Researchers Arvind Narayanan and Sayash Kapoor explain their influential "AI as Normal Technology" framework, which argues that AI should be understood through the lens of historical technology adoption rather than as an unprecedented superintelligence threat. They emphasize that AI's societal impact depends more on deployment and diffusion barriers than raw capabilities, challenging the rapid adoption narrative by showing that meaningful integration takes time despite instant access to AI tools. The authors address criticisms, clarify misconceptions about their thesis, and explain why finding middle ground with the opposing "AI 2027" worldview proves difficult due to fundamentally different causal assumptions about technology and society.
Table of contents
Normal doesn’t mean mundane or predictableA restatement of our thesisIf disappointment about GPT-5 has nudged you towards AI as normal technology, it’s possible you don’t quite understand the thesisWhy it’s hard to find a “middle ground” between AI as Normal Technology and AI 2027It is hard to understand one worldview when you’re committed to anotherReaping AI’s benefits will require hard work and painful choicesThe surreal debate about the speed of diffusionWhy AI adoption hits differentConcluding thoughtsFurther reading/viewingSort: