AI: Boom or Doom? | EYES ON GEOPOLITICS

The Team House - dee takos

In this episode of 'Eyes on Geopolitics,' host Jonathan Hackett and guest Anthony Vinci, CEO of VICO and author of several books on AI and national security, explore the geopolitical implications of artificial

Key takeaways

• AI is not just software—it can persuade, influence, and even control physical systems like autonomous vehicles or military equipment.
• Chinese 'open weight' models, while free and accessible, are often biased toward the Chinese Communist Party and pose significant security risks due to lack of transparency in training data.

Main topics

• U.S.-China AI competition
• Security risks of Chinese AI systems

Notable quotes

"If you're using a piece of AI to run an autonomous vehicle, in theory, they might be able to control that autonomous vehicle."

Conclusion

The episode concludes with a call for cautious innovation, emphasizing that while AI holds immense promise for

Transcript preview

Speaker 2 (0:00) Bye. Speaker 3 (0:08) I'd rather live in the United States where we have control over the technology and information that affects Americans. And when you have tech like that used to be fine for basically a century because America made the best technology and most of the technology. Now we're entering a world where China makes a lot of technology and in some cases may make better technology, but they don't want to abide by our use. Speaker 3 (0:38) our constitution, our rules around this. So I would just say, fine, like you can do whatever you want in your country, but we are not going to use your technology then. Speaker 3 (0:54) Hey Speaker 2 (0:54) everybody, welcome to another episode of Eyes on Geopolitics. I'm here today with Jonathan Hackett and our guest is Anthony Vinci. He's the author of The Fourth Intelligence Revolution, The Future of Espionage, and The Battle to Save America. He's got a pretty amazing resume. He previously served as the first CTO and Associate Director of Capabilities at the National Geospatial Intelligence Agency. Following earlier work as an intelligence officer in Iraq, Africa, and Asia, you could actually watch an episode with Anthony on the team house that we did probably a good six months ago now. Great episode. Check that out. I'll put a link in the description. You can check it out. I also put a link in the description for his book. He is now the co-founder and CEO of VICO, an AI company focused on decision-making and finance and national security, and is an adjunct senior fellow at the Center for New American Security. And to boot, he holds his PhD in international relations from the London School of Economics. So I try to get somebody here that's on the level of smart as Jonathan Hackett is because Jonathan's got a pretty good resume too. And I think I did a pretty good job with you, Anthony. So thanks for joining us today. Yeah, man. Thanks for having me. I want this whole talk to be about what's going on with AI. Me and Jonathan have been talking about this for a couple months, putting together a show and trying to find a guest that can give us the lay of the land of what's going on. I've joked with Jonathan, I think I joked with you on an email too, about saying this literally could be a three-hour conversation because there's just so many rabbit holes we can go down. Speaker 2 (2:31) Full disclosure, I am more on the bearish side of AI, particularly when it comes to just all the CapEx that's going on and where's it really going, how expensive the compute is compared to how enterprises are using them. It doesn't really look like there's a big, at least for the frontier models, it doesn't look like there's a big profit center. right now um but anyway let's just get into it oh i mean i don't know how you guys want to play it jonathan uh please like let's just you know improv this basically do we want to explain like what the frontier models mean like should we go more of a basic route for like the layman i'm a layman obviously too but i just read Speaker 1 (3:15) this insensibly i think it's helpful to have like a more general question of okay why should people care about the crossover question of china and ai Because there's the China national security threat that's existed for a long time, especially international security strategy, all the way back to 1987 when the first one came out. Now AI has entered the chat. And the question for the American listener is, why should they care about that? Speaker 2 (3:39) Yeah, I love it. Let's go because we can go in any direction here. So, Anthony, what are you tracking basically right now with AI? I mean, I know I sent you over that Financial Times article and that does seem like the big scary. Speaker 3 (3:53) Yeah, well, look, China AI I think is, that's the question. That's like the generational level question because I think about it like this. AI is not like software. AI does what people do, right? And it communicates information to you. It doesn't just show you something, but you interact with it, right? So anything, when you get on Claude, you get on ChatGPT, and you ask it a question, it can ask you questions. You can continue to ask it questions. And the thing is that it can persuade you to do things. Right? Like it can lean you in a direction. That's different than anything we've had before. Only people were able to do that previously. So that's one thing. The second thing is that AI, that's kind of LLMs, like AI you talk to. There's another form of AI that you're using for autonomous vehicles that can drive Waymo cars around or can fly drones around. That's also super important. There's a third kind of AI. that I think is, you were going to see more and more of, and it's kind of what I do in my company at Vico, which is using AI to help run businesses or even to help run the military. It's making strategic choices. It's allowing you to play through lots of decisions, right? Those three kinds of AI, that's kind of how the world works. You have People talking to people, now talking to AI. You have AI that's able to drive cars and other vehicles. And you have a system that can run companies. Now, these are the things that we obviously are competing with China around. And the question is going to be, are we going to use Chinese software or AI