China's Generative AI and AI Technology Landscape

China's Generative AI and AI Technology Landscape

Show Video

Good morning, everyone. Welcome to the inaugural webinar on US-China Generative AI competition by the Asia Society Policy Institute's Center for China Analysis. We are thrilled to have you join us today for the start of a series of webinars exploring tech advancement in AI. Our focus today will be the dynamics of competition between the U.S. and China in the realm of generative AI. As these two giants are shaping the world, their strategies clearly hold global implications. As we journey through this webinar series, I also invite you to tentatively mark calendars for two additional upcoming sessions. The specific dates and speakers will be announced later. In the second webinar, we will delve into the domestic economic implications of generative AI for China. Specifically, we will look at its potential impact on China's labor market within the context of China's current economic slowdown.

Looking further ahead, our third and final webinar will explore the effects of export controls, sanctions, and investment restrictions on China's access to advanced technologies. Additionally, we will also analyze how these factors might affect China's pursuit of domestic tech advancement. But first, allow me to introduce our stellar panelists and myself to you today. I am Lizzi Lee, honorary junior fellow at the Center for China Analysis at the Asia Society Policy Institute. Joining me today are Karen Hao, former tech reporter at the Wall Street Journal, Zeyi Yang, tech reporter from MIT Technology Review, Jenny Xiao, partner at Leonis Capital , and Qiheng Chen, a fellow honorary junior fellow at the Center for China Analysis here at Asia Society Policy Institute and also a senior analyst at Compass Lexicon. Thank you so much for being here. In the first few minutes of today's discussion, we'll explore the importance of AI in US-China competition. As you know, today, it isn't just a tech marvel, it's axis around which global power dynamics tend to revolve. Its impact reaches far beyond technology itself, influencing economics, national defense, societal shifts and other international norms.

So my first question goes to Qiheng. Why has AI taken center stage in the strategic rivalry between us and China? And how do their tech advancemenst align with each country's broader global ambitions? Well, first of all, thanks for having me here, Lizzi, and thanks everyone for listening in. And speaking of AI, you have to understand that it is a foundational technology, foundational to increasing productivity in industry, and that's what makes it so essential to China's goal of industrial upgrading and pursuing hig- quality growth in the future.

And also, technology has this dual use nature and has a wide range of military applications, and that mix makes it even more critical to the long term goal of the Chinese government. And it is one of the areas that China actually has a pretty decent shot at catching up the frontier. And we also have to look at this thinking that as we have seen from the Internet economy, that if you're a technology superpower and given the population of China, you're going to be an economic superpower. And with that, you are going to play an indispensable role in global governance and at the geopolitical stage. And the same thinking generalizes to AI, leading from technology to economic and geopolitical superpower. And in terms of how the two countries are aligning their technological advancements with the broader geopolitical and economic ambitions, on the China side, China is betting on quite a few strategic industries and AI being one of them. China is kind of taking a venture capital approach.

Even if China only has a 20% success rate at being very successful in 20% of those strategic industries, that's still a huge deal. And China definitely has the resources and the political will to bet and invest in AI technologies. Fascinating. Thank you so much. Please go ahead. And on the US side is taking a more market-oriented approach.

Although we're seeing the progressives in office and a revival of the industrial policy. So this long-term competition will play out in the next decade or even a longer time. And that's just to set up the stage, and I look forward to the discussion today. Fascinating. So Zeyi. I'd like to hear from you next. Given the rapid growth of AI, which sectors do you think are poised to reap the most benefits? And on the flip side, which sectors might encounter disruptions in the near future? Thank you, Lizzi. Actually, I want to answer the second part of the question first. I feel like right now the industry that feels most disrupted or threatened by the emergence of generative AI is the creator industry.

You are talking about marketers, writers, illustrators, models, or even influencers who feel like they have long thought that their industry cannot be replaced by machines. But now there is a relatively reliable competitor, namely, generative AI that can do their job. So that's why we are seeing a lot of the artists or people who are in the creative industry, fearing that maybe in a few years their job will be completely taken place by AI.

