US-China Technology Competition: Insights from Experts on AI and Supply Chains

Uh, okay, good afternoon, everyone. Welcome to CSIS. And for uh this event hosted by the

geopolitics and foreign policy department at CSIS. I'm Victor Cha, president of the department, uh Korea

chair and professor at Georgetown. And um today we're going to be talking about US-China

relations and US-China technology in particular, technology competition. Um uh we're very fortunate today to be

able to have this conversation partnering with the Yale Jackson School. Um my good friend uh Ted Wittenstein

uh has brought a group of wonderful scholars and students down from New Haven um to spend the

morning with us to talk a little bit about a recent trip that they went to China um and also to be in dialogue with

our scholars who are working on these issues. So, it's really been an excellent day

and um we're looking forward to uh sharing some of this conversation with you. I'm going to ask Ted to come up and

say a few opening remarks. Um and I should also say that um my colleague uh Henrietta Levin who is a senior fellow

in the Freeman chair in China studies uh just came off a stint at the State Department as the number two

uh for um uh China House at the State Department will be the moderator and facilitator of the

discussion. Just because I know Henry, you're you're going to be introducing everybody else and nobody's introducing

you. So, I thought I should do that courtesy. But let me uh give my uh colleague Ted Wittenstein who is senior

lecturer in global affairs and director of the Schmidt program on artificial intelligence, emerging tech, and

national power at the Yale Jackson School to come up and make a few remarks.

Well, uh thank you, Victor. Uh and thanks for all the work uh that you do in connecting the policy and

practitioner and academic communities. Uh the work that you've done here at CSIS is is really extraordinary and

there are a lot of parallels to how we think about uh the value of our students at our own Yale Jackson School. So, how

do we develop both a theoretical academic grounding in global affairs combined with those practitioner

insights that can be so useful in informing the policy prospect process and you've really done an outstanding

work in bridging that divide here at CSIS as well as your teaching and academic leadership at Georgetown as

well. At the Yale Jackson School, we are, you know, training the next generation of

global affairs leaders broadly defined. And when we say bridging the divide, particularly in our Schmidt program at

Yale, our AI emerging tech security program, we're talking about how do you connect different components uh of

students studying AI and emerging tech challenges from different vantage points. So, how do you bring together

the law, technology, policy, business, ethics communities that are examining how AI is having a transformational

impact on the international system. And we have a pretty broad-based uh academic research and teaching program at Yale

focused in this area uh with the support of Eric and Wendy Schmidt with for which we're very grateful. And so, the key

part here is how do we appreciate the extent to which these technological changes are truly revolutionary,

upending conceptions of world order entirely versus are they evolutionary? Are they still incremental changes that

are accelerating pre-existing trends in the international arena? And obviously US technology competition uh is a core

focus of our program as we think about the impact that AI has had as an accelerant in a lot of the challenges

that we find ourselves in the in the moment on US-China relations. And in particular, our Yale-Renmin dialogue on

AI, emerging tech, and US-China relations is now in its fourth year of a peer-to-peer exchange between both

faculty, practitioners, and students that bring our uh top minds uh I both to China as well as hosting a group on the

Yale campus uh at a time when peer-to-peer exchange is becoming very difficult as it relates to US-China

relations. Uh and this is something that we acknowledge and we engage in the sort of

uh spirit of competition and debate and dialogue and hopefully mutual understanding in ways that can uh

avoid miscalculation and inadvertent conflict uh at a particularly volatile moment. The exchange has in its roots uh

Dr. Henry Kissinger's uh late-in-life diplomacy focused on sort of thinking about AI and emerging tech as it relates

to sort of traditional challenges in US-China relations as well as Dr. Eric Schmidt and the work that he's done in

supporting uh the work that we've done. So, we're just back actually. So, I was with a delegation of four uh faculty and

practitioner colleagues along with 25 students. We were in Beijing, Suzhou, and Hangzhou uh between the week of

March 7th and March 14th, which included um both dialogue with uh Chinese

counterparts and uh former senior officials in Beijing, uh student-to-student uh debate on the wide

range of AI and tech uh and geopolitical related challenges, as well as traveling south to to Hangzhou where the six

little dragons uh of uh so-called of Hangzhou have made uh pretty enormous advances in uh robotics and AI and ways

that uh have caught the attention of US policymakers and the global conversation. So, I'm hopeful that we

can share some of the insights. Uh our students have had a really interesting exchange uh this morning uh hosted by

Dr. Cha and a number of his colleagues. So, very grateful to him uh and Henrietta and company for hosting

us and facilitating dialogue that can magnify the uh benefits that we have from

exchanges such as these. So, we do live in the moment where there is very limited uh US-China interaction at both

official and non-official levels. And so, when it does occur, we try to magnify the impact of what we're doing

and I think CSIS is the terrific place for convening a conversation such as this. So, I'm

really looking forward to the conversation. Thank you very much. >> [applause]

[clears throat] >> Great. Um well, thank you, Victor and Ted for

the opening remarks. Uh again, my name is Henrietta Levin. I'm a senior fellow here at CSIS with the Freeman chair in

China studies. I'm very honored to be moderating today's panel discussion uh joined by a very distinguished group of

experts and and leaders. Uh so, I'm particularly excited for the conversation because as Ted mentioned,

he is newly back from a I think very unique and insightful dialogue uh at the practitioner level uh in China. But but

the dialogue among the students as well, thing that we'll really want to dive into in a period of time where

people-to-people ties between the US and China have become so attenuated. So, um let me start by introducing our

fantastic slate of panelists. Again, we have Ted Ted Wittenstein, senior lecturer in global affairs and director

of the Schmidt program on artificial intelligence, emerging technologies, and national power at the Yale University

