Future Knowledge AI As Normal Technology
Feb 25, 2026
Sayash Kapoor, a computer scientist and Princeton PhD candidate who co-authored AI as Normal Technology and AI Snake Oil, argues AI is ordinary infrastructure, not magic. He contrasts generative vs predictive systems. He explains deployment bottlenecks, institutional barriers, policy priorities, and risks like biosecurity while urging clearer governance and realistic expectations.
AI Snips
Chapters
Books
Transcript
Episode notes
Don't Confuse Bench Scores With Real Competence
- Adjust expectations: benchmark model claims against real-world tasks rather than impressive test scores.
- GPT-4 passing bar-style exams doesn't equal competence at day-to-day legal work, so verify on domain tasks before trusting outputs.
I Stopped Typing Code Every Day
- Sayash stopped writing code daily after AI tools became practical, shifting to higher-level design work.
- He reports not typing code the last two years, reflecting programmer abstraction shifts enabled by AI.
AI Follows The Dynamo Pattern
- AI should be seen as a general purpose technology that diffuses through institutional changes rather than an instant miracle.
- The dynamo example shows 40 years of reorganization before electrified factories delivered widespread benefits.
