Nathan Goes to China #3: US-China Relations, the Art of the AI Deal & the Road to Pax Robotica

Episode Summary

Executive Summary: Nathan Gardels argues the US and China should seek a “Pax Machina/Pax Robotica” focused on shared AI prosperity, not victory. He contends China is a durable great power with deep AI talent, legitimate domestic support, and real safety concerns, so containment is futile. The US should stay strong, learn from China’s state capacity, and pursue de-escalation, interoperability, and selective cooperation to avoid an AI arms race.

Main Topics: Goal for US-China relations: Pax Machina (Priority: 5/5): The episode’s central thesis is that the objective should be mutual prosperity and safety through AI, not geopolitical victory or containment. Nathan frames this as a peaceful, abundance-oriented future where both countries share AI’s benefits while avoiding catastrophic risks. China’s long-run strength and AI competitiveness (Priority: 5/5): Nathan argues China’s resurgence is a return to historical normalcy, not a recent anomaly. He says China has deep human capital, substantial economic strength, and a large, competitive AI ecosystem that makes attempts to permanently slow it unrealistic. Chinese perspective: humiliation, sovereignty, and legitimacy (Priority: 5/5): He reconstructs how Chinese elites interpret technology denial and export controls as efforts to keep China down. This is tied to the century of humiliation, fear of dependency, and strong belief that technological self-sufficiency is essential to national security and legitimacy. What the US should learn from China (Priority: 4/5): The episode argues the US should emulate China’s state capacity, infrastructure execution, industrial policy seriousness, and willingness to build. But it should do so with American characteristics—preserving openness, immigration strengths, and democratic values. AI governance, export controls, and mutual legibility (Priority: 5/5): Nathan is skeptical of export controls because they slow cooperation, harden Chinese resolve, and risk decoupling the AI stacks so deeply that each side can no longer understand the other’s capabilities. Shared paradigms, standards, and collaboration are presented as safety priorities. Concrete proposals for de-escalation and cooperation (Priority: 4/5): He suggests unilateral trust-building steps, clearer guidance that safety research is allowed, possible chip-for-manufacturing/tech exchange, more open collaboration, and future mechanisms like confidential-computing-based audits, joint compute, and even conventions on autonomous weapons.

Key Arguments: The US should not aim to “win” the U.S.-China relationship; it should aim for peace, abundance, and broad access to AI benefits. China’s rise in both the broader economy and AI is not a temporary spike but a reversion to a long historical norm of great-power status. Attempting to permanently contain Chinese technological progress will only work briefly; China has the talent and state commitment to catch up. Chinese elites broadly see export controls and rhetoric about “winning the AI race” as attempts to keep China down, which strengthens hardline resolve. China’s AI community shows real safety awareness, open publication in English, and meaningful open-weight releases, making it a provider of global AI public goods. The US should learn from China’s strengths—state capacity, manufacturing execution, infrastructure, and industrial policy—without copying its political system. Export controls may buy time but also fracture shared AI paradigms, reduce mutual understanding, and hinder AI safety collaboration. A better strategy is selective recoupling: trade chips and know-how for Chinese strengths in batteries, solar, robotics, manufacturing, and other areas. The US should avoid banning Chinese models outright; a market/insurance-based approach to AI risk is more consistent with American values. The most dangerous future is one where the two countries’ AI stacks diverge so much that neither side can accurately assess the other’s capabilities. The US should support AI safety collaboration, transparent standards, and possibly joint/third-country compute arrangements to maintain legibility and trust. Both sides have real grievances, but future agreements are still possible if leaders stop assuming that cooperation is impossible.

Data Points: US GDP (nominal): about $30 trillion - Nathan cites headline U.S. output as a baseline comparison in the economic power discussion. China GDP (nominal): about $20 trillion - Used to show the U.S. still leads in dollar terms. China GDP (PPP): about $45 trillion - Nathan argues PPP better reflects lived economic scale and implies China may be larger in real terms. US vs China GDP (PPP comparison): China ~50% bigger than the US - His interpretation of PPP-based economic size. High-speed rail ticket cost: about $60 - Nathan’s Beijing-to-Shanghai train ride used to illustrate infrastructure quality and affordability. Pedicure cost in Shanghai: about $12 for ~45 minutes - Anecdote illustrating low service prices and broader consumer affordability. Pedicure worker’s phone: iPhone 17 - Used to suggest even working-class urban Chinese consumers can afford premium devices. Chinese AI five-year plan mention: 2016 - Nathan cites this as evidence that China has been strategically investing in AI for at least a decade. Count of frontier/near-frontier companies: 11 American and 12 Chinese - ChatGPT-generated rough comparison of companies shipping frontier or near-frontier LLMs. SuperGLUE final leaderboard year: 2022 - Nathan notes that Chinese companies were already highly competitive before ChatGPT’s launch. He Jiankui prison sentence: 3 years - Used to argue China can be conservative and punitive on controversial biotech/genetic experimentation. He Jiankui fine: 3 million yuan - Associated with the embryo-editing case. Trump-Xi summit timing: a couple weeks away - Frames the episode as a timely policy argument ahead of the meeting. AI safety hub founding board: 5 founding board members incl. a European professor - Example of China’s stated openness to international AI safety collaboration. Open letter scale: over 1,000 people - Nathan references an open letter advocating mechanisms to slow AI capability advances. Nuclear weapons stockpile: 10,000 or so - Used as a comparison for an AI-enabled “sword of Damocles.”

Pivotal Quotes: "our goal should not be to win, but rather to realize a future of peace and abundance" — Nathan Gardels: Defines the episode’s central normative aim for U.S.-China relations. "Technology races can raise the risk of safety catastrophes." — Dario Amodei (quoted by Nathan): Nathan uses this Anthropic-era statement to argue against escalatory AI competition. "We should jointly oppose overstretching the national security concept in the field of AI and placing one country's security over that of others." — Xi Jinping (quoted by Nathan): Nathan presents this as an olive branch and evidence of Chinese rhetoric favoring cooperative AI governance.

Implications: Listeners should expect sustained U.S.-China AI competition, but not inevitable conflict. The policy path Nathan favors is selective recoupling, shared standards, and safety collaboration to preserve mutual understanding and reduce the odds of an AI arms race.