Jackson School of Global Affairs. We have Jason Hsu, former legislator at large in Taiwan's Legislative Yuan, and

senior fellow at the Hudson Institute, and my CSIS colleague Kate Corcoran who is deputy director of the economics

program and Scholl chair in international business again here at uh CSIS and a uh until recently a senior

official at the the Commerce Department. Um so, with that, I'm hoping we can start the conversation with brief

opening remarks from each of our panelists. And Ted, hoping you can start off with the key takeaways, your

perspectives coming out of this very unique experience and dialogue in China. Great. Uh well, thank you so much, uh

Henrietta, and I'm happy to share uh just a few key takeaways from this exchange that we had in Beijing, Suzhou,

and Hangzhou in hopes that we can just have an interactive conversation among us as well as uh comments from all of

you in the audience. The timing of the trip, of course, was not something you can always prepare

for. So, it occurred uh in the context of uh the US initiating the conflict uh in Iran as well as

you know, questions about the future of any leader uh summit that might occur between Presidents Trump and Xi. There

was a sense at the time of the dialogue that the summit would be occurring at the end of March, just a week or two

after we were there. As many of you know, that dialogue was since postponed given the war in Iran.

Uh although as recently as this morning, some indications that that uh leaders summit has now been uh rescheduled uh to

take place in in May 13th or 14th. So, not too much of a postponement, which I think is important uh in terms of

head-of-state uh interaction uh that is likely to be more consistent over the coming year. So, Presidents Trump and Xi

are likely to have as many as four meetings uh over the course of this year, not just uh with President Trump

traveling to China in May, but a reciprocal visit by President Xi to the United States at sometime in the coming

year. Add to that the Apex Summit and the G20. There there likely going to be a number

of head of state interactions that might bring some level of stability to what's been a a volatile

period in US-China relations. I think the question of AI and emerging tech in some ways is secondary to the

high-level question of where the overall relationship is headed in terms of the economic level. So, I think one key

takeaway is if there is going to be some deal on tariffs or trade or export controls in the US and China, I think

that might allow for more interactive exchange on some of the AI tech challenges that I am going to flag for

you in just a moment. So, we'll have to see what happens as we approach this leaders summit. The Chinese are

definitely of the mind of trying to portray the US and China as the two premier

world powers deciding the most consequential challenges of the 21st century. So, whether you want to call

that some sort of G2 or framing of this, this is what I think China is really seeking to gain by these

head of state interactions at the US presidential level and trying to set an agenda for exchange that might bring

some level of consistency and continued exchange at staff and other levels of exchange to include the track two

dialogues and other things that I've been a part of. The dialogue focuses on how AI and

emerging tech is impacting US-China relations and I'll just maybe highlight three aspects of it that I think are

important and maybe sometimes lost in both the academic conversation on the Yale University campus as well as the

policy conversation in DC. The first is that we are rapidly moving towards a world of what you will call

embodied AI. So, we tend to focus on the greatest large language models, how advanced our chat GPTs or deep seeks are

as they relate to US-China competition, but the next version of US-China competition is already rapidly advanced

towards what you would call embodied AI. So, the fusion of language models with computer vision and robotics. And here

China really has some distinct comparative advantages. Our group did have an opportunity to tour

Unitree Robotics as well as Deep Robotics in Hangzhou. And these companies are

peer-to-peer capable with the United States. And even if they don't have the most advanced

large language models or chips, what China does have is a organizational integration advantage in how it could

fuse some of these technologies together. And with the economic benefit of robotics as it relates to industrial

processes as well as the dual-use military or intelligence counterparts that you might see continue to develop

in the relationship. The other part that I would also want to highlight is the conversation on AI and

emerging tech with US-China peers. Again, we've done this dialogue now for 4 years. It may have originally been a

conversation that looked at which language model is more capable or something like that, right? Like does

the US model is chat GPT that much better than deep seek or is the frontier closed models better than the

open-source Chinese models? This is not an inconsequential question, but what we've learned in the US-China exchanges

is that the whole suite of what you might call the AI stack, it's how you integrate these model capabilities into

a range of energy, infrastructure, applications, talent. And in this area China also may have some comparative

advantages relative to the United States that I think are worth thinking about and reflecting on. So, how do you

develop a whole AI stack? The energy and infrastructure needed to power AI, the critical minerals needed to build these

advanced systems, very important in robotics for instance. And the last piece I would just note

because obviously the war in Iran very much hung over some of the conversations that we've had with

Chinese peers. I believe China sees in the current conflict in the Middle East an opportunity to

position itself as a more responsible or predictable leader with respect to AI as it relates to

defense and military applications. And this of course is somewhat more of a PR diplomatic positioning than it's the

reality of how the CCP has engaged in a wide range of very concerning AI uses for its own surveillance and military

modernization, but it sees in the US use of military AI tools in Venezuela and Iran an opportunity for China to sort of

adopt the mantra of being a responsible stakeholder when it comes to the utilization and deployment of these

tools. So, this is going to be a broader diplomatic messaging challenge that's going to have to be confronted over

time. Very grateful to our Renmin University hosts for engaging with us on on a very serious

complex issues. And again, there's a student component here that I think is based on the premise that this is going

to be a a generational challenge. And Dr. Kissinger used to say in the last years of his life that he thought as he

looked at questions of AI and cyber and emerging tech that the United States and China were in many ways back in 1974.

