Ethan Mollick: AI bot replies systematically understate frontier model performance

emollick · x · 2026-09-13

Wharton professor Ethan Mollick observes that AI bot replies to his posts systematically understate frontier AI performance. In a prior post he described a two-way mismatch: organizations underestimate AI capability growth (often by a lot), while the AI tech world underestimates AI's real-world jaggedness (also often by a lot).

Related event: Wharton's Mollick: Organizations Underestimate AI, AI Circles Underestimate Reality(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →