The Last Invention The AI Skeptics
107 snips
Feb 20, 2026 
Arvind Narayanan, a Princeton CS professor studying AI policy; Gary Marcus, a cognitive scientist critiquing deep learning; and Ed Zitron, tech critic and newsletter writer. They debate whether AI is hype and grift, limits of neural nets and why scaling may fail, and why AI will diffuse through institutions slowly with human systems shaping its adoption.
AI Snips
Chapters
Books
Transcript
Episode notes
Critique Of Public Doomsayers
- Ed Zitron criticizes prominent figures like Geoffrey Hinton for public doomerism and speaking fees.
- He contrasts their media warnings with lack of tangible work on practical harms like data theft and layoffs.
Augment LLMs With Symbolic Systems
- Try integrating symbolic systems and world models into LLMs rather than relying solely on scaling.
- Combine different approaches to overcome reasoning and alignment failures in current models.
Dress Rehearsal Exposes Policy Failure
- Gary Marcus calls LLMs a 'dress rehearsal' revealing how money and power blunt regulatory responses.
- He warns investing heavily in poorly aligned tech diverts resources from building safer alternatives.
