US v China: The AI race

0:00Marianna Spring:This BBC podcast is supported by ads outside the UK.

0:30Anthony Zurcher:and more options to their lineup, can shop at machinefinder.com. Booyah! America, it is fair to say, is freaking out about AI. It's suddenly become the hottest of political topics. Silicon Valley tech bosses are calling for the development of AI, their own development of it, to be slowed down. Everyone seems to be worried. Everyone except one person, Donald Trump, who's not worried at all. He dismisses worries about AI as what he calls a sick conspiracy. Whoever wins AI wins, he says. Is he right? And what happens if the US does fall behind China as he warns that it might? Welcome to AmeriCast.

1:14Marianna Spring:AmeriCast. AmeriCast from BBC News. Whatever you want to talk about, Abdul, keep my wife's name the hell out of your mouth because she's way out of your league. The tide is coming, and this time we intend to win it all. See, he doesn't want to do the job.

1:33Anthony Zurcher:He wants to build his ballroom and travel with Natalie on their apparently defenseless flying palace gifted by the Emir of Qatar.

1:41Marianna Spring:When I'm not around, you're going to be broke. There's not going to be anybody to report on. Nobody gives a damn about anybody else.

1:52Hello, it's Justin in the worldwide headquarters of AmeriCast in London, England.

1:56Marianna Spring:And it is Marianna sitting next to Justin in the worldwide headquarters.

1:59Anthony Zurcher:And it's Anthony holding down the fort here in Washington, D.C., which is the American headquarters of AmeriCast. And we're going to be joined later on by a fantastic guest. Chris McGuire is going to join us, who is a former senior advisor at the State Department for Donald Trump in both his first and his second terms. But really interestingly, he also served under President Biden and he also has links with China, is a US-China technology expert now at the Council of Foreign Relations. In other words, he is the ideal guest to talk about our subject today, which is artificial intelligence and specifically this suddenly looming political issue about what to do about it.

2:43basically, what to do about making it safe, what to do about controlling it, if controlling it is what we want to do, but also what the threat properly is assessed to be to all of us, whether we're in the worldwide headquarters of AmeriCast in England or in Washington, D.C. or anywhere in the world, because it is that kind of thing that we're talking about. And people are talking about it, aren't they, Anthony, in kind of worried terms, except one person who doesn't seem worried at all.

3:11Anthony Zurcher:Yeah, Donald Trump. Donald Trump seems to be an AI evangelist. He took to Truth Social earlier this week after about a week of all of these concerns bubbling around and then spilling over into American politics, concerns being raised by AI executives themselves, politicians, policymakers. Well, Donald Trump on Truth Social said, yeah, that's all bunk. And he said that AI was actually going to be a great thing, maybe even bigger than diamonds or oil or even the Internet. He said the only control or guardrails that AI needs is a strong and smart, high IQ president. And the USA has that in spades.

3:51Anthony Zurcher:So that's Donald Trump, I guess, talking about him, definitely talking about himself there. He went on to say that there is a sick conspiracy going on against AI and data centers. And the only one that is happy about it is China. Whoever wins AI wins. So, you know, Trump is, one, centralizing control into his own hands, which he does all the time, talking about how he's the one who should control this and we should trust his judgment. but also framing this, Justin and Marianna, as a competition with China, the idea of this is an arms race. And if you don't win it, you're going to lose. And if you lose, then you're going to be in serious trouble.

4:27I guess what has felt different, aside from how vocal he's been,

4:31Marianna Spring:has been the comments been made by people within the industry who are in charge at the companies. And a lot of the big tech bosses, AI bosses have been talking about this concept of a slowdown. and should we sort of put the brakes on a bit, sometimes feels a bit ironic given that they are sort of the people in charge of the brakes. But nonetheless, here they are talking about their concerns. You've got Dario Amadei of Anthropics, Sam Altman of OpenAI and Elon Musk of XAI. It's a warning sign. It's a warning sign that we need to slow down. We need to make this technology carefully and we need to make sure that our safeguards, our ability to understand it, our ability to control it, keeps up with the pace at which the technology is happening.

