Skeptical take: OpenAI can't train large models, pivots to RL and inference

teortaxesTex · x · 2026-09-04

Quoting LinQingV's observation that AI scaling is shifting from storing intelligence in ever-larger models to extracting more from fewer parameters via compute-intensive inference, user teortaxesTex offers a pessimistic read: OpenAI simply cannot train a large model right now, gave up, and doubled down on RL — yielding mediocre knowledge scores but a leap in intelligence.

An unverified but widely relatable take in the ongoing scaling-bottleneck debate.

Related event: Debate Swirls as OpenAI Shifts to RL and Inference; Netizens Mock AGI Hype(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →