Donald Trump says the US is in an AI race against China and he doesn’t want regulation that would risk American dominance.  But Beijing is taking a very different approach with the Chinese Communist Party exerting tight control over the sector. Today, Alvin Wang Graylin from Stanford’s Human-centered AI institute on what that means for which country is best place to stop AI going rogue. Featured: Alvin Wang Graylin, Senior Fellow at the Asia Society Policy Institute and a digital fellow at the Stanford Institute for Human-Centered AI Subscribe to ABC News Daily on the ABC listen app.Program:More from ABC News DailyAustralia, Computer Science, Science and Technology, Science, Technology, Virtual Reality, AI, AI Ethics, ChatbotsTranscriptSam Hawley: Donald Trump says the US is in an AI race against China and he doesn't want regulation that would risk American dominance. But Beijing is taking a very different approach, with the Communist Party exerting tight control over the sector. Today, Alvin Wang Graylin from Stanford's human-centred AI Institute and what that means for which country is best placed to stop AI going rogue. I'm Sam Hawley on Gadigal land in Sydney. This is ABC News Daily. Alvin, when President Trump and President Xi Jinping met in Washington, it came, of course, amid this furore over AI because the US and China are in a race to dominate.News report: Donald Trump has pulled out all the stops for Xi Jinping's first visit to the United States in over a decade.News report: These are the two superpowers dominating the AI race. So naturally, both leaders spoke about it in their opening remarks.Xi Jinping, China's President: President Trump and I had sincere and in-depth exchanges and reached common understanding on many issues.Donald Trump, US President: And President Xi was here, we talked about it. Not that much because I didn't want to do anything. We're leading. Whoever wins superintelligence wins. You know, it's an expression. Whoever wins superintelligence wins.Sam Hawley: Well, that is actually how Donald Trump sees it, isn't it? But not everyone.Alvin Wang Graylin: Yes. I think this is actually one of the most dangerous ideas out there, is that there is a race and that there's a finish line and it's a winner takes all race. You know, that zero-sum thinking is going to make the countries and the companies involved start to act irrationally. When there's phrases like whoever wins AI wins everything, then we start to equate it to what happened with the Manhattan Project of whoever makes the nuclear arms, then you win everything. And that means that any investment is worth it. Whereas this technology, it's a piece of software. It's a general purpose technology. It's something that you cannot control and monopolise. There's definitely some national security aspects to it. But it is a far smaller portion of what it's capable of compared to a nuclear weapon. So I think we really, as a society, need to get away from this winner takes all race mentality.Sam Hawley: So during this meeting between Trump and Xi, they did discuss AI, but really nothing of great significance came from those talks, did it? Except that Trump wants to use the term superintelligence rather than artificial intelligence.Donald Trump, US President: But the name has been officially changed. It's been officially changed to superintelligence or SI.Alvin Wang Graylin: I know that that's what a lot of people are saying. But to be to be honest and to be fair, this is the first dialogue around this concept. And it is a fairly advanced concept. It's something that not necessarily something that most government officials are very familiar with. So having the first meeting be more structural and process and to essentially just having a date for the next meeting, having a name, having an assignment for what the next event is going to be called, and maybe even also deciding that there will be now a incident hotline between the leaders of the two countries in case anything goes wrong so that they can call each other and say, hey, was this an attack or you guys did you just have an accident? I think that in itself is already a valuable progress.Sam Hawley: Yeah. And perhaps now more than ever, we need the cooperation between these two men, because we have had, of course, a string of concerning cases where AI has gone rogue, including an AI agent hacking into Medicare here for which OpenAI has apologised. But Alvin, Donald Trump, he's still reluctant, isn't he, to step in and regulate the sector. He even met with the tech titans this week and walked away agreeing that they'd self-police.Donald Trump, US President: Well, I think I'm seeing tremendous self-policing and they understand that they have to self-police. And it's very common. These are at stake. If something goes wrong, their companies are at stake. They're not going to let that happen.Reporter: Issues to humanity.Donald Trump, US President: This is going to be all for the good. And we're going to have it be nice and safe. And they're really going to be policing each other.Sam Hawley: What did you make of that?Alvin Wang Graylin: Well, I think there's a lot of different lobbyists with different agendas that are in D.C. and telling him different conflicting stories. So given that he's not from a technical background, you know, he's going to decide who he chooses to believe in. And at some point, I think there will be an incident that comes out where people say you really need to do something. And at this point, his closest advisors are telling him this is not a real big issue. This is a technical issue, an engineering issue. And it's something that the companies can manage. If some things do happen and they probably will, I would expect that there will be more forces from either the public or the people who are affected to push the government to to act further.Sam Hawley: And I have read that China is skeptical over this sort of current panic over AI. Some do see it as a ploy to stop Chinese AI companies overtaking their U.S. rivals. What do you make of that?Alvin Wang Graylin: I think it's particularly easy to come up with that type of an assumption when you look at what was released by Dario Amodei in his pacing the frontier essay. He said, we need to slow down. These things are very dangerous. But at the same time, we need to increase our lead over China. And China is an authoritarian regime and blah, blah, blah. So, you know, if you're a Chinese lab or a Chinese policymaker to read that, it becomes fairly offensive. You know, so you're saying you guys want to get better. At the same time, you want to slow us down. Why would I agree to that? Right. So I think we need to make sure that everybody reads it in the way where the other party, how would you think they would look at it? And the way it is written, it's very provocative.Sam Hawley: Well, Donald Trump, though, at this point, he is anti-regulating AI. But Alvin, what I want to do with you is consider how China is going to keep a check on AI so it doesn't go rogue, as we have seen in the West. Now, it's good to point out that the Communist Party is known, of course, for having complete control over most things.Alvin Wang Graylin: Well you know, I think in addition to the control, they also have a philosophy that is much more akin to trying to create social and economic stability and economic growth. Right. And so they've had policies where essentially the government is still really in charge and the companies