Gary Marcus
Professor Emeritus · Author
NYU
Cognitive scientist and NYU professor who is one of the most prominent critical voices on LLM limitations. Founded Geometric Intelligence (acquired by Uber). Author of "Rebooting AI". Argues current LLMs are sophisticated pattern matchers, not genuine reasoners.
Current focus
Jul 23, 2026Gary Marcus argues that AI competition with China is tightening, warns of investment risks and oversights, and compares hyperscalers to struggling airlines, while suggesting some jobs may adapt to AI.
Do you agree with this position?
AI-distilled summary of recent news coverage — see sources above.
In the News
Gary Marcus says AI fatigue could hit coders but other jobs may be spared — and even become more fun
MSN · July 22, 2026
China has all but caught up. The US is not going to “win” the AI war. Here’s what we should do instead.
Marcus on AI | Substack · July 20, 2026
Gary Marcus on AI investment risks and oversight
MSN · July 6, 2026
Hyperscalers could end up resembling airlines—plagued by small margins, intense competition, and high expenses, AI skeptic warns
Yahoo Finance · June 29, 2026
China catches up
Marcus on AI | Substack · June 28, 2026
Core Positions & Ideas
The Algebraic Mind — Humans Use Rules, Not Just Statistics
2001In 'The Algebraic Mind' (2001), Marcus argued that human cognition relies on symbolic rules that cannot be captured by pure connectionist (neural network) models. This set up a decades-long debate with the deep learning community that the rise of LLMs has made more, not less, urgent.
Your take on this position:
Deep Learning Is Hitting a Wall — We Need Hybrid AI
2019Co-authored 'Rebooting AI' (2019) arguing that deep learning, despite impressive results, lacks the reliability, common sense, and causal reasoning needed for real-world AI applications. Proposed hybrid architectures combining neural networks with symbolic reasoning as the path forward.
Your take on this position:
LLMs Are Sophisticated Pattern-Matchers, Not Genuine Reasoners
2022One of the most persistent and documented critics of LLM capabilities (2022–present). Collects and publicizes systematic failure cases: reasoning errors, hallucination, compositional failures, inconsistency. His argument: impressive performance on benchmarks hides fundamental inability to reason reliably about novel situations.
Your take on this position:
AI Companies Are Systematically Overhyping Capabilities and Understating Risks
2023Documents gaps between AI companies' public claims and actual system performance. Argues that AI hype creates dangerous deployment of unreliable systems in high-stakes domains (medicine, law, autonomous vehicles). His blog and substack provide ongoing documentation of AI failures that companies downplay.
Your take on this position:
Essential Reading & Watching
Rebooting AI: Building Artificial Intelligence We Can Trust
Co-authored with Ernest Davis. A systematic critique of deep learning's limitations and a proposal for what genuinely reliable AI would require. Prescient on many of the reliability problems that emerged with early LLM deployments.
Read / Watch →
Deep Learning: A Critical Appraisal
A detailed academic critique of deep learning's limitations across ten dimensions including interpretability, causal reasoning, compositionality, and sample efficiency. One of the most cited critical papers on deep learning.
Read / Watch →
Gary Marcus's Substack (The Road to AI We Can Trust)
Regular essays documenting LLM failures, critiquing AI hype, and arguing for what genuinely trustworthy AI would require. Essential reading for a critical perspective on AI progress claims.
Read / Watch →
Recent Writing (8)
Breaking: Demis Hassabis endorses preflight safety testing for AI
Good news, for once.
July 14, 2026Read original →
Off for adventures
Leaving you with a couple laughs
June 29, 2026Read original →
China catches up
Has the US been focused on the wrong things?
June 28, 2026Read original →
The month Generative AI lost its mojo
June is not over, and anything could still happen, but a lot already has.
June 26, 2026Read original →
The Generative AI Fizzle™
Disclaimer: Anything can happen at anytime in the market; I don’t give stock picks, and as the saying goes, the market can remain irrational longer than you can remain solvent.
June 25, 2026Read original →
Accenture: Then and now, and how it may signify things to come
Blip, or one more data point that is on trend?
June 18, 2026Read original →
Breaking: Trump asks the impossible of Anthropic
Where do we go from here?
June 17, 2026Read original →
OpenAI’s lead is dwindling fast
As James Carville might have said, “It’s the lack of a moat, stupid”
June 16, 2026Read original →