So, it's going to take a long period of time to develop really even a common understanding of basic terms and

concepts, let alone trying to define guardrails and norms of behavior. So, we have to start somewhere and I'm grateful

to my own Yale students, some of whom are in the audience today for an opportunity to engage in that exchange.

So, thank you for inviting me to be here, Henrietta. Thank you. Turning to Jason next. Of course, when it comes to

US-China competition in the technology domain and the rapid acceleration of Chinese

capabilities that Ted has just spoke to, of course, Taiwan has a very strong interest commercially and from a

security perspective in in all of this. So, so what would you add from the Taiwan perspective to the discussion?

Absolutely. Well, thank you for for having me here. Just kind of build on what Ted has sort of laid out as to how

the landscape has changed and evolved and how Taiwan fits in with sort of the challenges and also some of the

crisis that are emerging. I think we need to look at the current competition between US and China has evolved from

innovation to industrial. And this has marked a significant deviation from how the world views tech and in terms of how

the competition and choke points are happening. Then the the biggest, probably most distinct point to to

really look at is is during the previous administration the ability to create strategic advantage is

by creating the chip denial. So, the single chip denial of high-end chips is able to was able to slow China down. But

then the competition has become the system-wide. Basically, ruling what Ted has said, it's going

from one single chip to full stack. China is now looking at chips to clusters to data centers to operations.

And then they measure each scale and each category differently. By very smartly and intelligently creating

specific choke points for the years to come. You know, one good example is magnet and and

critical minerals. And and and if you look at dissecting this whole tech stack, you you will ask how would US

adjust these potential vulnerabilities in the next few years to come and how does US work with partners and allies to

create a ecosystem so that it's strong enough to sustain the choke points that are coming

to happen. And I would argue that what China is really good at is to operationalize the compute at national

scale. And that has not happened in the past when a single chip denial gives US advantage. And for China, China

has moved beyond a single chip disability. And what they are looking at is how in what areas China can

accelerate and excel and then speed up and to get past US in a very quiet and somewhat rerouting the some of the

ban or prohibitions that US imposed on China. So, looking at the robotics and also EV, the

batteries, and also some other parts that China so focused on applicate applications

is very telling on where we should be laying out our competition lens.

So, the the next thing that I wanted to to talk about is the concept of a completely decoupling

is nearly impossible if you look at how the tech stack is structured and organized and supply chain is is also

functioning. That has created also potential loopholes and the vulnerability where China can come in

and create selective interdependence. And [snorts] this goes back to what I just mentioned that China selectively

choose strategic sectors to focus and then to invest and to accelerate and then to build exponential exponential

growth without, you know, being known to the outside world. So, we need to look at that and we need to look at how China

treats speed, which is more important than spirituality. Uh the the race to with AI uh US is leading in the frontier

model large language model, but China is so good at application layer deployment at scale. And to me, that's alarming and

that's very worrying when, you know, US uh tech startups using Chinese open weight uh

uh language models to train their LLM. So, to me, that that's something we need to

look at. And lastly, I will talk about Taiwan um as a part of the the system integrator now becoming a a choke point

in the in the AI competition era. Silicon Shield, as many of you know and also uh experts sitting on this stage

would understand that Silicon Shield itself is not sufficient for Taiwan anymore. It used to be that Taiwan

produces the chips that China want and also US want and that would deter uh China's aggression on uh Taiwan. But, I

would argue that China now is able to reroute around Taiwan and then to also reroute around the export control and

the bans on this uh advanced chips. And so, we need to really look at what areas that would become potential

vulnerability. And I would argue that two ma- uh important areas we need to be uh paying attention to. One is a

redundancy. Right now, if all of this, you know, tech stack or or ecosystem is creating single point of failure,

what if something breaks in the supply chain, it becomes a system-wide crisis. The other is diversification.

TSMC is investing in the US very heavily. They announced another 100 billion to build in the US. They've

purchased enough land to build six fabs across Arizona and they've announced their third fab in Kumamoto in Japan.

So, this global expansion is very important and it also it's also creating a new cluster of fabs. Whereas China

sort of structured itself around the area and application there they are very good at. And now US is looking at where

they could develop a solid and sound uh chip delivery system. And I would say that um it's very important to ensure

this diversification is successful and can deliver at a speed with which that um the replenishment and as well as the

industry that is in need of that the chips are um uh increasingly dependent on this uh one

company. Uh lastly, uh 90% of chips produced by TSMC are advanced uh chips 2 nanometers and TSMC is in- investing and

innovating, you know, sub-2 nanometers. 70% of the world's AI server are created by Taiwanese tech companies. So, we

cannot allow this system to fail, otherwise we are we will be basically blinding ourselves for the sight of the

whole picture by only focusing on one area that our competitor and opponent is basically by looking or bypassing and

then creating other areas of weakness for US and its allies. Great. I think a helpful reminder from

you and also from Ted to think increasingly about the technology competition on a systems level, looking

at integration among capabilities rather than the perspective that may have held sway in Washington and even a couple of

years ago focused on more individual components in in the supply chain. Um so, let's turn next to to Kate with all

of your experience and expertise in how the US government has looked at uh protecting its own high-end

technologies. Really welcome your initial thoughts on this competition, on the findings um that Ted saw in in China

and anything else you'd love to to start off with. Thank you. Uh so, thinking of the US, you know,