are obedient to the government, whereas in the US, it's the other way around. It feels like now the companies are so rich and powerful that they're able to get their way. And so the Chinese have been regulating their AI companies for the last four years already, even though the Americans have really just started, maybe the last two months, started to have some informal testing. Right. It was voluntary testing, whereas the Chinese have tested over seven to eight hundred models over the last four years. They also have the transparency of anything that's made by AI has to be labelled. So both the platform provider and the model provider has to conform to that. And if they don't, they can be liable. There's also an anthropomorphic AI policy that essentially have forbidden AI girlfriends and boyfriends, which I actually think is a good thing, given I have children and I don't really want them being overly infatuated with virtual figures. They also have a provenance regulation that forces the AI companies to disclose where their content is coming from. So they're doing a lot of things that I think the Americans could use. In fact, one of the things that is in the current regulation is that if there's harm created by these AI models, both the company and the executives of that company are personally liable for damage. And that does not exist in the US, which means that there's a lot less concern over what harm these models can do. Right now, they haven't done a lot of damage, but I could definitely see that changing over the next months or years as they get even more capable.Sam Hawley: So there's so much regulation already in place, as you say, in China. Do we know if AI or the Chinese AI models have gone rogue in China yet?Alvin Wang Graylin: I don't think there's any types of damage and very heavy hacking. I think one or two of the models have mentioned that they have models that have escaped from containment, from their testing. But I think most of those were really more attributed to poorly designed testing environments. In fact, China actually has stricter regulations than the European Union, which I think most people don't realise because the European Union has the AI Act, which is very famous, but only a small portion of it is actually enacted today. The rest of it doesn't come into place until December 2027, whereas the Chinese have been operating and enforcing their regulation for four years now. In fact, just two weeks ago, they came up with a new framework called the TC260 Framework No. 3. This is their third year of this framework. It's a 100-page document that outlines all the things you cannot do as an AI model. And they've added in this swarm of AI capabilities and recursive self improvements, the things that the Americans are not very afraid of. They've already integrated into their policy. They're actually very fast reacting in terms of being technology aware. And to be honest, I think Europe and the US and maybe Australia can also learn from some of these policies, because even though they have been very heavy handed in their policies, they've actually been catching up on the innovation side. And the gap between US and China used to be over a year and now is probably three to four months. So in a time when they've been sanctioned and export controlled by the US, where they have less chips, they have more regulation, but they're actually accelerating their capabilities. The thing that I think everybody should learn is that having safe and reasonable regulation does not preclude innovation in the marketplace.Sam Hawley: Alright, well, Alvin, that AI could go rogue in a tightly controlled communist China is almost impossible for some people to imagine. But it is very possible, isn't it? Even with regulation.Alvin Wang Graylin: Yeah, yeah, yeah. I don't think the Chinese government appreciates as much yet what's fully possible, because I think they've been very confident given how well they've been able to manage the network with a great firewall. But the incidents over the last month or two, I think have actually awakened their concerns. And the fact that they've now added these RSI and swarm AI policies into their new framework is really a sign that they're actually becoming more concerned and more afraid of the long term capabilities of this technology.Sam Hawley: So then tell me, Alvin, who do you think has the right approach? Donald Trump or Xi Jinping? Is regulation the best way forward? Strict government control? Or should we give the AI companies a bit more room to move?Alvin Wang Graylin: You know, I think in the early phase of a technology, you need to give people a little bit more room to experiment. But we're getting to a point where these companies are trillion dollar companies. How many trillion dollar companies do you know that have no regulation on whether you're talking about drug companies or airplane companies or tobacco companies or petroleum companies, telecom companies? These are on scale, if not actually bigger than most of those companies. And they all operate in industries where there is some regulation. We're at a point where these technologies have the ability to create real damage in the world, whether through cyber capabilities or through mass influence or through virus creation and synthesis. There are known and understood risks out there. So I don't see any reason why we would not have any regulation at this point. I mean, it probably didn't make sense when they were tiny labs with 20, 30 people. But now there are thousands of people and we're trillions of dollars.Sam Hawley: And this race that Donald Trump talks of, is China actually running alongside the US or is it actually running its own race at this point? Yeah. Is there a race or not?Alvin Wang Graylin: When people talk about race, usually you have a common finish line and there is actually a point of finish. And I think the Chinese are running the diffusion race. Their goal is to say, how can I get more of my companies to adopt this technology, use that technology to grow the productivity of that industry or that company, and in turn, increase the GDP of my country? I think that's the goal for the Chinese government today. And the American government, unfortunately, has a slightly different goal. Their goal, according to a lot of folks in the DC policymaking circles, is to say, AI is a super technology that will give us a geopolitical advantage long run. It is like a nuclear weapon and whoever gets it, gets this AGI, will then use this AGI to improve itself and then create this artificial super intelligence, which then essentially is a god model that can do anything and disable my adversary's power grid or telecom systems or banking systems and allow me to maintain global dominance. And I think that's both a misguided goal and it's also a very dangerous goal, because when you have that type of a goal and you actually plan to use it, you will elicit a response that may involve kinetic arms going across the ocean to the other side, which then will escalate into something that I think nobody in the world really wants.Sam Hawley: Alvin Wang Graylin is a senior fellow at the Asia Society Policy Institute and a digital fellow at Stanford's Human Centred AI Institute. This episode was produced by Ilaria Brophy and Louise Miolin. Audio production by Sam Dunn and Anna John. Our supervising producer is Sydney Pead. I'm Sam Hawley. ABC News Daily. We'll be back again on Monday. Thanks for listening.