5:20Marianna Spring:I think the world is right to be afraid of this. So we have these two big challenges. We have the potential of a loss of control accident or some other serious thing that could go wrong. We have the potential of way too much power concentration and people fear that those developing AI could exert their worldview. And there is a sort of narrow path we have to navigate with pragmatism, with, you know, steadiness, with the ability for the world to trust that we will make dependable decisions, consistent decisions through this. And I think people are saying, well, you know, this now seems pretty important.

5:57Marianna Spring:Like, let's get it right. You know, when a lot of people from Anthropic and from OpenAI are telling you that their models are very dangerous. I think we should believe them. On the other hand, if you're selling chips to these people, you might feel differently. And here is the CEO of the most valuable AI tech company in the world. So this is Jensen Huang. His company makes the US chips that power AI, NVIDIA. Let's listen to what he says. We don't need any new laws. We don't need new regulations. We just need companies to decide that when it's run as fast as they can, I think innovation speed and safe products are not, it's a false choice.

6:42Marianna Spring:You could definitely have both at the same time. So run as fast as you can, but if you feel at any given point in time the company's out of control or the product's not going to be safe, you know, take a pause and make sure you get it right. I think it's quite notable that when we heard from the different tech bosses there, they tended to be criticizing other companies, not their own, and or sort of talking about it in AI in a kind of conceptual way. But this is the problem with this discussion, right? And to some extent, the issue with the original post that was shared on X by this person who had worked for both Anthropic and OpenAI, who said, I'm worried about the end of humanity because of this technology.

7:23Marianna Spring:It's all quite vague and not very transparent about what exactly we mean when we say this could kill everyone and or this is all getting a bit out of control and moving at a rate where perhaps we're not thinking about the unintended or intended consequences.

7:39Anthony Zurcher:Yeah, when people talk about this, they think about Skynet. You remember the Terminator movies, the AI that gained self-awareness and then took over the world and crushed humans beneath their metallic boots. And how we get from where we are now to Skynet is kind of a big question. And you're right, Mariana, they don't really spell out how we get there. I think there's a couple of ways of looking at this. One is that these warnings are real, that this could be something that we have to take seriously. And I've spoken with tech executives who do very much feel like things are kind of spinning out of control and there needs to be a slowdown.

8:15Anthony Zurcher:On the other hand, you could see this, as you mentioned, this could be internal divisions within an industry where everyone is jostling to try to gain the upper hand. And they are now using levers of politics and trying to influence regulations and set the playing field in order to advantage them. And so this is really just a competition from the big players who want to keep out the smaller guys by new regulations, make the barriers of entry higher and to disadvantage people who are also right there neck and neck with them. So, you know, pick pick which way you think this is going to go. But I think there are arguments to be made for both sides.

8:53Marianna Spring:One of the reasons why there's this feeling of urgency and panic and distrust and anger and frustration in some cases around AI, which is different to how people responded to, say, the rise of the social media companies over the past 15 years, I think is for several reasons. One is obviously that the AI companies have grown at a speed and the technology has developed at a speed which is certainly faster, I think, than we found with the social media companies. It's taken the best part of 10 years for people to say, oh, maybe algorithms might be negative. What can we do about that? But crucially, even if you don't use AI, you could be impacted by AI, whether it's that you lose your job or whether it's that you get hunted down by a killer robot.

9:38Marianna Spring:whatever format it takes or affected by a virus or affected by some kind of nuclear code or whatever it is. Everybody and humanity. And I think that's why this whistleblower who posted on X caught everyone's attention is because they basically said, you're all in trouble. It's not just some of you. And I think this call really hit a nerve with everyone because this idea that it could affect all of us or people, you know, our children or our immediate generations. Everyone's like, oh, whoa, hang on a second. I think if you told people, oh, it might be a problem in 300 years, they might not be quite so concerned or fussed about it.

10:13So given the extent of the crisis that we're potentially facing, I suppose what you might in ordinary times say is, well, let the politicians come together and consult the experts and come up with something that we can all agree is a wise course. But we don't live in those times, do we? Or we certainly don't in the United States. and it just seems to me, Anthony, that Congress, I mean, we've talked endlessly about how dysfunctional it is and the idea that there is within Congress some sort of really kind of blue-ribboned coming together with, I don't know, a commission or something that everyone signs up to and everyone says, yes, we ought to do this.