And then on the flip side, I think the industries that are mostly trying to take advantage of generative AI are those industries that rely a lot on the creative work but do not see that as their base. I think one good example of this is the gaming companies where obviously gaming companies need a lot of these visual assets to keep their games running, but at the same time, they usually outsource that to illustrators. They outsource to other companies and right now they feel like this generative AI is a great tool for them to cut down on the cost, to cut on outsourcing and to run on a better profit profitability model, which is also where we are already seeing, let's say in China, some gaming companies are the most adventurous in terms of adapting and using generative AI technology. And also on the other side, I also want to say that not just industries using generative AI, but also industries supporting the development of generative AI are also very much in demand right now.

So we are talking about chips, data centers, everything computing. These are the companies that kind of form the backbone of the growth of generative and they have a lot of great prospects in terms of their business right now. Fascinating, great insights from both of you. Now, just dive into the state of China's generative AI development. As you know, they've achieved quite remarkable progress in recent years, especially in their capability. So, Qiheng, back to you. How has China's AI development plan influenced its journey in generative AI and where do models like GPT fit into China's broader ambitions?

But China has been looking at developing a very serious, as I would say, since about 2017 with a new generation of artificial intelligence development plans. And so that set goals for where China wants to be in 2023, 2025 and ultimately achieving global, leading tGPTbe a global leading player by 2030. But I would say the frenzy of chatGPT actually took China by surprise. If you look at the development plans, China didn't identify the specific technological path to achieving the ultimate artificial intelligence, simply said AI in general, and the chatGPT path, which is part of the transformer architecture family is only one of the several ways to achieving that end goal as it was originally proposed by Google back in 2017 and has developed to this day what we see as chatGPT. And it took China by surprise at the end of last year. And then we see investment pouring into this area and policymakers are also paying attention. Back to how generative AI like models fit into China's broader ambitions. I think China knows this is one of the areas they can buy heavily in.

And so they have a target to direct the state-led investment into. And this belief that if you know what they are going after and China's state-driven model can have advantage in catching up the technology frontier, whereas the Chinese model does less well when the target is moving and is kind of exploring the technology but doesn't really know where is the sure path to success. So the success of OpenAI just points -ut a path for China and but China is still definitely quite far from the frontier and definitely has a long way to go. Definitely has a long way to go. Thank you so much, Qiheng. Karen, I want to hear your perspective now. Can you help us highlight some of the notable milestones China has already reached in this generative AI realm? Thank you so much, Lizzi. I think some of the milestones that we're seeing China reaching right now are quite similar to the milestones that we see in the U.S. There's really a mirroring effect that's happening across both countries. So when openAI first released ChatGPT, we then saw Baidu be one of the main companies in China that announced that they would try to develop their own version and then they released Ernie bot.

We've seen since then that there's been a proliferation of large tech companies both in the U.S. and China and a ton of startups in both countries as well that are trying to catch up both to the other countries and the companies within their own countries. And so you see things like SNAP in the U.S. where they're releasing chatbot features to their app. And then you see Bytedance released a chatbot feature to their apps. And with the open source movement that we've seen towards a response to some of the more closed nature of AI models, we also see this mirroring of Meta releasing the llama model open source.

And then you see Baidu and Alibaba kind of responding with their own open source, large language models, I think the two areas that might be more distinct for China, one is right now there's there's a huge chip crunch in China because of chip export sanctions that were placed by the US in October on China. And so one of the milestones that I think we should be tracking to figure out whether or not China can overcome this challenge is how many of these generative models are actually being built on domestic chips. And thus far, I think one of the only contenders is Huawei has built a large language model called Pangu on its Huawei Ascend chips. But it is still it's a little bit difficult to tell whether or not China really does have a domestic chip that would be capable of bringing more generative AI applications broadly across the country. And I think the other thing is China is actually ahead in the respect of regulation, which we'll touch on, I'm sure, later. But China's made more aggressive moves in trying to regulate this technology both in some ways that we would not want to be learning from and in other ways that we would. And so that's sort of kind of the general landscape right now that we're seeing in China.

