a16z Video Watch posted an update
What Today’s Best Models Still Can’t Do in Math
Why it mattersDaniel Litt examines AI in mathematics through the Erdős unit distance result and models’ ability to calculate, combine techniques, and search large spaces. He argues that frontier models still struggle with intuition, theory building, and choosing fruitful questions, while considering how researchers can use them without weakening mathematical craft.
Discuss: If AI in mathematics can generate proofs but struggles with intuition and choosing fruitful questions, how should researchers use frontier models without outsourcing the development of mathematical taste?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.