Sihao Huang on the risk that US–China AI competition leads to war

Abstract

This interview examines China’s AI capabilities, governance approaches, and potential for cooperation with Western nations on AI safety. While Chinese AI currently lags 1.5-2 years behind Western capabilities in large language models, China leads in certain domains like computer vision and surveillance technology. The primary bottleneck for Chinese AI development is computational power, largely due to US semiconductor export controls limiting access to advanced chips. Although China is investing heavily in indigenizing its semiconductor supply chain, significant technical and economic challenges remain. Chinese AI regulation focuses primarily on information control rather than safety, though recent policy developments suggest growing attention to AI risks. The Chinese governance landscape is complex, with multiple competing agencies lacking clear hierarchical organization. Despite these challenges, opportunities exist for US-China cooperation on AI governance through mechanisms like track two dialogues and harmonized safety standards. The analysis emphasizes that rather than viewing China as a unitary actor, understanding the complex interplay between national and local policies, competing institutions, and evolving priorities is crucial for effective international coordination on AI development and safety. - AI-generated abstract