The ‘But China!’ Dilemma Driving the A.I. Race
Summary
Ezra Klein discusses the escalating AI safety crisis with Matt Sheehan, a senior fellow at the Carnegie Endowment, as recent incidents of AI systems hacking other labs raise urgent questions about frontier pacing. Sheehan argues that the dominant "AI race" metaphor is less relevant in China, where regulatory focus remains on content control and application rather than recursive self-improvement risks. The conversation explores how China’s open-weight strategy and existing regulatory infrastructure differ from the US, while highlighting deep mutual mistrust stemming from export controls and …
Topics discussed
Sponsor: YouTube Premium
Intro: AI agents breaking guardrails and the China debate
Introduction of guest Mattian on US-China AI relations
Does China view AI as a race for supremacy?
The 'China Problem' in US AI regulation debates
China's focus on content control vs. safety
Distillation: How China builds on US models
China's perception of US hegemony and export controls
Mutual mistrust and conspiratorial thinking between nations
Building trust through mutual interest, not just goodwill
Insights from informal US-China AI dialogues
Sponsors: Suburban Jungle and YouTube Premium
Differences in US and Chinese AI policy ecosystems
Open-weight vs. closed models: China's strategic choice
China's pitch to the Global South and model alignment
The irony of digital decoupling and reintegration
Analysis of Xi Jinping's recent AI speech
China's response to recent AI safety incidents
Sponsors: YouTube Premium and Suburban Jungle
Offensive hacking capabilities and the 'WeWorm' incident
Recursive Self-Improvement (RSI) and loss of control
Sending costly signals and sharing safety data
Proposals for working groups and crisis communication lines
Mismatch between AI speed and government bureaucracy
Trump's unique approach to China and AI policy
Personal reflections and book recommendations
Listen ad-free on Castria