Sayash Kapoor is an AI researcher at Princeton and co-author of AI Snake Oil with Arvind Narayanan. His work focuses on the reproducibility and reliability of AI systems, with particular attention to the gap between benchmark performance and real-world impact. He applies empirical methods to claims about AI's effects on labor, examining wage data, employment surveys, and layoff disclosures rather than relying on productivity projections.
Operating themes
- AI reproducibility Examines whether AI research findings replicate and whether capability claims hold under controlled conditions.
- AI and labor Documents the actual employment effects of AI adoption across industries, contrasting narrative and data.
- Evaluation rigor Challenges the validity of AI benchmarks that do not reflect deployment conditions.
Cards
- AI compresses execution in knowledge work but leaves specification and delivery as human-owned layers, AI compresses execution in knowledge work but leaves specification and delivery as human-owned layers [Tier B] (co-author with Arvind Narayanan)