China competition uh zoom from a zoomed out level, what's striking to me is just how how close uh

how how much China has surged ahead and narrowed the gap in so many areas over the last 15 years and especially the

last 5 years on the semiconductor sector uh both in terms of chips and but also the equipment. Uh the, you know, policy

makers that were looking for leverage points, you know, over China, looking for areas where uh they could, you know,

you know, seek to uh use export controls uh in a meaningful way to constrain China. 15 years ago had a a much uh

wider range of uh items and technologies to choose from than they do today. Uh and I think there's, you know, the

dialogues uh like like Ted is running are great for bringing that perspective of where China really has become a peer

competitor or a near peer competitor or where they are close to becoming one. Uh there's uh there aren't uh a lot of uh

these dialogues happening. Uh there's there aren't many Americans going to China. Uh there aren't many American

policy makers going to China. Uh and that perspective can get lost uh and I think there is still a tendency uh in

some circles to assume that the US must have uh you know, the the the best option or must have, you know, an

advantage where uh there increasingly that that is uh less uh generally the case. Um semiconductor manufacturing is

an interesting uh sort of example that we we look at a lot in the economics program here

uh you know, in terms of the impact of export controls and and and how that played into China's uh increased uh

abilities in that area. Um you know, the 2022 uh export controls uh that were brought both on the chips themselves and

on the manufacturing equipment uh and then were expanded in 2023 and 2024 um you were intended to, you know,

constrain China's ability, you know, in the AI race. Uh and it's uh one of those areas that's

really fascinating to me to look at because you can have very well-informed uh and well-reasoned arguments on both

sides of whether those were effective or not. Uh it depends on what you consider effectiveness. Uh you know, in terms of,

you know, are we ahead still uh in in equipment, uh in leading chips? Yes. Um but in terms of, you know, is China,

you know, gaining on us? Are they nearer? Are they able to uh use AI use AI across the stack, uh which I also

appreciate that perspective, uh is is less clear uh in terms of who is going to actually operationalize this

technology. Um in terms of, you know, having a military advantage, it won't be who has the best chip necessarily. That

won't be what makes your you your, you know, military superior if you have access to the best chip. It'll be how

you are able to implement that and organize that and, you know, intelligentize uh your your military

platforms. Uh so, that's something uh that uh I I appreciate uh from this dialogue. Um

I think I'll leave it at that for now. Great. Um well, I I may uh abuse my privilege as moderator to ask just a

couple of questions to the the panel, but also we want to open up the floor for any questions from the audience that

you would like to pose to the experts on stage. So, please start thinking about whether there is anything you'd like to

introduce into the conversation, but um in the meantime, I just want to start with two questions to hear from from

everyone on the panel. I I think um Jason, you mentioned this uh dynamic that I think is often discussed, at

least here in Washington, of the differing uh differing shape of the AI race when viewed from

the United States or viewed from China. And where in the US uh and certainly from the industry side, there's a sense

that we are racing to AGI. We're racing to the singularity. Um and that whoever achieves that outcome first may secure

some kind of enduring strategic commercial technology advantage. Whereas you see the shape of AI innovation in

China focused much more heavily on application. How can AI be integrated into everything

today. Um but I I think we we often uh maybe lack a a certain concreteness to what

that really means, what it feels like, especially on the ground in China. Um and so, really welcome your thoughts on

how this dynamic is playing out in practice uh especially in industry. And then just a second question as well. Um

so, I I think from all of you, there's a strong sense that the pace of innovation, especially in technology

relevant to AI, is perhaps exceeding uh the pace of our own strategic frameworks for thinking about uh protection of

those very technologies. And so, if the conversation was previously uh and still to some degree, but maybe more

exclusively in the past focused on the very high-end semiconductor capabilities um and the technology used to produce

them, maybe now there's just starting to be more of a conversation about embodied AI.

Um and Ted would love to hear more about how you're seeing China's focus in that industrial sector on the ground, but

then also what comes next. Um you know, do you have a sense of the

broader space in which China is looking to establish uh new points of leadership, new choke points.

Jason, I think you you mentioned you know the quiet acceleration past the US that China may be seeking. You know,

what what sectors do we expect them to focus on next and and do we think that will succeed? So, I'll just push all of

that back to our our panelists and Ted, why don't you start off? Uh sure. Well, those are

three excellent questions, each of which we could spend a lot of time focusing on, but I'll just comment on a few

points and Jason should also certainly chime in given his expertise in thinking about Taiwan's own role in the AI

innovation ecosystem. I think on the ground in China there's a sense that

as well as an anxiety about can we integrate the tool for the benefit of our economic domestic population. That

that is essentially the core challenge and my students had a lot of very interesting exchanges with their peers

about their anxiety about the economic outlook and situation overall in China. And so the the pathways for what it will

mean to advance within industry and fields of endeavor in China remains an undecided question. So, there are some

very promising companies that look to be harnessing the potential, but can they really push these tools out within their

own population? I think it's going to be the first challenge and that will affect sort of public sentiment about our

advances in AI and robotics sort of a net positive or negative. The public opinion data suggests that China in

general, Chinese citizens have a more favorable view towards AI and emerging tech than you've seen take root in the

United States. And that's because China is trying to democratize and push out the value

that they see in the future of these tools. And of course, they have an information ecosystem that's centrally

controlled and a variety of ways of messaging their audiences on this question.

