What Safeguarding Nukes Can Teach Us About Safeguarding A.I.
Summary
Host Michael Barbaro and guest David Sanger discuss the upcoming Trump-Xi summit, focusing on the urgent need for US-China cooperation to regulate artificial intelligence. Sanger draws a historical parallel to the Nuclear Non-Proliferation Treaty, arguing that while the AI landscape is more complex and faster-moving, a similar framework of containment and inspection is necessary to prevent civilizational catastrophe. The conversation highlights the significant challenges in achieving this, including differing national interests, the rapid pace of technological development, and the lack of a…
Topics discussed
Sponsor: YouTube Premium
Intro: Trump-Xi meeting and AI risks
US-China position on AI safety
Historical parallel: Nuclear non-proliferation
Comparing AI extinction risk to nuclear era
History: Manhattan Project and Trinity test
Hiroshima and Nagasaki bombings
The nuclear arms race and proliferation
The Cuban Missile Crisis
Kennedy's speech and Test Ban Treaty
Cultural impact: Duck and Cover and satire
Signing the Nuclear Non-Proliferation Treaty
Treaty terms and IAEA inspections
Outliers and the success of containment
Sponsors: YouTube, Capital One, Book Club
Why the nuclear model is hard to apply to AI
Political disagreements on AI threats
Economic incentives and chip export controls
Existing US-China AI agreements
Expanding scope: Bio and cyber weapons
Feasibility of an AI treaty and inspections
Time pressure: AGI vs. treaty timelines
Loss of human control over AI agents
Conclusion and sign-off
Sponsors: YouTube, Capital One, Book Club
News: Media access and Iran statement
Credits and outro
Sponsor: Capital One
Listen ad-free on Castria