Gary Marcus says LLMs still haven't produced new linguistic generalizations, echoing Chomsky
GaryMarcus · x · 2026-09-08
Gary Marcus recalls his 2023 Web Summit conversation with Noam Chomsky, where Chomsky said he had learned nothing about language from LLMs — and argues that three years later, events have proven Chomsky right.
He quotes linguist Colin Phillips: despite years of impressive and useful language models, requests for concrete examples of new linguistic generalizations credited to LLMs have met mostly silence. The post continues the long-running debate over whether LLMs genuinely understand language or merely pattern-match.
More from AGI Musings
- The Economist: AI Jobs Apocalypse Postponed as an AI Jobs Boom Emerges — daviddisco · 2026-09-08
- Garry Tan echoes investor: AI remains 'extremely early, underhyped and undervalued' — garrytan · 2026-09-08
- Turing laureate Judea Pearl turns 90, honored for twice revolutionizing AI research — zacharylipton · 2026-09-08
- GPT-4 Was Already AGI: Why AGI Technically Means AI That Isn't Hopelessly Brittle — pwlot · 2026-09-08
- Dwarkesh: Don't stop evals or punish models that get caught over HuggingFace incident — ajeya_cotra · 2026-09-08
- Engineer Joshua Saxe: Astra useful but far from AGI hype — a 'slot machine' AI with no memory of intent — joshua_saxe · 2026-09-08