systems to do these things? Are we going to use American systems? And then what about our... allies or just third party countries like that are neither aligned with the US nor China kind of sit in the middle. And what are they going to use? And we've seen this situation before with Huawei, with telecommunications equipment, where the question became, are you going to have your telecommunications system set up on Western equipment from Ericsson or something? Are you going to have it set up on Chinese equipment? The downside risk there was that Chinese equipment, they can spy on it and steal all of your data. It's completely unsecure. Well, we have the same situation with China right now. is that if you use a Chinese AI system, they can spy on it, they can access your data. But it's even worse now because they could do two other things. One, in theory, they could use that AI system to control what you're using it for. So if you're using a piece of AI to run an autonomous vehicle, in theory, they might be able to control that autonomous vehicle. That could be really problematic in war, for example. But the even more nefarious possibility is that Chinese AI, remember the part where you can talk to it, where you use chat GPT, where you ask it a question, say, about how to vote in this coming midterm. If it's Chinese AI system, they can mount an information attack against you and can actually persuade you to do something. So those are like really serious consequences for like the average everyday American citizen. Do you really want your vote, for example, in the midterm to be biased in some way because of your interaction with the Chinese system? Do you really want this Waymo car that you're driving in to potentially be controlled by China and say in a state of war to be used against you? Maybe it, I don't know, runs people over or something. Or American military equipment to be harmed by a Chinese AI system. And worst case scenario, if you're running your business and you're using AI to do this or the government is being run by AI, do you really want it to be compromised by Chinese AI system? And I think these are like really serious Speaker 1 (8:07) questions that we have to address. And I think also too, I was actually reading an article recently, I think this week about open weight models that China is kind of pioneering and getting out there a lot of, and people are seeing them, their lower costs and all these, there are all kinds of benefits and advantages on one ledger. And as I'm reading more about what open weight actually means, my first thought before I dug into this was that open weight was the same as open source. And I've learned that that's not true. Can you talk a little bit about why that's problematic? Speaker 3 (8:37) Yeah, the yeah, open weight means you're getting the open source means it's a piece of software that you can use and you can modify and you can do whatever you want with and build on. So like Linux, for example, open weight means that they have provided the model, but not the underlying data in the model. So it does mean that you can use it now and you can build that. You can build an app, say, on top of it. But you don't actually know the data that went into training it. You don't know exactly how it's going to respond. So the danger is that it could respond in a way that is biased. So kind of notoriously, the Chinese open weight models are biased towards the Chinese Communist Party. Like if you ask a deep seek model or Kimmy or choose your Chinese open weight model, a question about Tiananmen Square, it's going to give you the party line. Right. It's going to say what the CCP wants you to hear. And that's problematic on its own. And what but really what open way and open source mean is that it's free. And so what China has done is they've made these free models. deep seek and a couple of other companies have gone out and done this and so what's happening is they're essentially giving like this low-cost solution that u.s or other country companies can build on top of and so it's kind of similar to the huawei thing again right huawei was able to produce routers and other telecommunications equipment at a fraction of the cost of western equipment so people started using that kind of equipment and these these systems were compromised in a way or could be compromised by Chinese intelligence and so forth. Well, same thing with AI. They're sort of giving away these open source models that are effectively free so that people will build their own software and apps and so forth on top of them, but they're compromised. Speaker 1 (10:43) And actually, that brings up a good point. I was working somewhere over the summer where we got fable from Anthropic for like two days, and then the government shut it down. And there was this kind of disruptive moment where the government was trying to the US government was trying to figure out how do we regulate this? Do we put export controls on it? How dangerous is it and all these questions. But at the same time, China is looking at this. In the opposite way, they're trying to get as much access to people as possible, whereas the U.S. seems to be trying to be more careful. What's kind of the right way to balance these two competing philosophies? Speaker 3 (11:14) Yeah. So there's a downside to letting AI models out, right? Which is that they can be used to do dangerous things, right? You could take an AI model and in the case of Fable and Mythos, the Claude models, that they could be used to hack, to perform an offensive cyber attack. And that could be really dangerous. Imagine you just sort of let everybody in the world have that and now these criminals criminal hackers were able to use this stuff and then they can go and hack into banks or into healthcare systems and so forth. Could be super dangerous. So there is a debate just on that. Like when and how should we release these models? And it gets worse, by the way, than just hacking. As these models get more powerful, it's possible you could use them to create like a bioweapon. Right. That a person with little to no biological science experience could actually create a bioweapon by following the instructions like you would with, you know, a cooking recipe from an AI system. And that like that's like terrifying to even