Thank you so much, Karen. And now let's turn to the driving forces behind the tech advancement we see in AI: the institutions, the companies, and the researchers. Jenny, over to you now. Can you please shed light on the major corporate players in China? Who are driving innovations in generative AI? And how are Chinese research institutions contributing to research and nurturing talents in this field? Thanks for the question. I think that's a really interesting issue to discuss. A couple of points that I want to make here. One is there are the usual suspects, the big tech companies in China, that are pushing generative AI development, but it's usually more for product development and not for pushing the frontiers of generative AI. That's mostly because these corporations have been suffering quite a lot from the tech crackdown a couple of years ago, and a lot of them were cutting their AI research and development teams a couple of years ago. And I remember just before ChatGPT came out, everyone in China, like all the big tech companies, Alibaba, like Tencent, all of them were like, Oh, we're just going to let go our AI researchers or get them to do product and do something profitable. So I think the funding from the corporate side, from the major corporations, is a little bit unreliable given how they are struggling themselves. I think in addition to the corporate side, the big corporations, there's a lot of startups in China are trying to do generative AI.

This has been a trend over the last six months or so, and a lot of people that I know have come to me or to other venture capitalists. They were like, Oh, I'm starting this AI company in China. Do you want to invest? And unfortunately we don't invest in China. And we just said no to everyone. But there were a lot of people out there who were really interested in starting AI startups. A lot of people, at least like ten teams, were like, Oh, I'm going to be kind of like the openAI of China. And my take on it is that they're not very consistent. So they're interested in AI for like two months and then afterwards they're not going to work on this, which is unfortunately what happened because a lot of these startups are pivoting or they decided to drop the whole idea of building generative AI models altogether.

You've seen this really prominent co-founder of Meituan who came out a few months ago saying he's going to do a big generative AI startup and he's given up on that and Meituan has bought his generative AI startup. And a lot of prominent startup founders in China were like, Oh, I'm going to start an AI startup that build generative AI models. And a lot of them have rolled back their ambitions a little bit over the last few months. I think this is a field that is definitely less hot nowadays compared with like six months ago. And on the institutional side, like educational institutions and government institutions are probably playing a larger role in developing the fundamental technologies.

And in particular, I think the universities, Tsinghua and Peking University, they're doing a lot of work. And also you're seeing government private sector alliances such as BAAI pushing out a lot of really interesting LLMs, and they were actually the first to go into the sector before it was hot. They started pushing out models back in 2020 and these are GPT3 like models and they've been working on it for a couple of years now. So I'd actually say they're probably at the frontier of LLM R&D, whereas the corporations are more interested in the application side. Speaking of what corporations are interested in, are there specific products or use cases from these Chinese companies that have garnered attention from investors? I think compared with the US, there is definitely a lot more... the use cases in China are more concentrated. So everyone in China, like every other generative AI company in China is a chatbot and every other company is an image generator. It's just like very, very concentrated in a handful of sectors. And I think this is because their models are just aren't as good as the models that American companies have access to. And therefore, they're not able to build more complex use cases on top of their technology. So it's really like very, very simple use cases.

And I think what is interesting, though, is in Silicon Valley, like, if you look at how many people are building LLMs versus building applications, there's like a lot more people building applications versus building the foundational model. There's a handful of major companies that are building foundation models and like hundreds of application companies. Whereas in China, you kind of see, a lot of people building on the foundation layer because there is less of a clear winner in the foundation layer. So a lot of people are pouring money into building the openAIs of China, trying to become an API company, whereas in the US, fewer companies are doing that. And in China there's a lot of application layer companies, but these application companies are usually concentrated in a couple of very simple use cases rather than the complex use cases that we're seeing in Silicon Valley. Fascinating. Thank you so much, Jenny. And as we start to dig deeper into the unique strengths and challenges that China US faces and brings to AI competitions, Zeyi, here's a question for you How do you think the research landscapes differ between China and the United States? And how do the talent pools of China and U.S. compare in terms of research and applications?

I actually think there are a lot of similarities between the research environment in China and US. There are a ton of research talents and there are actually a lot of collaborations going on to really research institutes working with each other. I will point you to a report published this year by the Stanford Human Centered AI Center. What they found after going through all the scientific publishing of AI, is that China has published way more papers than everyone, however, in terms of the impact, the citations, the US still leads there. So we are seeing that a lot of Chinese researchers are working really hard and very diligently, but they're still catching up to the knowledge advantage that US researchers have. The other thing the report also pointed out is that even though there are still collaborations going on and happening, the pace actually slowed down.