So, you see a big divergence still. Guangzhou is a bustling robotics tech hub, but can it really expand out beyond

that domestically within other core provinces and industries remains to be seen. The second piece

will be whether these Chinese companies have an ability beyond China to create sort of marketplace and

adoption and integration into global systems and processes. So, China is really banking on an ability

to be the sort of AI robotics supplier in the global south block of nations over the course of the coming next

decade or two. And so, what advantage will that truly have economically as well as will it be fully adopted and can

you kind of create a Belt and Road investment initiative that really encourages or similarly an adoption of

these tools over time, I think it's going to still be a significant challenge. And can these tools push into

western European or even American markets in the context of export controls or dual use

military or security concerns? Unitary, for instance, is positioning itself to be

the next Huawei or ByteDance TikTok when it comes to robotics. They're going to push out relatively inexpensive humanoid

or quad robot tools that can be fashioned in a lot of ways and adopted using Nvidia chips as its backbone,

but still potentially harnessing data and having a relationship with its own parent

government minders in a way that will continue to raise alarms and red flags in Washington as well as in a number of

other capitals. So, what comes next? I think we have exceeded a lot of the existing policy

frameworks already when it comes to embodied AI. Just as an example, the supply chain

questions, right? So, in robotics, the supply chain challenge, the choke point is not the chip. It doesn't have

to be at the moment an extremely high-end export banned chip and Nvidia low-end

chips are powering a lot of the advances that you've seen take place in

the six little dragons of of Hangzhou. And I think the challenge here is going to be

uh is the choke point is it better off to be in this AI innovation ecosystem in

China? And that's Nvidia's position. It's better to be in the system and have a US influence in the system than to try

to restrict it entirely. And this is going to have to play out. The other piece is just worth focusing

on on the critical minerals. So, the robotics supply chain is very different. So, you need heavy magnets, you need

motors powered by these heavy magnets, you need tools to make those heavy motors.

There are many pounds of heavy magnets in a humanoid robot that we were examining. And so, because of

that, China has a very significant advantage on that end of the spectrum. Now, there are companies like MP

Materials in the United States that are doing great work to try to build a vertical integration supply chain, but

even then on some of these critical minerals, it's it's nowhere near the capacity

currently. So, we don't really know who's going to win these competition challenges because

of the comparative advantages that go on both sides. And so, this is I think a reason why the US policy framework in

the Trump administration has been to try to unleash as much as possible the innovation breakthrough potential in

Silicon Valley by removing sort of perceived regulatory barriers because this is something that China can't do,

right? We we do not US can do sort of a breakthrough frontier capability on this question.

And China will try its best to mimic that. But, you know, even if China has a 75% as good capability, if across the AI

stack they have an organizational integration ability that the United States doesn't have, then the 75%

capability may be just as good. So, this is how we've been thinking about these issues and teaching them in our Yale

students and I think in the policy discussion it needs a much more nuanced examination of the competition dynamics

across all these different stack elements that I mentioned. And you you mentioned that some of the

most advanced Chinese embodied AI capabilities are running on Nvidia chips, US designed chips. But, to be

clear, that we're not talking about the semiconductors that are subject to US export controls. I assume we're not

even talking about H20s, you know, higher end chips that are newly authorized. These are Yes. lower edge

capabilities. >> Yes. And we may ultimately reach a time, you know, three or five years

from now where the high end piece in a very sophisticated robotic tool could be a choke point advantage. That's

not what we're seeing at the moment with advances in humanoid robotics for China. So, it's a risk that the United States

really will fall behind the edge here because even with a frontier capability, it may not

be as readily adopted or integrated. And this is going to be a significant challenge as we think about supply

chains or other pressure points against China in this regard. One last thing I'll add about robotics,

you know, people said data is the new oil. This was the rage at the time in AI, you know, maybe seven or eight years

ago at the sort of beginning of our understanding of these technologies. In robotics, there is very limited data to

power robotics relative to large language models. So, the reason why large language models have advanced is

cuz global internet data can be sort of harvested, tagged, and coded. That is not true for a lot of vision movement

types of data. And China has huge amounts of that data because of its internal surveillance state that the

United States does not have. So, China will have again an ability to harness motion human interaction data to power

robotics in a way that will be harder for the United States to replicate. So, it you know,

data is not the new oil, but robotics data is going to be important. And what we

were just saying now, of course, three years ago we were saying chips are the new oil, right? But now as we're talking

about the robotics, we're saying, well, maybe the chips are not the new oil. And as I've been thinking about the

Strait of Hormuz, I actually think oil is the new oil. So, we're back in a situation where I think

we've come full circle to appreciate the energy infrastructure demands required to power AI. And if this is something

where China can outcompete the United States, we're in a lot of trouble. So, I am hopeful that US policy makers will be

seized with this as we think about our own industrial policy as it relates to competition. And we could have a whole

other conversation on China's dominance of new energy technologies and how that complicates the US competitive position.

Jason, what what would you add? Yeah, again, I I would look at this question from the system perspective. It is true

that China's AGI is slowed down by with a lack of advanced chips, but there are really four critical components to

investigating where the system creates vulnerability. Number one is infrastructure piece.

Number two is the diffusion and deployment. Number three is fusion of hardware and software.

And number four is the weaponization of standard. You know, with respect to

infrastructure, we need to ask number one, without advanced compute,

what China is doing to advance their AI, if they are not doing AGI, but what areas that they are accelerated? Can

