🔬Tao Criticizes OpenAI's 719 AI-Written Math Manuscripts
TL;DR
Mathematicians say OpenAI's batch release of 719 AI-generated manuscripts falls short of new community standards. Only 10 include the model's chain of thought, and 42% of proofs were not formalized.
Mathematicians say OpenAI's batch release of 719 AI-generated manuscripts falls short of new community standards. Only 10 include the model's chain of thought, and 42% of proofs were not formalized.
Key Points
The IAS-hosted Advisory Group on Mathematics and AI (nine members) published guidelines at the end of September
Terence Tao said problems are being solved by "AI prompters who have no interest in the broader field itself"
A Cambridge and King's College London paper found at least two mismatches between a Navier-Stokes proof's text and its Lean code
The group's first recommendation: stop testing advanced math problems on proprietary models
Why It Matters
A Lean proof only helps if the formal statement matches the claim, and model-written formalizations can drift. Without human understanding at release, nobody can tell a result from a plausible-looking one.
Quick Facts
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,567 builders reading daily.