陶哲轩在菲尔兹颁奖现场：数学迎来百年新危机
Merged summary
TL;DR - Terence Tao argues that capable AI could shift mathematics from proof scarcity to proof abundance, creating bottlenecks in verification, explanation, peer review, and integration into accepted knowledge. He calls for mathematics to value human understanding and stewardship, not merely proof generation.
- AI solved 7 of 10 research-level First Proof problems at publishable quality under controlled evaluation, reportedly costing $10–$1,000 per problem.
- Tao identifies a five-stage pipeline—generation, verification, exposition, publication, and canonization—whose slower human-led stages may be overwhelmed by AI output.
- He advocates disclosing AI use and holding human authors fully responsible for correctness, citations, and expert-level explanation.
- AI-assisted results should not be published unless their authors can clearly and accurately present and defend them.
Sources (1)
陶哲轩在菲尔兹颁奖现场：数学迎来百年新危机
TL;DR - Terence Tao argues that capable AI could shift mathematics from proof scarcity to proof abundance, creating bottlenecks in verification, explanation, peer review, and integration into accepted knowledge. He calls for mathematics to value human understanding and stewardship, not merely proof generation.
- AI solved 7 of 10 research-level First Proof problems at publishable quality under controlled evaluation, reportedly costing $10–$1,000 per problem.
- Tao identifies a five-stage pipeline—generation, verification, exposition, publication, and canonization—whose slower human-led stages may be overwhelmed by AI output.
- He advocates disclosing AI use and holding human authors fully responsible for correctness, citations, and expert-level explanation.
- AI-assisted results should not be published unless their authors can clearly and accurately present and defend them.