Exploring the Pitfalls of AI Predictions

The episode begins with Kapoor explaining the origins of AI Snake Oil, tracing it back to his PhD research at Princeton on AI's limited predictive capabilities in social science domains. He shares how he and co-author Arvind Narayanan uncovered major methodological flaws in civil war prediction models, which later extended to other fields misapplying machine learning.

The conversation then turns to the disconnect between academic findings and media narratives. Kapoor critiques the hype cycle around AI, emphasizing how its real-world adoption is slower, more fragmented, and often augmentative rather than fully automating human labor. He cites the enduring demand for radiologists as a case in point.

Kapoor introduces the concept of “AI as normal technology,” which rejects both the notion of imminent superintelligence and the dismissal of AI as a passing fad. He argues that, like other general-purpose technologies (electricity, the internet), AI will gradually reshape industries, mediated by social, economic, and organizational factors—not just technical capabilities.

The episode also examines the speculative worldviews put forth by documents like AI 2027, which warn of AGI-induced catastrophe. Kapoor outlines two key disagreements: current AI systems are not technically on track to achieve general intelligence, and even capable systems require human and institutional choices to wield real-world power.

On policy, Kapoor emphasizes the importance of investing in AI complements—such as education, workforce training, and regulatory frameworks—to enable meaningful and equitable AI integration. He advocates for resilience-focused policies, including cybersecurity preparedness, unemployment protection, and broader access to AI tools.

The episode concludes with a discussion on recalibrating expectations. Kapoor urges policymakers to move beyond benchmark scores and collaborate with domain experts to measure AI’s real impact. In a rapid-fire segment, he names the myth of AI predicting the future as the most misleading and humorously imagines a superintelligent AI fixing global cybersecurity first if it ever emerged.

Episode Contributors

Sayash Kapoor is a computer science Ph.D. candidate at Princeton University's Center for Information Technology Policy. His research focuses on the societal impact of AI. He previously worked on AI in the industry and academia at Facebook, Columbia University, and EPFL Switzerland. He is a recipient of a best paper award at ACM FAccT and an impact recognition award at ACM CSCW.

Nidhi Singh is a senior research analyst and program manager at Carnegie India. Her current research interests include data governance, artificial intelligence and emerging technologies. Her work focuses on the implications of information technology law and policy from a Global Majority and Asian perspective. 

Suggested Readings

AI as Normal Technology by Arvind Narayanan and Sayash Kapoor.

Every two weeks, Interpreting India brings you diverse voices from India and around the world to explore the critical questions shaping the nation's future. We delve into how technology, the economy, and foreign policy intertwine to influence India's relationship with the global stage.

As a Carnegie India production, hosted by Carnegie scholars, Interpreting India, a Carnegie India production, provides insightful perspectives and cutting-edge by tackling the defining questions that chart India's course through the next decade.

Stay tuned for thought-provoking discussions, expert insights, and a deeper understanding of India's place in the world.

Don't forget to subscribe, share, and leave a review to join the conversation and be part of Interpreting India's journey.