An interview with Arvind Narayanan and Sayash Kapoor on their new book AI Snake Oil, which is based on their popular newsletter about AI's shortcomings
A New Book by 2 Princeton University Computer Scientists X: Eric Topol / @erictopol : Is #AI snake oil? Some of it is, as asserted by @random_walker and @sayashk in a new book published today @PrincetonUPress. It's the subject of a new Ground Truths podcast, with transcript and links to key papers. Open-access, no ads. link in profile [image] @wired : In their book ‘AI Snake Oil,’ two Princeton researchers pinpoint the culprits of the AI hype cycle and advocate for a more critical, holistic understanding of artificial intelligence. https://www.wired.com/... Sayash Kapoor / @sayashk : 📣AI SNAKE OIL is out today! Writing this book over the last two years has been a labor of love. @random_walker and I are very excited to hear what you think of it, and we hope you pick up a copy. Some reflections on the process of writing the book 🧵 [image] Arvind Narayanan / @random_walker : 📢 AI Snake Oil is out today! I have too many feelings because this book has been five years in the making. In 2019 I gave a talk on AI Snake Oil and tweeted out the slides. I had no idea my career was about to change. Within a couple of days I had 30 or 40 invites to write a [image] Reece Rogers / @reece___rogers : Fascinating conversation with @random_walker and @sayashk about their new book, AI Snake Oil. It pinpoints the culprits of this AI hype cycle and advocates for a more holistic understanding of the technology: https://www.wired.com/...
Context & Ripple Effects
Narayanan’s critique predates the book: in an earlier interview, he characterized ChatGPT as a “bullshit generator” while developing a taxonomy for assessing AI claims. The book turns that continuing line of criticism into a more durable public reference point.
The release also sits within coverage of an AI hype cycle traced back to the transformer era and a later policy argument for treating AI as a normal technology rather than humanlike intelligence. That makes the interview less a standalone dissent than part of a widening debate over how AI claims should be evaluated.
First-order effects
- Narayanan and Kapoor gain a book-length platform for their newsletter’s critique, extending its reach beyond its existing readership through Wired and the Ground Truths interview.
- Readers and AI decision-makers get a consolidated critical framework for separating credible uses from overstated AI claims, rather than relying on promotional narratives alone.
Second-order effects
- The book adds pressure on AI vendors, buyers, and commentators to make narrower, testable claims about system capabilities and limits when facing a more visible critique of hype.
- Its framing can strengthen demand for operational evaluation and accountability practices, aligning with the broader need for ethical oversight beyond peer review.
Third-order effects
- If this kind of critique becomes more influential, AI competition may shift incrementally from broad capability narratives toward evidence of reliability, fit for purpose, and real-world performance.
- The policy debate may also become less centered on hypothetical humanlike AI and more on the concrete risks and governance of deployed systems, as argued in the normal-technology framing of AI.
The trend: AI discourse is moving toward tougher scrutiny of capability claims, with critics seeking to anchor adoption and policy in demonstrated performance rather than generalized intelligence narratives.