I think that actually speaks to the larger environment of scientific collaborations freezing and slowing down, which I believe Karen can talk more about. But that's like one of the things we're seeing: they might go on more kind of separate routes from now on because of the geopolitical tensions, everything else. The last thing I would note is that I think in both countries we are seeing that a lot of times the companies actually take a more leading role because there's a very high cost that comes with developing generative AI models because of the need to spend on ships, the training and running of the algorithms. So sometimes the institutions don't have that much financial backing to justify spending on that year over year. And the companies, especially companies that feel like they have a chance to be like the first mover or to have an advantage in the industry, they are actually more financially capable of investing it, but we also see them cooperating with the academic researchers by funding their research to get their research to use their product. So we're seeing like really all of them tangled together. But I think in this case, the U.S, and China don't differe that much.

Not not that much of a difference. Thank you so much. To Qiheng. I know this relates to your research. Can you talk a little bit about data accessibility and data availability in China? And how pivotal is that issue? Is that an issue for China's progress or further progress in AI? The conventional recipe for success in AI is talent plus data plus computing power. I would say data is one of the three areas that China has the most ability to mobilize resources and have control over. And China for the past several years, has been engaged in the effort to optimize the allocation of data resources, meaning that alloting the data that are now used by one party and making that available to other parties that have better use cases for the data.

And we have seen initiatives to establish data exchanges to facilitate this kind of allocation of resources. But the degree of success, I would say, is still questionable at this point. To my knowledge, the vast, vast majority of the exchanges still have off the exchange, meaning that it happens directly between the two parties not through the exchange, but China's also trying very hard at unlocking the valet of public data, pushing government agencies to release public data for training and for use and applications. It's it's still a work in progress. But I would add a caution to this conventional thinking that China sits on a vast trove of data. China's PC or internet ecosystem may not be as developed or has as long a history as the US, you may not have an ocean of data to scrap from and to fit into language models. More data is in the mobile ecosystem, which is more closed off and is controlled by the big corporations. And so it still remains to be seen if China's really abundant in training data for large language models.

That's a really important caveat. Thank you so much, Qiheng. Jenny, I want to hear your thoughts on this. And you mentioned these distinct paths between the fundamental LLMs and applications between the U.S. and China. Are there are there sort of niches or specializations that each country is leaning toward at this moment in generative AI, as far as you can see? Yeah, I think a lot of traditional trends that we're seeing in the mobile era also apply to the generative AI, specifically in the US. A lot of companies are building B2B products. These are parties that serve other businesses. These could be like products for the financial sector, for health care, for serving like doctors, or say, serving lawyers like legal type products, whereas in China, most of the products are to see their products directly serve customers.

And the big difference between these two types of products is that U.S. products tend to be more hype- driven. So people are super interested in this app and everyone's like downloading it, whereas like B2B products are more like, Oh, we need to do enterprise sales to get all these big companies using our products. And China has always been very good with the 2C play, but very bad in 2B because a lot of Chinese companies don't want to buy software and this has been a bottleneck for generative AI companies in China and they're like, Oh, I really can't sell software.

No one's going to buy software, they only want to buy my products if I provide consulting. And this is a trend we continue to see today in the generative AI era. On the 2C side. I've actually seen some pretty cool applications. A lot of people are building chatbot apps for gaming and PCs and also people are building, like AI girlfriends or boyfriends that was like a whole trend a couple months ago. And also people are building stuff like AI psychologists or whatever. And there's a lot of really cool 2C products. And also in addition to language generation, a lot of folks are building video generation or short form videos. That's something that TikTok brought to the US that's getting pretty big in China. I've had like a couple of companies come up to me and they're like, Hey, I'm building this short-form video generation platform or app. It's also a form of generative AI, but like less talked about in the US and like there's a lot more interest in that in China.

But I just want to quickly touch on a question that we were talking about earlier, which is data accessibility. I think what is interesting here is I heard a couple of companies tell me or tell other households that they have exclusive access to some government data in China because they have like a close relationship with some government officials or they were guaranteed by a local government that you can get access to like all the hospital data in their province or all the hospital data in that certain region. And that's really interesting because for US companies that kind of data is really, really, really hard to get. A lot of health tech companies in the US are really struggling to collect data and they will absolutely envy the data access a lot of Chinese companies have. Fascinating. Thank you so much, Jenny. We already touched upon this a little bit, but let's talk more about regulatory landscapes in both countries.

