Skip to content

Spend the Lead on Safety: Biden AI Advisors Ben Buchanan & Tantum Collins on the Fate of the World

The Cognitive Revolution

1h 49m
Transcript

Original podcast

Where to listen

Summary

Written by a model from the transcript; not yet reviewed.

Former White House special advisor Ben Buchanan and former NSC director Tantum Collins discuss their book, The Bitter Struggle, exploring how a small group in the Biden administration anticipated the rise of AI and pushed for advanced chip export controls against China. They explain their theory of denying computing power to build a strategic lead, which the United States and its allies can then spend on safety and careful pacing. The conversation addresses the balance between tech optimism and realist geopolitics, the challenges of international diplomacy with China, and the risk that AI could drive centralized administration to the point of a slow creep toward autocracy.

  • The United States maintains roughly a hundred to one advantage in producing advanced computing power and about fifteen to one in access to it. 1:17
  • Export controls and hardware supply chain control allow democracies to deny computing power to strategic competitors and build a safety lead. 30:21
  • Building stabilizing global institutions that cede a portion of national power is the historical path for managing revolutionary new technologies. 47:27
  • Centralized administration enabled by AI creates a bitter temptation where democracies slowly creep toward autocracy on their own. 4:12

Chapters

  1. 0:00
    Episode Introduction and Host Setup

    Hosts introduce the podcast, the AI voice clone, and set the stage for the discussion.

  2. 0:33
    Compute Lead and Safety Strategy

    Buchanan explains the U.S. compute advantage and the plan to use it for AI safety rather than a race.

  3. 2:00
    Policy Formation and Ideological Tensions

    Exploration of how policy was shaped in the White House, the tension between idealism and realism, and the influence of private sector vision.

  4. 3:30
    Export Controls and Global Impact

    Analysis of export controls, their effectiveness, and the broader geopolitical implications for AI development.

  5. 41:52
    Future Cooperation and Strategic Positioning

    Hosts discuss potential future U.S.-China cooperation on AI safety and the importance of maintaining a strategic advantage.

  6. 45:12
    Closing Reflections and Takeaways

    Final thoughts on the challenges of AI governance, the role of democratic institutions, and the path forward.

  7. 48:06
    US Institutional Legacy and AI Governance

    The speakers discuss how the U.S. historically builds global institutions and how that model applies to AI governance, emphasizing the need for stabilizing bodies that include China.

  8. 49:48
    International AI Safety Institute Initiative

    They outline the creation of the AI Safety Institute and its network, inviting Chinese researchers and discussing timelines for global cooperation.

  9. 53:01
    Challenges of Rapid AI Development

    The conversation shifts to the urgency of AI progress, the risk of recursive self‑improvement, and the difficulty of keeping pace with safety measures.

  10. 55:22
    China’s Role and Compute Controls

    The speakers discuss China’s AI ecosystem, compute restrictions, and the potential for a bilateral deal to maintain U.S. lead while engaging China.

  11. 56:59
    Future of AI Policy and Legislative Momentum

    They conclude by reflecting on the growing political momentum for AI policy and the need for proactive, coordinated action.

  12. 1:11:21
    Government‑Industry Coordination and Regulation

    They debate how government can better understand and regulate private AI firms, proposing embedded evaluators and mandatory pre‑deployment testing.

  13. 1:19:37
    Political Salience of AI in U.S. Elections

    They examine how AI has become a major issue in recent elections, noting increased public concern and legislative attention to data centers.

Highlights

  • 1:01

    The Core Theory of the Casestrategy

    The guests explain their core strategy: because the US and its allies control the advanced chip supply chain, they can deny China compute, build a lead without reckless racing, and spend that lead on safety.

  • 4:07

    The Bitter Temptation of Centralizationinsight

    Tantum Collins outlines how AI could make centralized administration so effective that democracies slowly drift toward autocracy on their own unless public oversight and popular sovereignty scale up.

  • 7:15

    GPT-3 as the Policy Turning Pointstory

    Ben Buchanan identifies the release of GPT-3 in May 2020 as the technical turning point that convinced him to sound the alarm and urge the government to take compute scaling seriously.

  • 24:50

    The Maximal Sustainable Level of Idealisminsight

    The guests debate how to find the sweet spot between utopian tech ambitions and hard-nosed realist geopolitics, avoiding naive optimism while escaping destructive zero-sum traps.

