The Artificial Human AI's Bubble Trouble?: 3. If the bubble bursts, where next for AI?
Jan 28, 2026
Gary Marcus, cognitive scientist and author warning about overreliance on large language models. Adrian Lepers, Hugging Face monetization lead working on open-source model deployment and economics. They discuss whether the AI investment boom is a bubble. They contrast flashy LLMs with practical, smaller models. They explore market correction, consolidation, and how specialized AI might survive and reshape the industry.
AI Snips
Chapters
Transcript
Episode notes
Specialized Models Often Beat Giant LLMs
- Large language models (LLMs) dominate attention and investment but many practical AI uses are narrow and better served by smaller models.
- Adrian Lepers argues specialized, cheaper models often outperform LLMs for real-world tasks and lower operational cost.
Cost And Speed Favor Smaller Open Models
- Cost and speed make smaller open models attractive because each request to a huge model carries a high price and latency cost.
- Adrian Lepers highlights that cheaper, faster models encourage broader internal adoption within companies.
Correction Leads To Market Consolidation
- Market corrections naturally follow periods of overheated investment until valuations align with reality.
- Adrian Lepers expects consolidation where few companies win rather than many sustaining extreme valuations.