Princeton's Narayanan at ICML: AI is augmentation, not automation — plenty of work left for humans
sayashk · x · 2026-08-27
Arvind Narayanan, Princeton CS professor and co-author of AI Snake Oil, delivered an ICML keynote titled "What will be left for us to work on?", addressing widespread anxiety about adapting as AI capabilities grow.
He made three arguments:
- The AI as Normal Technology framework is correct and useful for thinking about AI's impacts — absent a future discontinuity like recursive self-improvement, many bottlenecks remain between capability gains and job automation; evidence suggests AI is better seen as augmentation than automation.
- Human effort will shift toward less verifiable tasks — from developing models to scaffolds, and from building to evaluation and monitoring.
- Long-term, as purely technical skills are devalued, research work will migrate from problem solving to question asking and conceptual progress; in industry, relational skills, domain knowledge, and aesthetic/normative judgment will gain importance.
Narayanan and Sayash Kapoor have also compiled a curated reading list of essays on the framework, organized into foundational essays, applications, and technical explainers. Slides with an annotated transcript are publicly available.
Related event: Princeton Professor: AI Augments Rather Than Replaces Human Work(2 posts)→
More from AGI Musings
- COLM paper: LLMs claim multilingual support but fail on low-resource languages — anas_ant · 2026-08-27
- The Voluntarism Problem in AI Oversight: Incentives and Distortions — BlancheMinerva · 2026-08-27
- Bill Gates stakes his reputation on AI plus humanoid robots reshaping the economy — BenBajarin · 2026-08-27
- Robotics Moat: Data, Hardware, and the Deployment Layer Flywheel — chris_j_paxton · 2026-08-27
- Speculation: Anthropic Hiding Mythos Signals It Won't Release AGI Publicly — BLUECOW009 · 2026-08-27
- People will be disappointed when they realize AI isn't a bubble — ZeYanjie · 2026-08-27