Qiheng, how is the regulatory environment in the US shaping the development and deployment of generative AI technology and also for China? So the U.S. is still in the very early stages of developing concrete laws on AI, and they are still in the discussion stage of how to go from the general principles to very concrete harms and to map out what existing regulations can be applied to harms originating from AI and what new regulations are to be developed. We have seen a whole of government approach to addressing the governance of AI. You see quite a few agencies within the U.S. government claiming that AI falls under their regulatory power, notably on the discrimination and bias side. You had a joint statement from FTC and 3 other agencies and also the White House, pretty notably around the end of last year issued the blueprint for a Bill of Rights that basically laid out five principles and mapped out how local and federal efforts are tied back to those five fundamental principles. And again, about a month ago, you see the voluntary pledge of several leading tech companies that they are pledging to watermark their products.

So you have seen this industry-led bottom-up effort coupled with the top-down approach: some of the guiding principles formed the early stage of the regulary landscape in the US. But it's going to take quite a few years for the Congress to really materialize any high regulation if there were to be any. And in this vacuum, the tech companies will just need to navigate the water and have robust internal compliance and risk management programs and help shape the discussion of regulation. Fascinating. Thank you so much. Karen, I also want to hear your insights. In what ways do you think China's regulatory approach toward AI differs from the approach employed by the United States?

China is much more aggressive and it's also much more top-down. Like Qiheng mentioned that in the U.S., it's really driven bottom-up by the companies. But in China, it's the government that's really setting the pace on what they want to see generative AI used for and where they do not want it used for the ways that it should be developed and should not be developed. And we've seen this just generally in a lot of like cybersecurity and Internet regulation.

China sort of tries to keep up with the clip of technology development, in part because of a fear that when you have new technologies that could really influence the information ecosystem, they really want to try to contain that as quickly as possible. So you see in a lot of the regulations or the laws that come out, whether it's a cybersecurity law or the deepfake synthesis regulation that was just enforced in January this year. You'll see this language like for any algorithm or any generative technology that has social mobilization capabilities, the companies need to make sure that they kind of button up their technologies and don't end up influencing the Internet environment, the digital ecosystem to misalign with the party. But at the same time, because of this aggressive clip, China has also been quite ahead in some other areas that are more aligned with sort of Western Democratic or liberal values.

Like watermarking was one that Qiheng mentioned, where a lot of U.S. companies are exploring this idea of how do you inject some kind of digital fingerprint into these generated outputs so that it's easier to track on the Internet so that you can reduce misinformation risk. That's something that the the EU has also introduced into their regulatory frameworks, and now that's something that China has also introduced, whereas the US still doesn't really have like a formal law at the federal level that is looking at requiring something like that. So this is what some China policy analysts have sometimes told me, like the sweet and sour approach of China's regulation. Like you get a little bit of parts that you're a little concerned by. And then other parts are actually very clear that China is listening to the global regulatory discussions and trying to pluck things from the EU discussions or the U.S. discussions that they think would also be quite beneficial for their environment. Fascinating. Thank you so much, Karen. Another common question I receive about China's AI capability is the effect of the censorship regime, which can have a significant influence on information flow and the direction of technology.

So Zeyi, your thoughts on this. How does China's approach to Internet censorship intersect with its development? Are there instances of AI technologies being utilized or adapted for surveillance or censorship? Yeah. So actually, when Jenny just mentioned there, in both countries, a lot of generative AI are used for 2C products. So this means everyone can access those and ask questions and get your answers. The problem that comes with that is that there's a lot of unpredictability when it comes to what people are going to do with the models.

And that is a big risk for Chinese companies because China has had this, like very comprehensive censorship system that goes into basically every social media platform. So they also need to kind of like adapt the machine into the emerging generative AI products that's going to introduce more unpredictability, introduce more sensitive conversations. So I think what we have seen so far is that the Chinese companies have been very careful in navigating their landscape, especially when there have been a lot of Chinese companies releasing their products. But very few of them have opened them to be completely accessible to the public. And they're required to register with a real name. They're not required to register with a business entity.

So they kind of want to control the risk in terms of what they're going to generate and what they are going to attribute to these companies when it comes to contributing in the production of this kind of sensitive discussions. So that's very interesting. And we didn't really see it here. The other thing is that we are seeing how they approach this question from a technical standpoint, like do they just block certain keywords from being asked, which is more like the search engine, like Web 2.0 censorship mechanism, or do they use something else to try to prevent the models from saying things that a company doesn't really want to hear? For example, I think recently Bytedance released like a week ago their own kind of chatGPT alternative. And people have realized that if you ask the question that's about Chinese political leaders, the Bytedance bot will actually generate the answers first. And you kind of see its answer.