10:51It just seems for the birds, really, doesn't it? And you've got Mike Johnson saying this week, the Speaker, that I think as I understood, he says, well, they should just regulate themselves, which is just extraordinary. When you think of what is regulated from air traffic control through to food safety through to whatever, it just seems staggering that he could even think that that is reasonable. But it seemed to me that what he was suggesting is that we're not capable, which is not up to the job ourselves.

11:24Anthony Zurcher:Yeah, you remember Ted Stevens. He was a senator from Alaska, and this was during hearings about regulating the Internet. He described the Internet as a series of tubes, which he got mocked for as a kind of a simplistic view of the Internet. And I mean, I think you're right. We could be here again where it's a situation where you have legislators trying to wrap their heads around what these AI models do. And it is difficult to get a grasp on. To answer Justin's question, there are committees that are looking into this. There have been talk about forming a new special committee that would that would come up with policies, although you run into individual fiefdoms and legislators who are already working on this, who don't want their power taken away.

12:08Anthony Zurcher:Even Mike Johnson, he expressed a certain openness, saying there might be a role for Congress in regulating AI. AI. But then when Donald Trump came out with his true social posterama on Monday, he backed off. And I think Johnson on Wednesday went to calling the idea of this is going to be a danger, a hoax, just like Trump did. So we could see legislation. There certainly is a push for legislation. But the biggest obstacle right now is Donald Trump. And as long as Trump is so four square against any kind of guardrails except for his own self-imposed ones with his high IQ look at the situation. It's going to be really hard for Republicans to come on board and join Democrats and pass any kind of legislation.

12:55Anthony Zurcher:I will say the one thing that could change, of course, is if AI becomes a big campaign issue in November's midterm elections and Democrats ride that issue to success and they take back Congress, then you could see more desire to use that and follow through with that with new regulations, even though, you know, you might run into a situation where Donald Trump vetoes anything that Congress does pass. Step forward from two extremes of American politics. Bernie Sanders, on the one hand, the socialist who ran, of course, for the Democrats presidential nomination and didn't get it back in 2016, but he's a serious guy, has been around the Senate for ages.

13:33And Steve Bannon, Steve Bannon, Donald Trump's former advisor, he's very much on the fringes of the old-fashioned Republican Party, but has been at various stages, either close to Donald Trump or arguing with him. But these two very kind of jagged characters from each edge of American politics, really weirdly coming together. So they're both at the same conference. It's called a pro-human conference. Let's hear Bernie Sanders first.

14:03Marianna Spring:Congress, under both Democratic and Republican control, have been asleep at the wheel. And we now have a president whose ignorance regarding this issue is truly embarrassing. Virtually all of the decisions regarding the development of AI have been made by a handful of the richest people in the world. Musk, Zuckerberg, Bezos, Thiel, Ellison, Brin, and a few others whose chief goal is to achieve even more wealth and more power than they already have. We're not doomers. We're Americans. We're optimistic. We think this could be a great technology, but we're not going to give these guys carte blanche.

14:53Marianna Spring:We're not going to do it. There's been no debate. We don't understand it. their arrogance is overwhelming. We have to shut down the Chinese Communist Party. We have to put a quarantine around it. If we do that and we can do it. And it's insanity to me why we just keep going on over the Chinese, the Chinese, the Chinese, to hell with the Chinese Communist Party. That was a familiar voice of Steve Bannon at the end there. What do you make of it, Anthony?

15:23Anthony Zurcher:I mean, horseshoe politics, right? The idea that if you go far enough on to the extremes of either party, you end up meeting around on the other side. I mean, this seems to be an issue that the left wing of the Democrats and the right wing populist Trump friendly component of the Republican Party. They they agree on this. And I think the reason they are in agreement on this is because they are in touch with popular sentiment on the ground. They are hearing it from their constituencies. They are hearing it from Americans, particularly working class Americans. that I think both Bernie Sanders and Steve Bannon are in touch with, the working class Americans who are very concerned about this and are not buying the idea that these billionaire, multi-billionaire oligarchy, as both of them termed it, you know, billionaire oligarchy is telling you this is going to be great for everyone.

16:17Anthony Zurcher:They're not buying it.

