Wes Roth
@WesRoth · 20 juil. 2025
Google DeepMind just dropped a major milestone in AI reasoning.
AlphaGeometry, trained entirely without human demonstrations, solved 25 out of 30 Olympiad-level geometry problems, outperforming all previous methods and approaching the level of an IMO gold medalist.
No
Elon Musk
@elonmusk
The AlphaZero lesson
16:58 · 20 juillet 2025 · 89,3 k vues
122
35
727