And then once it's done, you will see it being blocked and say, actually, this is a politically sensitive topic and we could not generate it for you. So this is like a new territory. Like, people don't really know how to censor AI model. So they're trying all kinds of ways. But I feel like soon that they're going to find maybe this industry standard in terms of how to censor this kind of conversations. Unfortunately, I think that's going to happen soon. And again, to another part of the question. I don't think we have seen much of how generative AI has contributed innovatively to surveillance because it's very much in development.

But we're more seeing a continuation of the previous challenges to freedom of speech, to access to knowledge that are replicated on generative AI, because all these companies are so fearful of the political liabilities that they may have for introducing models without having some kind of guardrails there. Fascinating. Thank you so much, Zeyi. Jenny, back to you. How does the general public in China perceive generative A.I., particularly in the context of state censorship? Are they getting around it or are they just complying with censorship? What are the workarounds? What are your strategies and reactions? Yeah, first of all, I don't think a lot of the general public are very concerned about the censorship. I think a lot of them are very aware of the censorship, including the ones that have played around with these tools. Their reaction is, Oh, this, this tool like very dumb, right? It's not a very good chatbot. It's not as good as the chatbot in the West and they're not aware that that's because of the censorship layers behind the scenes.

I actually have an interesting anecdote. A Chinese CEO once complained to me that his Chabat can't even count from 1 to 10 because there's eight and there's nine, which I not only think is a really funny story, but it also tells you how dumb the Chinese chatbots are and I think the general public usually complains about it from a user experience perspective, like the user experience is not great the chatbots are kind of dumb. They're not really aware of the deeper causes, and if they are, they're probably used to it because this is what the Chinese Internet looks like. This is what Baidu, what all the search engines look like. So it's not particularly strange to them. But I do think the AI community in China is very aware of censorship requirements, and this is one of the main barriers to them releasing better chatbots is one of the barriers for them to make better products have made more innovative uses of their products. Fascinating. Thank you so much, Jenny. So we are coming to a close of the formal discussion part ofhour session.

But I really want to hear Karen's perspective. This came up during our offline conversation. We talk a lot about US-China competition. Let's just not forget that there are still opportunities for collaboration. What kind of opportunities for US-China collaboration do you see, Karen, between the two countries in the realm of AI? Thank you, Lizzi. I really appreciate the opportunity to talk about this, because I do think that the competition lens sometimes dominates the discourse so much that we really do forget how much cross-border connection there is and how much we should be fostering that in cases like generative AI, where it is a really transformative technology and it's going to be global and borderless and you're going to want to have some kind of like international norms around it. I think in AI already we've seen before generative AI became the term of the day we were already seeing so much US-China collaboration. There's this Macropolo study that showed that globally, like a third of the AI talent comes from China, a third of it comes from the US, and a third of the talent in the US comes from China.

Just looking at people to people flow, there's a constant exchange of ideas that's happening and there can be continued collaboration on beneficial applications like AI and drug discovery, scientific advancement, education applications, you know, financial obligations, things that are just generally beneficial, or like helping combat climate change, improving health care outcomes. And the other area that I think cooperation is really important is the risks and the harms. And we've seen some efforts with this. The Beijing Academy of AI hosted a conference that Jenny actually pointed out that brought together some of the leaders in China and the US on AI research. So Sam Altman appeared digitally. There was an executive from Anthropic that appeared digitally. There were professors from MIT, from Tsinghua, from Baidu.

From aall of the major firms that we're talking about, some of the ethics and harms and making sure that together, these two big superpowers, are actually communicating with one another on what should be the guardrails moving forward for developing this technology. So I think both the benefits and the risks are really, really important areas of cooperation. Fascinating. Qiheng, the world is not just the United States and China. How might other countries, you know, position themselves in this whole US-China competition or collaboration? Where are the other countries in the picture? So my base case is that the technology companies will come from China, the US and then the rest of the country will just need to see where you are standing. So EU will still be super relevant because of its regulatory power and competency. And we also have some other developed countries that have a very open system of open engagement with the United States that's going to benefit from access to US technologies while also open to governance cooperation with EU and the US.