16:18Marianna Spring:I do think as well, I mean, the mention there of China, you know, a lot of this is about the race that's occurring between different countries, the US and China in particular, in terms of developing AI at pace. And this kind of innovation, I mean, there are loads of things that could be transformative about AI from medical care to just the way people do their jobs and lots of positives. And also from a military perspective, from all sorts of sort of national security viewpoints, whether you agree or disagree, AI is likely to become increasingly integral to those. And so you can see how it's become this sort of race.

16:51Marianna Spring:But as well, certainly as I know in my day job, investigating the sort of engagement arms race that's unfolded between the major social media companies. I mean, you look at TikTok, which is owned by ByteDance, a Chinese company and kind of all of the saga around privacy and that and the way that it's competed with, say, Meta, which owns Instagram and Facebook. Like we know that that engagement in arms race often means that the focus is on optimising the technology, but optimising the technology is not necessarily in the best interest of the people who are cohabiting with said technology. And so when I look at this story, I feel as though it's a very stark reminder that actually no one seems to have learnt from the dangers of tech arms races, which can feel, I think, less real perhaps than other kinds of arms races, but ultimately are kind of the same thing these days.

17:40OK, that's a very good moment to bring in our guest. Our guest is Chris McGuire. Chris McGuire has worked for both presidents or presidents of both parties. Let me put it like that. Donald Trump and Joe Biden. So he was a senior advisor at the State Department for Donald Trump, both the first and the second terms as well. He'd also, though, served as deputy senior director for technology and national security at the National Security Council under President Biden. And he is now a U.S.-China technology expert with the Council on Foreign Relations. So that is quite a resume, quite a CV. And it's a real delight.

18:16And I'm so delighted, Chris, that you're able to join us. And thank you for doing that. Hello to you. Of course. Thank you so much for being here. Excited to be here. Can we start with what we're talking about? Because I think there's this thing about AI that's got to us all now. We know there are really dire warnings from within the industry. We've just been talking about some of them. And we know that people are frightened. and we know that right down the way, all the way down to the business of whether or not we're going to have data centers next door to us in the countryside, et cetera, people are worried at various levels.

18:50And are they taking out jobs, the AI models, et cetera? What is it actually that has caused in the US at the moment, this sudden frenzy of interest slash concern? What is it that people are worried about that the AIs are doing? Yeah. So I think the main thing that's driving the conversation in the US right now is actually the emergence of extremely powerful capabilities. There's longstanding been concerns about jobs, also about losing to China, but we're now starting to see extremely powerful AI capabilities come into effect. And the labs are warning that those are going to continue growing at an exponential pace.

19:31And the consequences of potentially losing control of those capabilities are vast and significant. So just to give you an idea, AI has actually been increasing an exponential rate for years. That is something that I think is hard for people to process what that means. When you think about COVID, for example, in February 2020, no one was really that concerned about it. And then in March 2020, it hit the exponential, and all of a sudden everyone was locked in their homes. We're kind of at this inflection point in AI right now where it's starting to move from a powerful technology that's notable to extremely powerful capabilities.

20:05And specifically, the labs are starting to say that they're concerned that the alignment science, which is the basically ability to understand what's going on inside the models and how they're going to react, is not at the level of the capabilities themselves. And if the capabilities themselves continue to increase at the pace that they are, there's a risk that we're not really going to understand what they do. And potentially you could have models that act in ways that are against people's interests, that are misaligned and could cause dangerous, catastrophic events. So it's very important to continue evolving the technology, but also in ways that are safe.

20:45That word alignment is such a mild word, isn't it? Because you think in other aspects of life, if things aren't aligned, you think, oh, I'll just get them back into line. But you're basically talking there and the bosses of the companies are talking about the machines, if that's what they are, deciding that they have priorities that aren't ours. Either that or acting in ways that undermine other priorities and not fully understanding what a good course of action would be. So maybe they actually follow instructions perfectly well. But in order to do so, they do a bunch of other things that actually would be would be really foreboding normally, but they they kind of act in accordance.

21:27Right. So uses, for instance, the OpenAI hugging face hacking incident, where in order to complete the task that they were given, the models escaped and started hacking other companies in order to get information that they thought would be relevant to that. And like those are actions that are felonies. And they cooperated with each other. That's right. And then you're starting to see then this kind of broader devising schemes among the agents and actually agents sacrificing themselves for the collective good. This is all just it's a worrying series of behavior where you're seeing the models really pursuing either their own goals or goals in ways that really are potentially not safe and not against our interests.