think about. So we want to have some control. China like does have some control. Even on the open weight models, right, the CCP has established guidelines and controls to those companies like the DeepSeek, for example, around information that they want certain information about Tiananmen Square or the Uyghurs, which is a minority group that's persecuted in China. They want to control that information. So they are putting in those controls. But then these open weight models are being released to everybody and. they don't have, there's no way for the U.S. government to put its own controls, like say around cybersecurity issues. Now, this wasn't a problem until recently because the U.S. models like Claude and ChatGPT and Grok and so forth were by far the most powerful models and could do the most powerful things. And these open weight models were, you know, a generation or two generations behind. And so they weren't as dangerous. But now they're sort of catching up. And as they catch up, they might get as dangerous as a Claude model for hacking and for other purposes. And there's no way for the U.S. government to regulate it or control it in any way. And in theory, China might let people, you know, allow a model out that could be used for cyber hacking. Speaker 2 (13:45) What do you guys take? What's your take on, you know, Microsoft announced maybe a few weeks ago, also Microsoft being a big investor into open AI, that they're going to start using DeepSeek or at least look into using DeepSeek for their some of their enterprise stuff. I mean, that's got to ring alarm bells for everybody, you know. Speaker 3 (14:04) Yeah. I mean, look, my personal take is that just like Huawei, it should effectively be made illegal or at least very problematic for a U.S. business to use these things because for all the reasons I mentioned, like these things, they're compromised in that sense. And there's a danger in that. They're already large American companies using. Chinese open weight models, Airbnb uses them. There's a few other big ones. And then there's lots of small startups that are using them. Microsoft saying they're going to use it is not great. I suspect that really they would be using it for something that they consider to be low security risk. And that's fine. But then who's making that call about what's low security risk? And we are in effect Speaker 3 (14:58) creating the same situation as Huawei, where we're going to kind of build up their open weight model companies over time, and then we won't be able to compete. You Speaker 1 (15:09) mentioned something really interesting. You said like Airbnb and these smaller companies are using deep seek. You know, I used to teach at our human intelligence collection course and we taught about ubiquitous technical surveillance and how to do things under a UTS threat and so on. That was before AI. And now I'm very curious about how it's not just a vector. I mean, in ubiquitous technical surveillance, you have multiple vectors that overlap that make the thing a threat or a problem for you. And it seems that. It goes beyond the vector problem when you're using AI. Let's say you're booking on Airbnb and there's certain behavioral patterns that you, the user, use on Airbnb that are almost exactly the same behaviors or exactly the same behaviors when you're undercover under a different persona on some other platform. And AI is able to see that and correlate you immediately way faster than a human analyst could. I already mentioned, what would you say to the average American? But I'm curious also to the average case officer who might be listening, because I know there are some who do listen to the show. What would you kind of caution them on? Speaker 3 (16:07) Yeah. I mean, well, look, even for the average American, by the way, I don't think the average American wants a Chinese company or a Chinese security service or the Chinese government to know what they're doing and to be able to match the records from Airbnb to other things in their life. We don't want our government doing that. Right. Like your average American like me is like. at least a little bit libertarian leaning in that sense. We don't want our government to spy on us. Why would we want the Chinese government to spy on us, right? So I think everybody, but yeah, for case officers, look, I was a case officer, you know, and by the time I stopped doing that, the UTS issue was just starting, right? And so I only really saw a little bit of it. And I can't imagine starting from scratch today, what it might must be like. And the, you know, even without AI, you can match people's patterns, right? Just literally like old school, big data analytics. I crunch enough data about you. I can detect patterns. The, the AI makes it easier and more ubiquitous, right? Like I don't even need to, you know, I can, I can do it like that example you brought up is a great example, actually. Like my, my habits. It might not, by the way, even be what I'm doing on Airbnb. It could even be like the way I move my mouse or the way I type, right? There are signatures that are, everybody has. These patterns can be taken from one place and then, and then apply to another. So as a case officer, you're trying to act and cover somewhere. And, you know, even if you. even if you used a false name for everything you did in work, and in fact, even if you used a false name for everything you did in your commercial life, well, your pattern still might be there, right? Like maybe you're using your wife's account to log into Airbnb, but you're the one on the mouse and making the decisions. They would still be able to recognize these patterns. So it's like becomes nearly impossible to kind of like. Keep yourself out of there. Now, I actually wrote a whole chapter about this in my book because I think where it goes to be a case officer, I think you just have to sort of assume that anything, any pattern is or could be tracked. And then I think we will enter a world where you just have to sort of live in pattern. And what it is to kind of be in cover and so forth completely changes. Like there's no more dock swapping and all that stuff we saw in spy movies. Like that's over. And like now, now you just have to sort of live your life as that person, like kind of closer to what it was like