Interesting piece. Agreed here, and I’d say it’s a framework that applies to most technologies (perhaps a bit controversially in this publication, also to semiconductors). The thing that a country wants to secure is the future dominant technology, not the current generation of technology.
Effectively, diffusion as defined here is a matter of embracing embedding a technology into one’s culture and allowing that to drive it forward. Inherently, that will mean that the current generation can keep getting stolen, but what matters is the next. I think that’s pretty clear for AI, but I’d say that a similar kind of thing is also present in semiconductors (e.g. Japan’s dominance of DRAM in the early 90s with Intel exiting and instead focusing on... CPUs). Its hard to say what the net tech transition is, but it’s fairly certain it will happen, and ideally we don’t protect US/allies tech in such a way that it stagnates and actually pushes China towards disruptive/non-incremental innovations.
Anyway, it’s interesting that you say that regulation is light in China. I’ve been hearing the opposite in my circles in terms of requirements (explicit but often implicit) that the AI outputs have to comply with. That’s been somewhat stifling along with a general wariness in the entire tech ecosystem both to the recent crackdown and current economic conditions.
"Anyway, it’s interesting that you say that regulation is light in China. I’ve been hearing the opposite in my circles in terms of requirements (explicit but often implicit) that the AI outputs have to comply with. That’s been somewhat stifling along with a general wariness in the entire tech ecosystem both to the recent crackdown and current economic conditions."
I've read that consumer models are most affected by China's LLM approval mechanism. It's overseen by the Cyber Security Administration, and mandates model-specific evaluations. Particularly challenging for public-facing, consumer-centric solutions like chatbots. But this won't hold back military innovations!
Interesting piece. Agreed here, and I’d say it’s a framework that applies to most technologies (perhaps a bit controversially in this publication, also to semiconductors). The thing that a country wants to secure is the future dominant technology, not the current generation of technology.
Effectively, diffusion as defined here is a matter of embracing embedding a technology into one’s culture and allowing that to drive it forward. Inherently, that will mean that the current generation can keep getting stolen, but what matters is the next. I think that’s pretty clear for AI, but I’d say that a similar kind of thing is also present in semiconductors (e.g. Japan’s dominance of DRAM in the early 90s with Intel exiting and instead focusing on... CPUs). Its hard to say what the net tech transition is, but it’s fairly certain it will happen, and ideally we don’t protect US/allies tech in such a way that it stagnates and actually pushes China towards disruptive/non-incremental innovations.
Anyway, it’s interesting that you say that regulation is light in China. I’ve been hearing the opposite in my circles in terms of requirements (explicit but often implicit) that the AI outputs have to comply with. That’s been somewhat stifling along with a general wariness in the entire tech ecosystem both to the recent crackdown and current economic conditions.
"Anyway, it’s interesting that you say that regulation is light in China. I’ve been hearing the opposite in my circles in terms of requirements (explicit but often implicit) that the AI outputs have to comply with. That’s been somewhat stifling along with a general wariness in the entire tech ecosystem both to the recent crackdown and current economic conditions."
I was surprised when you mentioned that!
I've read that consumer models are most affected by China's LLM approval mechanism. It's overseen by the Cyber Security Administration, and mandates model-specific evaluations. Particularly challenging for public-facing, consumer-centric solutions like chatbots. But this won't hold back military innovations!
https://digichina.stanford.edu/work/how-will-chinas-generative-ai-regulations-shape-the-future-a-digichina-forum/