Margaret Mitchell: Assuming LLMs Should Be the Basis of AGI Is Worth Questioning

mmitchell_ai · x · 2026-09-22

AI researcher Margaret Mitchell argues that the assumption that LLMs should underpin AI/AGI deserves scrutiny: LLMs are inherently stochastic, making them harder to predict and control—a critical concern from a risk/harm standpoint, including x-risk. She notes she's far from alone, pointing to Fei-Fei Li, Yann LeCun, and longtime critic Gary Marcus as established voices questioning LLMs as the foundation of modern AI—a notable dissent to the dominant scaling-toward-AGI narrative.

Related event: Researcher Margaret Mitchell questions betting AGI on LLMs(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →