Keen On America 2025: The AI Year Scripted by Gary Marcus in 2024
Dec 23, 2025
Gary Marcus, a cognitive scientist and AI researcher known for his critical views on AI models, dives into the realities of artificial intelligence. He discusses the limitations of large language models and the economic bubble around companies like OpenAI. Marcus predicts a turbulent future for AI in 2026, emphasizing the need for regulatory oversight and safer AI practices. He also highlights the competition between the U.S. and China and advocates for neurosymbolic approaches as the way forward. Will we finally see practical robots or merely demos? Tune in to find out!
AI Snips
Chapters
Transcript
Episode notes
Shift From Pure LLMs To Modular Systems
- Gary Marcus argues AI progress is shifting toward more modular, neurosymbolic systems and world models rather than pure LLM scaling.
- He predicts history will view this shift as corrective movement away from five years of overreliance on LLMs.
2025 As A Year Of Disillusionment
- Marcus sees 2025 as a year of retreat from AGI hype and some disillusionment after an earlier surge in AI enthusiasm.
- He cites replicated studies showing many companies get little ROI from current generative models.
Nvidia's Growth Depends On Sustained Demand
- NVIDIA benefited from strong demand but faces risk if chip purchases slow and an oversupply emerges.
- Marcus warns the chip boom depends on sustained demand tied to AGI expectations that may not materialize.