Delip Rao: scaling was necessary but not sufficient, OpenAI got lucky

deliprao · x · 2026-09-07

AI researcher Delip Rao argues scaling was a necessary but by no means sufficient condition for today's AI. OpenAI's faith in scaling laws was misplaced, and the field got lucky because later approaches—many invented outside OpenAI, like function calling and ReAct—worked synergistically with blind scaling.

Without those inventions, Kaplan's "scaling will solve everything" thesis would have hit a ceiling quickly, far short of current reality, Rao says.

Related event: Brundage and Delip Rao Clash Over AI Misuse Predictions and the Merits of Scaling(5 posts)→

Original post →

More from AGI Musings

AGI Musings channel →