Even in the AI sector

they said relevant future technologies

"Their models are as good and require less compute."

In a sense, this is exactly why I think that China will inevitably overtake the US in this field too. Even besides things like Deepseek, which distills LLMs dominant in the US into a more streamlined model, China also seems to be putting more resources into researching AI that doesn't need to start out with mahoosive amounts of data.

In some ways their AI might even be better at this point. I've talked with Claude users complaining that their AI kept skipping steps and I haven't had that problem with GLM 5.