Beyond the Hype: Artificial Intelligence, Institutional Failure and the Limits of Prediction

Artificial intelligence is increasingly presented not merely as a technological development but as a force capable of transforming institutions, labour markets, public administration and democratic life. The legal significance of that transformation lies less in the abstract question of whether machines will become “intelligent” than in the concrete question of who is permitted to deploy automated systems, on what evidence, against whom, and with what safeguards. Arvind Narayanan and Sayash Kapoor’s AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference enters this debate at an important moment. Published by Princeton University Press in 2024, the book distinguishes between different forms of AI and challenges the tendency to treat the label “AI” as a guarantee of technical competence. Its central warning is that institutional actors can be persuaded to purchase or deploy systems that appear scientific while being unable to perform the tasks for which they are marketed.