Let's take a step back from the very basis and ask you, Noor, how would you characterize China's approach to AI development? And if we benchmark it against the United States, what, what are the differences?

Yeah, I think that's a really important question.

And before I get to your question, Francesca, um, let me kind of, you know, give a disclaimer that I'm here in my personal capacity as a researcher, and also my view is not reflective of the whole commission.

But certainly the topic, um, is very relevant to, to our congressional audience and also the broader policy community.

I would say upfront that the thesis of this paper is that the US and China are running two fundamentally different theories of how AI leadership should be won.

The US, you know, on the one hand, kind of bets on a singular transformative breakthrough, right? We see a lot of, uh, AI labs are, frontier labs are, are running, um, towards AGI.

On the other hand, China bets on a kind of ubiquitous adoption and iteration, right? This means achieving AI leadership through diffusion, right? Not just one super intelligence model.

How will that change, if it will change China's AI offer in the future?

And your sense of to what extent is China actually innovating versus just building off of or taking from existing US models?

me just clarify real quick that um you know the 80 number that i mentioned is um an overall vc um investment um not just you know ai but ai is certainly a growing um it's growing share of china's investment overall and to your question about you know whether um we have seen um um more Efforts, more innovation coming from Chinese AI labs research, more efforts to kind of distill or generate outputs using US models to train their own models.

I think that's an important and complicated issue.

It is also constantly evolving as well.

I think on distillation in particular, It's useful to think about what we know and what we don't know.

What we know is that knowledge distillation is a common technique.

It's also a legitimate training method as well among researchers in the AI community.

Could you just unpack this a bit for our folks who may or may not have time to read your paper? What is the key takeaway and how do these two loops drive China's ecosystem?