Arvind Narayanan
Computer scientist; co-author of 'AI Snake Oil' and 'AI as Normal Technology' (with Sayash Kapoor)
Princeton University
Skeptics
Narayanan argues AI should be treated as a 'normal technology' whose effects unfold over decades through slow institutional change, not as a looming superintelligence. He thinks doom probabilities are unusable for policy and that the best protection is resilience and preventing a few companies from concentrating power.
Declines to give a number
Classifies catastrophic misalignment as a 'speculative risk' and explicitly rejects using existential-risk probabilities for policy as 'unviable'; argues 'superintelligence' conflates intelligence with power (AI as Normal Technology, April 15, 2025).Source
On timing
AI is a 'normal technology' like electricity: invention can be fast but diffusion is 'limited by human, organizational, and institutional change', so economic impacts materialise over decades, not years (April 2025).
SourceWhat they call for
- Resilience over nonproliferation
- Prevent concentration of power
- Sector specific safety regulation
- Transparency and monitoring
- Open source AI
- Defences against misuse downstream rather than model restrictions
In their own words
Policymakers should "increase resilience by preventing the concentration of power" through distributed access, rather than concentrating control of AI.
15 April 2025
Real concern is systemic risks from normal deployment—bias, job displacement, inequality, authoritarianism—rather than catastrophic scenarios.
15 April 2025