Gary Marcus

A Baltimore native whose father, Philip Marcus, is an MIT alumnus, Marcus skipped his last two years of high school, graduated from Hampshire College in three years, began a doctoral program at MIT at 19 and received his PhD in brain and cognitive sciences at 23; he credits Steven Pinker and Noam Chomsky with forming the foundation of his research.⁠[1]

At New York University he became a professor of psychology and neural science and director of its Center for Language and Music, researching language, computation, artificial intelligence and cognitive development; his books include The Algebraic Mind and Kluge: The Haphazard Evolution of the Human Mind.⁠[2]

Guitar Zero, published in 2012, recounts how, on the eve of his fortieth birthday and with no discernible musical talent, he learned to play the guitar, using the experience to examine how people of any age can master new skills.⁠[3]

With Zoubin Ghahramani, a University of Cambridge professor of information engineering, he founded the artificial intelligence startup Geometric Intelligence, which vowed to surpass the deep learning systems being developed at companies such as Google and Facebook.⁠[4]

On December 5, 2016, Uber agreed to buy Geometric Intelligence, of which he was CEO, using its 15 employees as the core of the new San Francisco–based Uber AI Labs, which he would lead; the startup had been developing machine-learning techniques inspired by cognitive science that rely on less data.⁠[5]

On January 2, 2018, he posted the paper “Deep Learning: A Critical Appraisal” on arXiv, setting out ten concerns about deep learning and arguing that it must be supplemented by other techniques if artificial general intelligence is to be reached.⁠[6]

After the paper appeared, Thomas G. Dietterich, a former AAAI co-chair and NIPS chair, countered it in a series of tweets, arguing that deep learning learns representations and is a new style of programming—“differentiable programming”; Yann LeCun backed Dietterich and suggested Marcus might have mixed up deep learning and supervised learning. The two had debated publicly at NYU in October 2017.⁠[7]

In a February 2019 interview he said he was not saying deep learning is “terrible” but that it is not sufficient in itself and should be one tool in a toolkit; he said LeCun was attacking his credentials rather than his arguments, and described their relationship with the word “frenemy.”⁠[8]

In June 2019 he and iRobot co-founder Rodney BrooksRodney BrooksRodney Brooks is an Australian-born computer scientist and roboticist known for behavior-based robotics. He directed MIT’s Artificial Intelligence Laboratory and CSAIL and co-founded iRobot, Rethink Robotics and Robust.AI.Open the full entry launched the robotics startup Robust.AI to build an industrial-grade cognitive platform for robots; Marcus became CEO and Brooks CTO, and the company raised a seed round from Playground Global and other investors.⁠[9]

In 2019 Pantheon Books published Rebooting AI, written with Ernest Davis. The book argues that AI engineering relies too heavily on deep learning built mostly on statistical models and too little on cognitive processes, and that AI without true understanding of the world cannot reliably drive vehicles or empower robots.⁠[10]

On October 3, 2019, he appeared on Lex FridmanLex FridmanLex Fridman is a Soviet-born, US-raised computer scientist and podcast host who works as an MIT research scientist on human-AI interaction and autonomous driving. His Lex Fridman Podcast is known for hours-long conversations with guests including heads of state and technology leaders.Open the full entry’s podcast in an episode titled “Toward a Hybrid of Deep Learning and Symbolic AI.” The episode description introduced him as an NYU professor emeritus, founder of Robust.AI and Geometric Intelligence, and a critical voice on the limits of deep learning.⁠[11]

In late December 2019 the Montreal Institute for Learning Algorithms (MILA) hosted a two-hour debate between him and Turing Award winner Yoshua BengioYoshua BengioYoshua Bengio is a professor of computer science at Université de Montréal, founder of Mila and a deep learning pioneer who shared the 2018 Turing Award with Geoffrey Hinton and Yann LeCun; since 2023 he has turned to AI safety research, chairing the International AI Safety Report and founding the non-profit LawZero in 2025. He appeared on the Lex Fridman Podcast in 2018.Open the full entry. The two found common ground on hybrid AI systems that combine deep learning with other methods, but differed on terminology and the history of the methodologies.⁠[12]

On March 20, 2022, he wrote in Nautilus that deep learning might be approaching a wall, arguing that scale was not everything, hallucinations were not going away and large language models would not lead to AGI. He recalled that the machine learning community reacted with strong ridicule, including a jab from Sam AltmanSam AltmanSam Altman is an entrepreneur and investor, and the co-founder and chief executive officer of OpenAI. He previously served as president of Y Combinator and founded the location-based social service Loopt. He also helped found Tools for Humanity and Merge Labs, has invested in energy and life sciences ventures, and has driven public discussion of AI governance, employment, and wealth distribution.Open the full entry; in November 2023 he wrote that some industry figures had moved toward his view.⁠[13]

On May 16, 2023, he testified alongside OpenAIOpenAIOpenAI is an artificial intelligence research and product organization founded in 2015 that develops the GPT model series, ChatGPT, Codex and a developer API, and conducts research on image, speech, video and intelligent agents. After its 2025 restructuring, the nonprofit OpenAI Foundation continues to control the for-profit OpenAI Group PBC. Its growth has involved large-scale fundraising and compute partnerships, and has also been accompanied by controversies over governance, copyright, privacy and AI safety.Open the full entry CEO Sam Altman and IBM’s Christina Montgomery before the Senate Judiciary Subcommittee on Privacy, Technology, and the Law, saying the new AI systems would create “persuasive lies” at an unprecedented scale and threaten democracy; he called for independent scientists to conduct clinical-trial-like safety evaluations before wide release and said something like CERN, international and neutral, might be needed.⁠[14]

After the hearing he said current AI is not an existential threat but a serious risk to democracy; he called for an FDA-like process in which large-scale deployment would require trials weighing benefits and risks, and called MetaMetaMeta (Meta Platforms, Inc.) is a publicly listed American technology company, renamed from Facebook in 2021, that operates social communication, digital advertising, virtual reality, smart glasses and artificial intelligence businesses.Open the full entry’s release of LLaMA “a little bit careless.”⁠[15]

In 2024 MIT Press published his Taming Silicon Valley: How We Can Ensure That AI Works for Us.⁠[16]

The book argues that current AI systems were released prematurely in a race for market share and that Big Tech has blocked regulation through political alliances and promises of self-regulation, and calls for public pressure to bring laws based on transparency and accountability, including a boycott of generative AI companies that use creators’ work without consent.⁠[17]

On September 24, 2024, he appeared with Peter NorvigPeter NorvigPeter Norvig is an American computer scientist, former Director of Research at Google and head of the Computational Sciences Division at NASA Ames Research Center, now a Distinguished Education Fellow at Stanford HAI, and co-author with Stuart Russell of the AI textbook Artificial Intelligence: A Modern Approach.Open the full entry at Stanford HAI to explain why current AI is morally and technically inadequate.⁠[18]

X: Public account: https://x.com/GaryMarcus (opens in a new window)⁠[13]

Robust.AI: Official website: https://robust.ai (opens in a new window)⁠[9]