Kedrosky criticizes Huang's AI safety logic in Klein interview
Paul Kedrosky argues Jensen Huang's approach to AI safety avoids the collective-action problem that makes individual responsibility insufficient.
What to know
- Jensen Huang and Ezra Klein clashed on fundamentals: what kind of system AI is, what risks it creates, and how to know about both.
- Kedrosky argues Huang's position—that firms should simply not ship unsafe models—ignores the economic prisoner's dilemma: each lab has incentives to move faster than collectively optimal, with gains immediate and costs delayed.
- Huang's framing treats safety as an engineering problem firms should solve individually, but Kedrosky contends individual responsibility is insufficient without mechanisms that force coordinated restraint.
“Every participant has powerful incentives to move faster than is collectively desirable. The gains from defecting are immediate and enormous; many of the costs are delayed, distributed, and poorly handled by ordinary litigation.”
Paul Kedrosky, Analyst · paulkedrosky.com ↗
Jensen Huang Nvidia CEO
Ezra Klein New York Times columnistPaul Kedrosky Analyst and critic
How it unfolded 1 development · click the chart to see its coverage articlesposts
-
1
Alec Stapp flags factual errors in Huang's interview
On X, Alec Stapp states that Jensen made false claims and cites Zvi for explanation of Huang's errors regarding Chinese open source.
“Huang kept answering like an engineer-in-a-box: unsafe models should not be shipped; safety is an engineering problem; more compute will make systems safer and more useful.”
— Paul Kedrosky, Analyst · source -
Jensen said a bunch of false stuff in his interview with Ezra. Good explanation from Zvi on what he gets wrong about Chinese open source:
-
-
background
Kedrosky: Huang's safety position masks a prisoner's dilemma — Paul Kedrosky analyzes the interview, arguing that Huang's engineering-focused safety framing—"don't ship unsafe models"—fails to address the structural incentive problem driving labs to move faster than is collectively desirable.
-
background
Ezra Klein interviews Jensen Huang on AI safety — The New York Times' Ezra Klein interviewed Nvidia CEO Jensen Huang, with the two talking past one another on core questions about AI risks, regulation, and employment.