  • 42:22

    Geneva Talks and Chinese Skepticismstory

    The guests recount meeting with Chinese officials in Geneva in May 2024, where the delegation treated shared AI risk discussions as a plot and showed little interest in collaboration.

  • 1:09:08

    Defensive Advantage in Cybersecurityinsight

    The guests argue that American frontier labs turned powerful cybersecurity capabilities over to defenders rather than using them offensively, contrasting this with how an autocratic nation might act.

Quotes

  • “Because the US and its allies control the supply chain for advanced chips, we can deny China computing power and build a lead without racing recklessly ourselves.”
    Unnamed speaker · 1:01
  • “Chips for controls, he says, is the end game.”
    Unnamed speaker · 3:25
  • “We want to think about sequencing these actions so that we can produce outcomes that get the things that are clearly good for the world and avoid things that are clearly bad for the world.”
    Unnamed speaker · 28:51
  • “There was always just enough virtue in this republic to save it. Sometimes none to spare, but still enough to meet the emergency.”
    Unnamed speaker · 38:05
  • “We spent a lot of time having engagement with the Chinese in the Biden administration.”
    Unnamed speaker · 41:54
  • “think there's a strong history here of the United States saying, look, we're ahead.”
    Unnamed speaker · 48:24
  • “And this was the first attempt to give the government real technical expertise to manage”
    Unnamed speaker · 50:01
  • “We spent a ton of time thinking about the ways in which overregulation harmed nuclear power and all that kind of stuff.”
    Unnamed speaker · 1:10:05
  • “I think that the real aliens here are the AIs, not the Chinese.”
    Unnamed speaker · 58:01
  • “We know the solutions here are gonna at some level be geopolitical, if not also political.”
    Unnamed speaker · 1:34:57
  • Anthropic 15:25–16:43
  • Outsystems 16:07–18:07
  • Tasklet 30:49–32:36
  • Granola 32:30–33:21

About this episode

Former Biden administration AI advisors Ben Buchanan and Tantum Collins join Nathan to discuss their book, The Bitter Struggle, and their experience shaping U. S. policy on compute supply chains and export controls. Drawing on their backgrounds in government and DeepMind, they unpack the geopolitical strategy of building an allied compute lead over China in order to spend that advantage on safety. The conversation probes the clash between frontier lab idealism and national security, bilateral talks over AI risks in nuclear command, and persistent technical hurdles like reward hacking. For full show notes, links, and references, read the episode page:https://www.cognitiverevolution.ai/spend-the-lead-on-safety-biden-ai-advisors-ben-buchanan-tantum-collins-on-the-fate-of-the-world/ Sponsors: Claude: Claude is the AI collaborator built for problem solvers, extending your thinking across coding, modeling, and strategic workflows. Get started and discover Claude Pro at https://claude.ai/tcr OutSystems: OutSystems is the leading agentic systems platform, empowering enterprises to build, coordinate, and govern AI agents and mission-critical applications securely. Learn more and start building your agentic future at https://outsystems.com/tcr Tasklet: Tasklet empowers your business with AI agents that connect to your tools and automate recurring workflows with no code required. Visit https://tasklet.ai and use code cog rev for $50 in free credits Granola: Granola is an AI-powered notepad that captures your meetings and turns them into clean, searchable notes with clear decisions and action items. Try it completely free on your next call at https://granola.ai/tcr CHAPTERS: (00:00) About the Episode (05:04) Conviction in scaling laws (11:22) Inevitability versus human choice (Part 1) (15:30) Sponsors: Claude | OutSystems (18:07) Inevitability versus human choice (Part 2) (19:50) Private labs and ideology (26:22) Spending the lead wisely (Part 1) (30:52) Sponsors: Tasklet | Granola (33:25) Spending the lead wisely (Part 2) (41:52) US and China diplomacy (54:05) Pacing frontier AI development (01:09:03) Regulating the frontier labs (01:18:39) Data centers and energy (01:24:19) Resisting the bitter temptation (01:32:50) Managing catastrophic emergency scenarios (01:38:45) Policy lightning round (01:44:16) Episode Outro (01:47:42) Outro PRODUCED BY: https://aipodcast.ing

Transcript

0:00
1:49:42

20,377 words · 3 speakers

Topics

  • AI Governance
  • Geopolitics
  • Export Controls
  • AI Safety
  • Semiconductors
  • Democratic Institutions
  • National Security

Page updated · transcript and summary are machine-generated