Gary Marcus

@gary-marcus · 10 positions · On ppll ↗

Cognitive scientist, professor emeritus of psychology and neural science at NYU, and the most persistent public critic of large language models. Author of The Algebraic Mind and Rebooting AI; writes the Marcus on AI newsletter.

Gary Marcus did not write this page.

We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.

What Gary Marcus says most about

LLMs 7

Almost every major figure in AI has now arrived at the critique of large language models he began making in 2019.
GPT-5 was an ordinary incremental advance, and the gap between what was promised for it and what shipped is the story of the field.

See all 7

AGI 3

Anyone who thinks large language models are a direct route to transformative AGI is kidding themselves.
A language model is no substitute for a well-specified conventional algorithm, so it cannot simply be dropped into a complex problem and trusted.

See all 3

Apple 2

A language model is no substitute for a well-specified conventional algorithm, so it cannot simply be dropped into a complex problem and trusted.
Success on the easy version of a task is evidence of nothing, because it seduces you into believing the model found a general solution when it did not.

See all 2

OpenAI 2

GPT-5 was an ordinary incremental advance, and the gap between what was promised for it and what shipped is the story of the field.
Scaling the 2019-2023 recipe — same architecture, bigger model, more data — is not enough; further progress depends on new architectural ideas.

See all 2

scaling laws 2

Scaling the 2019-2023 recipe — same architecture, bigger model, more data — is not enough; further progress depends on new architectural ideas.
The LLM line of research will reach a capability plateau.

See all 2