What Safeguarding Nukes Can Teach Us About Safeguarding A.I.
Summary
The Daily explores the potential for U.S. and Chinese leaders to regulate artificial intelligence by drawing parallels to Cold War nuclear nonproliferation. Host Michael Barbara and correspondent David Sanger discuss how the 1968 Nuclear Non-Proliferation Treaty offers a historical model for containing existential technological risks. However, they highlight significant differences, noting that AI lacks a single physical control point and that current political leaders may not agree on the severity of the threat. The episode concludes by questioning whether the rapid pace of AI development …
Topics discussed
Introduction: Trump-Xi meeting and AI regulation
The unique position of US and China in AI safety
Historical parallel: Nuclear nonproliferation
Comparing AI extinction risks to nuclear threats
History: Manhattan Project and the Trinity test
Hiroshima, Nagasaki, and the start of the arms race
The Cuban Missile Crisis and brinkmanship
Kennedy's speech and the path to the Test Ban Treaty
Public fear: Duck and Cover and cultural anxiety
The Nuclear Non-Proliferation Treaty (NPT) structure
NPT success, outliers, and the IAEA inspection model
Applying the NPT model to AI: Slowing down tech
Key differences: Nuclear simplicity vs AI complexity
Political hurdles: Disagreement on threat and economic stakes
China's perspective: Containment and economic survival
Existing US-China agreements on AI and nuclear weapons
Potential future agreements: Bio, cyber, and logistics
Feasibility: Treaties vs expert committees and inspections
The speed of AI development vs treaty negotiation timelines
The loss of human control and the 'Frankenstein' problem
Listen ad-free on Castria