China train their model with good enough chips? Yes, they can. Can China just stack up the you know, lower end chips

and then train the model? Yes, they can. Um Is China building data centers? Yes,

they do they are. And where are they building it? Do we have the data where they are building? Are they building it

outside of China? Yes, they do. Are they building these cloud data centers in the China friendly countries or dependent

countries or and then using the compute to reroute their compute capability back to China and with the access to high-end

chips, they might. And in fact, I'm sure they they are. And number three is, you know, with the energy

being the biggest choke point in AI infrastructure and compute development, is China having the advantage of almost

free or cheap or unlimited energy? I think they do. So, if you look at this, you really realize the infrastructure

layer that US, you know, is very likely to fall behind because of China's ability to basically bypass all the

barriers that US created, but now move the competition to the system. Secondly, diffusion and deployment. I think in the

today's AI world, it depends on how fast and how pervasive the diffusion can be with the models at at a at a

fingertips. So, if China is able to push their model through influencing, through using the small models, but every

household in the Chinese influenced countries and territories are using Chinese white label um um AI apps. And

then we are we will be looking at a world where not only is bifurcated in technology, but the narratives and also

the content and as well as um the um the AI generated um feedback is drastically going to be dangerous. So,

that diffusion rule and as well as the diffusion race, I feel that US need to catch up on as well. Number three is the

fusion of hardware and software that Ted has mentioned. Taiwan is so good at hardware. We build, you know, AI

servers, you know, companies such as Foxconn makes almost every iPhone in your pocket

and Mac MacBook laptops and etc. But you know, we've come to the age where what's happening inside your device

itself is a universe of compute and as well as algorithm, as well as the potential AI enabled um actions and

results. So, we're we're looking at the on-premise compute, on-prem edge computing. These are the type of things

that small AI device or AI enabled device such as, you know, cell driverless cars, autonomous vehicles, um

your phones that would essentially become a node that itself would be malleable to the malicious AI. So, that

that's to me is dangerous when US looking at reindustrialization, but then China is focusing on

fusing the hardware and software and creating, you know, it's essentially a swarm of AI enabled technology products

that can propagate around the world. And and and and then last piece I feel is the most critical piece is

that China is now weaponizing. It's a It's a standards of AI. Um Beijing creates the world's AI standard

council, basically looking at what rules or what developments, parameters, alignments that need to be

included in their own universe, in a universe that China is going to dominate. And that itself will coalesce

by around the countries that would be subjected to to to China's influence. So, these are these four key points I

think would be the deciding factor um the race plays out. And also, if we look at, you know, how long this race would

have some sort of end state or is the AGI the end state? Is the singularity moment end state? I argue none of those

because we're looking at three years out, five years out, 10 years out, something will happen in this area in

one single category or device being medical or military use AI or anything that would create such a big um areas of

a of a weakness that China will exploit. And we need to look at this from a system point of view.

And on standards in particular, cuz I I agree that that can be such a potent tool to lock in strategic or commercial

advantage, especially in these emerging fields. It How can the US catch up? Yeah, that's

a really good question. I I think what pre the Biden administration tried to create the the the chip alliance is is

brilliant in in terms of thinking, not only to structure and organize the supply chain and but also think about,

you know, working with allies to create more comprehensive and cohesive alliance. And not only to make sure that

US can have access to high-end chips, but also to think about how the US companies

and as well as allies can compete commercially with China. I think US needs to lower the barrier to compete

and create commercial interest. So, companies that are following the US export ban now lose the market in China

or competing with their Chinese competitors. You know, for example, the semiconductor

equipment makers is probably the one that suffers the most because now China have this homegrown indigenous

semiconductor equipment makers that are going to shipping out of China but and also propagating around the countries

that can start making um chips that will be used or by China or with China. And and so, I think

you know, by working with allies to create that alliance both on the diffusion and deployment as well as the

fusion of hardware and software is very important. Okay, what would you

Yeah, just a few thoughts. So, somewhere I see China really taking advantage and sort of turning potential weaknesses

into strengths is just the use of feedback loops. So, you know, how much of their you know, they've really leaned

into open source software and open source models, which you know, really encourages adoption, encourages

development. Um you know, in the the closest sort of comparison you would have you know, in the leading US model

Llama, you know, if you reach a large enough number of customers, you know, then you start to have to you know, have

a license and have to pay. Um so, it's it's it's really you know, not sort of not really the same approach. It'll be

interesting to see as they develop if they stick with you know, this open source. Um

My my sense is they will because the advantages they're gaining from it despite the potential commercial loss

are so large in terms of what they're getting in terms of that you know, group development, group buy-in, group

feedback loops. Um the same thing's happening with the semiconductor manufacturing equipment. Um You know,

the the huge growth there you know, in the past you know, few years, I think it came in large part from mandates to have

you know, at least 50% of their equipment be you know, domestic Chinese equipment, which is you know, really

requiring that they use equipment that is sometimes lesser, known to not be as good as what they could obtain

you know, from the US or other you know, from other nations. But because they have to use that equipment and that

equipment gets you know, tested, added to lines, gets you know, all of that experience, that in turn builds that

equipment. That helps those manufacturers, you know, Narenco, really get to where they are and catch up. Um

so, I looking and kind of expecting to see those similar types of you know,

mandates across other sectors that will require use of Chinese versions even where they are less capable as a way to

to bridge those gaps. Um on the US side, I'm looking to see kind of somewhat expecting to see more

leaning in on our sort of import controls through things like the ICTS authority. Um which you know, the

biggest sort of I one of the biggest rules through the ICTS authority is called the connected vehicles rule at

least colloquially, which is you know, one of one of the reasons you know, why you can't buy a BYD vehicle in the

United States. Um but as we are moving from a place where we are, you know, firmly in the lead, where you would

where you might expect to use an export control, we might see more shift on trying to stop those feedback loops by

by trying to stop things like having Unitree Robotics you know, widely you know, in use across

the United States where they would be you building off of that data to to accelerate.

Um and so, I I want to contextualize the conversation within the US-China

relationship as well. And and Ted, you started off with a few comments on the the summit

between President Trump and President Xi that was supposed to be happening in about a week. Now looks like it will be

in a few weeks mid-May. It is clearly a meeting that both sides have prioritized very highly.

