting math research toward an industrial model with defined roles, offloading computation and verification to machines and opening the door to larger, coordinated projects.
- Recent benchmark results, including International Mathematical Olympiad–level performance, were cited as evidence of the capability jump from Tao’s earlier description of AI as a very inefficient graduate student.
- In Tao’s classes, homework scores have risen while closed-book exam results have fallen, leading him to predict that routine weekly assignments will be the first casualty of widespread AI use.