And then you may also have another group of countries that I think the top concern is about more about people's access to AI technology. So that's one dimension of looking at other countries. Another dimension is looking through the lens of geopolitics. I am quite worried about that. We're going to see a bifurcation of the ecosystem. One is US-led and one is China-led. I don't see much interaction between the two, not just in terms of talent and technology, but also in terms of competence. Governments that may have a demand for censorship. technologies might resort to Chinese companies to fulfill that demand.

And this bifurcation will have much wider implications in terms of the global AI competence. Fascinating. Thank you so much, Qiheng. Karen, back to you again. What are some of the ethical and societal considerations both countries should prioritize as they advance in the realm of AI? There are so many. Well, one thing that's top of mind for I think a lot of people right now is this challenge around misinformation. So a lot of these generative AI. tools, they are not truthful because there is no sort of knowledge base that it's drawing upon. This is called hallucination in the technical world. So that's a problem that really needs to be thought about just whoever is developing generative tools, whether it's US, China or elsewhere, also protecting workers' rights.

I think we're seeing a really big wave now of like with the protest in Hollywood of workers in many industries feeling a lot of squeeze. Zeyi also mentioned influencers feeling this pressure where they think that they could be replaced by generative AI or even the threat of being replaced is starting to put pressure on their wages, their job security, their economic outcomes. So those are areas that I think both technologists and policymakers around the world should really be collaborating on to figure out how do we ensure that as we move into an increasingly automated, generative world, that we still have dignified, long-term jobs? And I think the last that I'll mention is like data privacy and consent. There's also a really big global debate now around whether or not we should be allowing all of our data to be consumed by these technologies to create them. And artists are kind of protesting this, journalists are protesting this, and comedians, actors are all sort of challenging this idea that their content that they put online under different circumstances should then be just used wholesale in the production of these technologies.

So I think that's another area that regardless of where these tools are developed and even for jurisdictions that aren't necessarily developing the tools but are actively thinking about how to regulate when the tools come to their jurisdictions, that should be a really big area that should be sorted out. Looking into the future a little bit, Jenny, what disruptions or innovations could emerge on the global tech landscape due to US-China competition, due to the development of AI, as far as you can see? Yeah, I'll talk about disruptions first. Just going back to Qiheng point that there might be a bifurcation in the world with the US and China. There's going to be like a glass sphere and then there's a China-led AI sphere. I have no doubt that that's going to happen.

And in fact, this is already happening. Just give you like a very, very concrete example. Nowadays, it is almost not impossible, very, very, very hard for us investors to invest in China. And this have a couple of downstream effects. One is China's AI landscape is really going to lack the venture capital that's needed to grow a lot of these applications, a lot of funds in China that have a very strong AI focus are telling me that, oh, they can raise capital from the US, but because traditionally 90% or 70% of their capital, their money come from US university endowments and these endowment firms are pulling out of China-based funds because of national security risk and because the Biden administration came up with a couple of policies that prevent American investors from investing in China. And also there is this very big venture fund in China that is also like a branch of the US fund. They received a letter from Congress a couple of weeks ago warning them about investing in China and also asking them to appear in front of Congress to explain what exactly they're investing in and what they're doing.

So on the capital side, I already see these as two distinct spheres. It's very hard for Chinese investors to invest in the US, either because there's all sorts of KYC rules that US companies put in place. And these are like KYC, know your customer. That means they have to do background checks on all of the investors. And the US has a lot of very strong, very strong like investment restrictions for China-based investors of investing in the US. So I really think these are like two very distinct spheres. A lot of American companies are suspicious of people with Chinese work experience, especially if they work for Bytedance or Baidu.

They're always worried, Are you going to like, steal my technology and give it to a China-based company? So I think there is going to be a lot of disruption and there's going to be two spheres. And on the innovation side, I think what's interesting is I think the competition is definitely pushing hardware innovation in China as all of these companies are switching. These China based companies are switching to their domestically produced hardware that pushes innovation on the on the China side and also on the American side, because there is this risk or fear of China. A lot of government agencies are pouring more money into supporting AI research, especially fundamental research. And I think, you know, there might be interesting innovations that are coming out from this competition, but I'm more worried about the potential for a US-China arms race.

With that we conclude the first part of our discussion today.

2023-08-31 14:08

Show Video

Other news