You know, I think there was initially questions about whether the Iran war might cause China to feel uncomfortable

having such a high-profile engagement. But clearly that is not the case. Clearly China views the summit as an

incredibly important opportunity I think to continue stabilizing the US-China relationship on what China sees as their

terms coming out of Busan. And potentially to press for additional concessions from the United States as

well on more of a strategic level as the US policy goals appear to be more economically focused. And so in that

context, I wonder what what all of the panelists think about what China is trying to get from the US on technology

at this stage. And so Ted, maybe you could start off with observations from your your

dialogue. Sure. Well, I'm glad this leaders summit actually is now scheduled and not

continuously postponed. Of course, we don't know until it actually happens given the the volatility in the

relationship. I do think it's very important for the United States and China at a leader level to remain

engaged despite all the differences. And so even with the Iran conflict which the United States and China have very

different views at the moment. The two sides really need to demonstrate I think globally that despite even some

highly contentious even kinetic action among our partners in the world that we're committed to dialogue

and trying to manage areas of disagreement so that they don't spiral into miscalculation or inadvertent

conflict. What does China want to gain? I think part of it is really just the global presentation and stature.

And I think that that really cannot be underrated on the Chinese side. It's very important in their system

for President Xi to be viewed on par with President Trump as sort of the

world's two premier leaders solving global challenges. That's a core part

of China's own information ecosystem and messaging about sort of China's ascendance to the world stage. So in

some level even meeting four times in this kind of high-level way over the course of the year and everyone talking

about the meeting and the goal of the meeting is very much

playing to how President Xi wants to project strength in China's role in the world.

In terms of what they would specifically like to get out of this, I think trying to bring some level of stability

on the question of the tariffs and the trade war is going to be paramount.

So hopefully a lowering of the barriers and a continuation if not a permanent continuation of some of the pauses that

went into effect following some of the more recent retaliatory measures. Overall, the economic stability

will be the core message from which the technology will flow. So I do view the AI technology piece as sort of secondary

to whether or not the economic overall tariff trade war aspect can be paused. What China certainly seeks is for its

technology companies to have access to global markets and not be

constrained by US efforts to hinder their development as you had seen in the chip export controls. So they are eager

for their robotics and other high-tech industries to continue to seek non-Chinese partners in

the world. And even if those aren't in the United States, China would likely desire to see a world where the United

States is not actively thwarting its efforts to deploy or develop technology

relationships in the global south as well as even with some potential European partners.

The dual-use military and security issues are real. So I I don't want to downplay those, right? So for every

unitary robotics that is sort of a public-facing generally

and entertainment or research-oriented robot, you can envision

companies that are through China's military-civil fusion deploying these tools for their own

surveillance and military modernization and desire to to you know, continue to constrain if not

move even more aggressively towards Taiwan. So these challenges are real. It's going to

be very difficult I think for China to overcome those risks and perceptions because it's really an

issue of mistrust. So China is not transparent about how its private companies interact

with its government and security apparatus. China is not transparent in terms of how its own

domestic surveillance and and centralized control priorities lead its companies to

support a wide range of policies that I think a lot of US citizens would be very uncomfortable with their data

or other aspects being utilized. So because of this, I think you're likely

to have more of a pause. Secretary Rubio recently said about the summit he was hoping to bring strategic

stability to the US-China relationship. I think something more akin to a strategic pause is likely going to be

part of the goal here trying to buy a year of interaction while the United States focuses on other priorities in

its own hemisphere as well as in the Middle East. And from China's perspective, of course, this is

going to be a real challenge for the United States if America is focused again in the Middle East for an

extended period of time or in our hemisphere in a way that allows China to again kind of thwart any

type of pivot to the Indo-Pacific region. And China is no doubt watching as US

military personnel and assets are are moved out of Korea and Japan. So from the Chinese perspective buying a year of

strategic pause to assess and then to sort of see if the US is really going to truly pivot from a

security standpoint to the Indo-Pacific or not is likely in China's advantage as it thinks about these types of

head of state commitments over the coming year. Jason.

I I think the Trump-Xi summit would be more of a symbolic meeting than substantive. I don't think

anything substantive will come out of this meeting more than more than just you know, a symbolic handshake and just

to make sure that we meet and things you know, cool down a bit. But if this meeting were to happen 6 months ago or

year ago, I would argue that Xi would want this meeting more much more than Trump did. And I would say

that the Beijing used to be very anxious about

decoupling and the recoupling. Now they are not so anxious anymore. Now if you look at the the AI race and also the the

chips control, Trump approved of the H200 chips to sell to China. Look at how many

China or Beijing has placed order. Not that many. And so it is clear that China is now trying to foster their homegrown

champions whether it is Huawei or other other companies that are focusing on the AI applications. And we also need to be

very clear that I am the dual-use components that these companies are doing. They are

facade is a commercial application, but then the real use is find this robots are deployed in battlefield and what

type of capabilities they are capable of doing. In Trump-Xi previous meeting, I know

former National Security Advisor Jake Sullivan worked hard on getting Trump and getting sorry, getting Biden

President Biden and President Xi to agree on none of the AI nuclear AI enabled mechanism can be activated

through unilateral or as as we develop these technologies. I I hope President Trump can

also reinforce this point because I think as we evolve and develop this technology, somewhere some point in some

area will be a black box that we don't know and we don't know how to reverse. And then more than anything else,

I think what um President Xi is trying to gain from this

meeting with President Trump is trying to get some sort of a promise and a political concession

on Taiwan. What he'll be asking to President Trump is can you delay or postpone your US's arms sales to Taiwan

in exchange of some sort of a trade agreements or us buying more stuff from you, more Boeing jets, more soybeans,

and and and so forth. There are two issues that Beijing care deeply about,

Taiwan that Washington that they feel is very upset with. One is the militarization of Taiwan.

