The ultimate goal for US-China relations should not be to "win" a zero-sum AI race, but to achieve a "Pax Robotica"—a future of shared peace and abundance. This reframes the relationship from competition to collaboration, avoiding the creation of new AI-powered existential risks.
Viewing China's rise in AI as a recent surge is shortsighted. For most of history, China has been a leading global civilization. Its recent period of technological lag was an anomaly, making its current strength a predictable reversion to the historical mean, not a new phenomenon.
China's national strategy is deeply rooted in its "century of humiliation," where technological inferiority led to foreign domination. This historical trauma drives its non-negotiable goal to never again be reliant on foreign powers for critical technologies like AI and semiconductors.
From the Chinese perspective, US semiconductor export controls are not just a competitive tactic but an attempt to enforce technological dependency. This directly triggers deep-seated fears from their "century of humiliation," strengthening their resolve to achieve complete self-sufficiency at any cost.
The greatest risk in US-China tech competition is not one side getting ahead, but their technology stacks becoming mutually unintelligible. This lack of legibility creates a volatile environment ripe for catastrophic miscalculation and conflict, making a shared AI paradigm a strategic necessity for peace.
By consistently releasing powerful open-weight AI models and publishing research in English, China is strategically positioning itself as the leading provider of global AI public goods. This soft power play builds international goodwill and directly challenges the narrative of a closed-off, extractive nation.
The Chinese Communist Party's popular support, its modern "Mandate of Heaven," is built on its proven ability to deliver stability, safety, and prosperity. Western assumptions that the populace secretly craves democracy overlook this deeply rooted, performance-based social contract.
Instead of trying to replicate China's top-down industrial policy, the US should forge its own path by developing "state capacity with American characteristics." This means focusing on projects that align with US culture, such as deploying autonomous vehicles rather than building a national high-speed rail network.
The demographic reality of the one-child policy has created a potent, often-overlooked deterrent to war for China. With many families having only one child or grandchild, the political and social cost of military casualties is exceptionally high, making large-scale conflict a massive gamble for the government.
China views US concerns over its "civil-military fusion" as hypocritical, pointing to America's own long-standing military-industrial complex and Defense Production Act. They see this as a motivated double standard used to justify unfair trade actions, which erodes US credibility and trust.
A significant unintended consequence of US export controls is the creation of a climate of fear that prevents American AI safety organizations from collaborating with Chinese counterparts. This self-imposed barrier hinders global efforts to manage the risks of a technology that affects all of humanity.
China's temporary lag in the large language model race wasn't from a lack of skill, but from a pragmatic business perspective. They viewed early models like GPT-3 as an expensive way to generate "bad poetry" with no clear use case, highlighting a less ideological, more application-focused approach than Silicon Valley.
Instead of a purely defensive posture, the U.S. should pursue a proactive strategy of reciprocal tech transfer. This would involve trading American strengths like semiconductors for Chinese expertise and onshore manufacturing in areas where the U.S. lags, such as batteries, solar, and robotics, effectively recoupling the economies.
Emerging technologies like confidential computing offer a novel path to building trust between rivals. They could allow for mutual verification of AI systems or even sensitive government documents, with an AI providing trusted attestations (e.g., "no backdoor found") without exposing the underlying secret data.