Narayanan and Kapoor: AI Is Normal Tech, Not a Superintelligence

Arvind Narayanan and Sayash Kapoor want to pull the AI debate back to earth. Their essay in the Knight First Amendment Institute sets out "a vision of artificial intelligence (AI) as _normal technology_," framed to serve as "a description of current AI, a prediction about the foreseeable future of AI, and a prescription about how we should treat it."

The numbers do the arguing. "In August 2024, 40% of U.S. adults used generative AI," they write, but that reach "only translated to 0.5%-3.5% of work hours (and a 0.125-0.875 percentage point increase in labor productivity)." Electric dynamos, they note, were "everywhere but in the productivity statistics" for nearly 40 years after Edison's first central generating station.

What slows AI down in practice, the authors argue, is a "capability-reliability gap" that "shows up over and over" and remains "a major barrier to building useful AI 'agents' that can automate real-world tasks." Benchmarks hide it. GPT-4 "reportedly achieved scores in the top 10% of bar exam test takers," they write, but "this tells us remarkably little about AI's ability to practice law." The self-driving industry already proved the point: "Cruise is set to shut down in 2025, while Uber was forced to sell off its self-driving unit."

Model-level alignment, they say, has "proved to be extremely brittle." The useful defenses sit downstream, in deployment environments and institutional controls. The paperclip-maximizer scenario is not dismissed, but placed in its own category, grounded less in precedent than in open epistemic uncertainty about whether the risk is zero.

Two researchers in our Who's Who tracker have shared the link.