Curious what people think about China’s actual AI capabilities. It is clear some of these models, even if they are very capable, are partially trained on synthetic data from smarter OpenAI/Anthropic models. If the US paces the frontier, what does this mean for China’s capabilities? I could see a couple scenarios. (For the sake of brevity, I address Moonshot/Deepseek/Alibaba/Z .ai together as Chinese labs, let’s assume we discuss the single most capable Chinese AI lab) 1) Most of China’s advances were from targeted distillation, so a major slowdown in American models would also result in a major slowdown in Chinese models. 2) China has extremely capable AI labs, they just started later and thus are still in the process of catching up. In this case, it’s only maybe 6 months in a pause or 12 months in a slowdown before Chinese models catch up to and surpass American models. If (1) is true, the US unilaterally pacing the frontier will work out fine, and we don’t need buy-in from China. If (2), however, we really need buy-in otherwise China will just surpass and accelerate faster. I’m not super well informed but my understanding places things closer to (1) than (2) currently. Curious to hear any other thoughts.