AI Under Trump? The Stakes of 2024 w/ Joshua Steinman [Pt 2 of 2]

Episode Summary

Executive Summary: The episode is a long-form debate about whether a potential Trump presidency is preferable for AI policy and national security. Nathan argues Trump would intensify China tensions, accelerate arms racing, and behave too erratically for a volatile AI era, while Josh Steinman—drawing on Trump administration experience—argues the U.S. must confront China as an adversary, prioritize domestic capability, and that trust-building or de-escalation is unrealistic. The core disagreement is whether AI risk is best reduced through cooperation and restraint or through hard power and strategic competition.

Main Topics: Trump, AI, and the 2024 election (Priority: 5/5): Nathan frames the election as a referendum on Trump specifically as it relates to AI risk, leadership discipline, and whether a loose cannon is appropriate during a period that may include AGI. U.S.-China AI arms race (Priority: 5/5): The central debate concerns whether an AI arms race with China is inevitable or avoidable. Nathan wants de-escalation and mutual restraint; Josh says strategic competition is already underway and must be met with American strength. AI existential risk and timelines (Priority: 5/5): Both speakers treat AI as potentially transformative soon, but Nathan emphasizes tail risk and governance while Josh downplays the policy significance of the threat relative to geopolitics and industrial competition. Trust, sanctions, and chip export controls (Priority: 4/5): Nathan criticizes chip bans and escalation as premature and counterproductive to trust-building; Josh defends sanctions and argues advanced compute should not be handed to a strategic adversary. Leadership style and executive decision-making (Priority: 4/5): Nathan argues that in crises the person matters and Trump’s lack of rigor makes outcomes worse; Josh says Trump is highly strategic and that media narratives obscure his competence. Positive-sum social contract and abundance (Priority: 3/5): Nathan repeatedly argues AI should be used to create abundance, shared benefits, and a global positive-sum order rather than zero-sum nationalism. Media bias and insider experience (Priority: 3/5): Josh claims mainstream coverage distorts Trump and government reality, and says his firsthand White House experience showed Trump to be unusually sharp and ahead of advisers.

Key Arguments: Nathan argues that if AGI is plausible within a few years, presidential choices materially affect AI outcomes and should weigh heavily in voting decisions. Nathan says Trump would likely worsen U.S.-China relations through inflammatory rhetoric, tariffs, and decoupling, increasing the chance of a destabilizing AI arms race. Nathan believes AI risk is real but not inevitable; competent, stable leadership can shift probabilities downward compared with impulsive leadership. Nathan argues chip bans and decoupling are escalation that China will interpret as containment, while trust-building and later option value would be more prudent. Josh argues the Chinese Communist Party is a hostile strategic adversary and the U.S. must assume competition, not partnership, in technology and security. Josh says sanctions and export controls are necessary because allowing advanced chips to China would aid an opponent in a real arms race. Josh maintains that Trump is highly intelligent, communicates well, and often thinks several steps ahead of his advisers, contrary to media portrayals. Josh argues policy is made through leverage and concrete power, not aspirational language, so “words on a sheet of paper” are insufficient to prevent dangerous outcomes. Nathan contends that global cooperation is essential for AI safety because militarized competition raises the odds of catastrophic misuse, including autonomous weapons and escalation. Josh counters that domestic capability, continued pressure on China, and U.S. chip manufacturing are the practical plan, while broad trust-building is unrealistic and naive.

Data Points: AGI timeline cited: 2–3 years - Josh says powerful AI systems could appear as soon as the next two to three years. AGI near-term possibility: 2027 - Nathan references the idea that AGI in the 2027 timeframe should be politically decisive. AI policy threshold: 10^26 - Nathan says the Biden administration's 10^26 reporting threshold has aged well as a 2023 rule. Potential U.S. energy-grid limit: Until about 2026 - Josh references others' claims that the U.S. grid can support current AI power growth only until roughly 2026. Chinese power expansion goal: 3x to 7 terawatts/year in 15 years - Josh cites this as evidence China has major domestic power needs and is expanding aggressively. U.S. intellectual property theft estimate: Over $1 trillion - Josh cites this as the amount of stolen IP attributed to China over the past 20 years. Trump admin tenure: 2017 to 2021 - Josh served in the Trump administration throughout the term. War/arms-control reference: 20,000 deployed nuclear weapons - Nathan says the world has far too many nuclear weapons and that this level is unhealthy. National security role: Senior Director for Cyber Policy and Deputy Assistant to the President - Josh describes his White House role coordinating cyber, telecom, supply chain, and cryptocurrency policy. Decade reference: 10+ years - Josh says he wrote about drone assassination dynamics in graduate school about ten years ago.

Pivotal Quotes: "I can't support putting a loose cannon in power as we head into such a potentially pivotal period." — Nathan Leven: Nathan explains why, despite agreement on some AI-adjacent issues, he will vote for Kamala as a rejection of Trump. "If we end up in an AI arms race and we end up seeking strategic advantage over each other, we are going to all lose." — Nathan Leven: Nathan argues that avoiding an arms race with China is the key AI governance objective. "The physics of the environment." — Joshua Steinman: Josh describes his view that great-power competition with China is an objective reality, not a choice.

Implications: The conversation highlights a deep split in AI governance: cooperative restraint versus hard-nosed strategic competition. For listeners, it shows that election choices may shape AI security, chip policy, and U.S.-China escalation far beyond traditional politics.