AI Summary
5 min read“Every time you use their product, they lose money.” That is the blunt arithmetic Gary Marcus offers for the generative AI industry. The cognitive scientist and longtime AI skeptic sat down with Prof G Markets to argue that the technology investors are betting trillions on is fundamentally unreliable, its business model is unproven, and the industry has locked itself into a single, flawed approach. Marcus is no Luddite—he founded a machine-learning company acquired by Uber and has spent decades in AI research. But he insists the current path is dangerous, both economically and socially.
The Core Flaw: Next-Token Prediction Is Not Intelligence
Marcus’s central technical argument is that large language models (LLMs) are “basically next token predictors.” They are trained to guess the next word in a sequence, not to understand the world. “That’s an interesting thing to do,” he says. “It’s part of what humans do. But it’s not all of what cognition is.” The result is a system that can mimic human conversation convincingly but has no stable model of reality. When pushed outside its training data, it “will do really stupid things.”
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Prof G Markets
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (03:19) **Gary Marcus: The Skeptic Who Works Inside AI** - Marcus, an NYU professor and founder of a machine-learning company acquired by Uber, explains why he's spent years warning about the technology's limitations and dangers.
- 2 (06:33) **Why Generative AI Is "Inherently Unreliable"** - Marcus explains the fundamental technical flaw: LLMs are just next-token predictors, not reasoning systems.
- 3 (10:18) **The "Over-Attribution of Intelligence" and the Trillion-Dollar Bet** - Humans are evolutionarily ill-equipped to judge machine intelligence, leading to massive over-investment.
- 4 (12:09) **The "Mountain Peak" Metaphor and the Innovation Monoculture** - The industry is stuck on one peak (LLMs) and needs to go back down the valley to find a better path.
- 5 (14:08) **The "No Moats" Regime and Price Wars** - Marcus predicted in 2024 that LLMs would run out of headroom, leading to a commodity market with no differentiation.
- 6 (15:19) **The Paradox: Dumb Models That Are Also Dangerous** - Marcus reconciles the view that AI is "dumber than we think" with the need for regulation.
- 7 (20:07) **The Regulatory "Sea Change" and the End of the Free Ride** - Marcus discusses the shift from a no-regulation stance to active investigations, driven by real-world harms.
+ Full timestamped outline available in the app
Show Notes
Ed Elson is joined by Gary Marcus to discuss why he’s concerned about the fact that we’re all-in on AI. They explore why he argues generative AI is inherently unreliable, whether the concerns surrounding Anthropic's Mythos model are justified, how policymakers should approach AI regulation, and the biggest misconception about the technology that he believes needs to be corrected.
Gary Marcus is a leading voice in artificial intelligence, author, and professor at NYU Stern.
Subscribe to the Prof G Markets Youtube Channel
Check out our latest Prof G Markets newsletter
Follow Prof G Markets on Instagram
Follow Ed on Instagram, X and Substack
Follow Scott on Instagram
Send us your questions or comments by emailing [email protected]
Learn more about your ad choices. Visit podcastchoices.com/adchoices
More from this podcast
Prof G Markets →