Secondly, is the internationalization of the Taiwan issue. I know these two issues with regard to

Taiwan are on the table even, you know, before this meeting was take place. So, I think that this will be mentioned and

this might be a decisive factor whether Xi Jinping would agree to some of the terms that Trump lay out or trying to

get some wins. I don't think buying more jets or buying more soybeans from the US make any

difference to Beijing. It is very clear that they want this to happen, especially when Trump now clearly needs

some victory or some quick win for whatever it's worth that's happening around the world and he needs this with

with Xi. And it seems like whether we're talking about concessions, Beijing is trying to

extract from the Trump administration on technology or even as you mentioned on Taiwan, seems like there's two two ways

that could look. The first is an actual deal, right? Trump and Xi come out and say, "We we hereby I'm giving you all

the chips. You know, this is our new Taiwan policy." In a way, my assumption is that's actually less likely Yeah. in

comparison to a more subtle, almost a shadow deal where as Ted mentioned, this isn't just one meeting. This is a

cadence of meetings, potentially four between the two leaders in a single year.

President Trump has made very clear that having nice meetings, having a warm relationship with Xi is one of his top

policy priorities. And so, you could see the Chinese also maneuver that cadence of meetings and always getting ready for

the next meeting to potentially enforce a kind of informal restraint on US policy, US steps to defend its own

national security on technology or even in other domains. So, it would it would be much less flashy in China, but over

the course of the year could have a similar impact as well. Um

Kate, what would you add? Yeah, I agree. I think the the focus of this meeting and the ones to come for

China is going to be trying to keep things in this period of detente, keep things in this period where the US

essentially has, you know, agreed to you know, not take any actions to offset the balance.

As for the the rare earths, the critical minerals, you know, sort of restrictions that they've been placed and then paused

as part of the deal, the US vulnerability to those is going to remain for for several years. You know,

there's there are, you know, multiple you know, great large-scale initiatives to, you know, undo that, but it's going

to take time. I think China is very aware of that leverage that they have and the ability

to use that. I think they are going into this scene that they feeling that they have an

upper hand and the ability to potentially get concessions, but at a minimum to keep things where the US

cannot escalate. You know, over you know, the last few years there was, you know, a you know, a sort of a steady

escalatory policy both through new export control policies, you know, other DOD policies against China and I think

they have found a way to to pause those and cut those off. Well, I'd like to open the floor to

questions from the audience. If there are any questions from I see let's start with Dr. Cha

and then I see one in the third row after after that. Really fascinating conversation

doing this morning and today I feel like I've taken a whole course in this already.

But I have two >> my class. >> [laughter]

>> Two quick questions. One for Kate based on the very last thing you said, which is

when we think about sort of critical minerals and the vulnerabilities there for the United States

like so in these large-scale efforts that are being undertaken, you said it'll take

several years. Like how many years are we talking about and what level of dependency are we

trying to get down to? So, the famous case is Japan in after 2010 where they took them about 10 years to

go from 80% to 50%. So, what would it be in the US case? And then second um um

for Ted in your conversations over there with them, so you know, there's a conversation here

about AI that we all know very well in terms of the domestic politics of it where there's

the concerns about how it would affect um workers, jobs, these sorts of things,

the displacement of workers. Um we had Jensen Huang here and he told us that's exaggerated that, you know,

this So, I'm curious like do you is there a parallel conversation going on in China? I know that the the discussion

is much more controlled in China, but just in your interactions with these folks, where where where does that fit

in the whole conversation? I have I have like two other questions, but I'm not going to ask them.

>> [laughter] >> Okay, all right. So, so on the critical minerals piece, it's a it's a very good

question in terms of what how much dependency is an acceptable amount of dependency.

And right now it's clearly, you know, too high. In terms of how quickly to get that down

it's I think it's possible within maybe a two to three year time frame with heavy allied support. So, if you know,

the some of these larger initiatives, Paxil, other critical mineral initiatives

you know, that are bringing in that working with allies, if those come to fruition um

which is it's a bit of a different approach than a lot of other US foreign policy right

now um to do these broad allied efforts, but if those actually are able to you know, really move forward and at scale

and at pace, I think a two to three year time frame is is is possible. Um if in terms of going it alone in terms of

achieving actual the true US production you know, capability on that, I think 10 years would be

the bare minimum. Um I I have detected I had not been in in

mainland China in in two years with this iteration of dialogue that our Yale group recently had

and I did notice particularly on robotics, I think it is creating a lot of anxiety about the future of work

among young people in China. So, I don't want to understate what this change might mean

for Chinese students, counterparts to my students as well as actual blue collar workers, right? So, the large language

models, the gigantic AI tools that you're seeing bombard the conversation in the United

States, those are most likely to replace or challenge a lot of white collar type

jobs in the United States. In China, once you start talking about industrial robotics,

it's blue going to be blue collar jobs in certain types of significant manufacturing and other core

industries. And and as a result, that has an ability to create domestic unrest, concern. The provinces at the

moment all have sort of different strategies and directions. Hangzhou has a relatively hands-off

approach, which is very different than what you're seeing in Beijing and in Shanghai. So,

the central government is going to have a significant challenge of evaluating across all of China's provinces which

types of industrial AI applications are proving most successful, which aren't, which are

creating challenges and unemployment and disenchantment among youth as well as mid-level