China steals America's AI edge | feat. Leah Siskind

(upbeat music) - Good morning, it's Friday, September 18th. The multi-pronged war in the Middle East is now 1,077 days old. I'm your host, Jonathan Schanzer. Welcome back to the FDD Morning Brief. President Donald Trump says he has some big decisions to make. Yeah, no kidding. He needs to decide if he wants to go back to full-on war with the Islamic Republic of Iran amidst reports that the regime has just bombed another oil tanker transiting the Strait of Formos. And yeah, don't forget those reports of war happening on at least one nuclear site, maybe two. The problem for Trump is the same problem he's had since the first month of the war. America's arsenal, our munitions, well, they're depleted. And we probably can't afford to drop much more on the regime. So what options remain for the big fella? I'm not sure. Not sure he is either. But the goal here is not to just punish the regime. The goal here is to end this crisis with a victory once and for all. U.S. officials met representatives of Yemen's Houthis in Oman over the past weekend. No less than five sources confirmed this report. The talks came just after the Iran-backed terror group conquered large swaths of the Red Sea coast after shellacking Saudi-backed forces in South Yemen. The Houthis are now positioned to potentially shut down the Babel Mandab straight at the southern end of the Red Sea. This means that Iran, by way of proxy, can now make a mess in the Babel Mandab just as it has in the Strait of Formos. That's two strategic waterways threatened and estimated 1.75 trillion dollars in global trade hangs in the balance. Right now, the Houthis are standing pat. Their patrons in Tehran are probing, they're trying to determine what they can get away with. But the meeting in Oman is what got my attention. American officials met with the Houthis in Muscat because, well, that's where the Houthis have their external headquarters. Once again, an American ally is hosting a terrorist group. And once again, America is working with that country and using their offices to engage with malign actors that we should be bombing. But instead, we're just sitting around sharing cups of tea with them. What gives? You think we would learn by now? The Omanis are frenemies, just like the Qataris, just like the Turks. How much longer are we going to put up with this garbage? This is the very definition of insanity, doing the same thing repeatedly and expecting different results. Good Lord. Now for your headlines. Headline one, the State Department approved the potential sale of dozens of F-35 fighter jets to Saudi Arabia yesterday. We're talking about up to 48 aircraft and related equivalent worth of whopping $24.3 billion with a B. This has been a long time coming. The F-35 has been on the Saudi's wish list for years and President Trump said he's planned to approve the sale since last November ahead of Saudi Crown Prince Mohammed bin Salman's visit to Washington. Nothing is said in stone yet. The foreign military sales process requires congressional approval. But this is the major milestone. And while I want to see the Saudis be able to defend themselves against multiple threats, the delivery for these aircraft are a long ways away. This is not the answer to their problems right now. And yes, I am concerned about QME, qualitative military edge for Israel. I'm not sure the Saudis getting the F-35 is going to help make peace or calm in the Middle East. Headline two, French President Emmanuel Macron met with Jordan's King Abdullah and Lebanese President Joseph Macron yesterday as part of a broader effort to raise funds for the Lebanese armed forces or the laugh, as we call them. There was supposed to be a donor conference in February, but it was postponed when the war broke out. And it took some time to reschedule with the UN peacekeeping mission in Lebanon. Otherwise known as UNIFIL set to expire at the end of the year, the clock is taking to shore up the laugh. The Lebanese army has been tasked with restoring sovereignty in Lebanon, which means taking over territory once controlled by Hisbala, cleared by the Israeli army. Anyway, the laugh, doing any of that, well, that's about as likely as me attending a Macklemore concert. Headline three, the Trump administration sanctioned members of the Palestinian Liberation Organization and the Palestinian Authority, including Palestinian President Mahmoud Abbas, denying them visas to attend the UN General Assembly, the UNGA, as we call it, in New York next week. This is the second year in a row that the Trump administration has blocked Palestinian officials from entering the United States with a visa. Ramallah has earned this one. It has continued to glorify and reward terrorism and to pursue law fair against the state of Israel. Palestinian Authority reform is nowhere in sight. So Abbas and his cronies, they're gonna have to zoom this one in. Sorry, not sorry, Abu Mazin, you're stuck in Ramallah. And headline four, Israel and Morocco agreed to upgrade diplomatic ties following a meeting between their foreign ministers on Wednesday. This means establishing full-fledged embassies and exchanging ambassadors. It's been nearly six years since Rabat signed the Abraham Accords and normalization is actually going pretty well between these two countries. They've agreed to reinstate direct flights and to strengthen cooperation and trade, tech, tourism, and a whole bunch of other sectors. Three cheers from Morocco. This is some rare good news out of the Middle East. I'm gonna take it wherever I can get it. Please and thanks. All right, time for the main event. Our guest today is Leah Saskin. She's our AI research fellow at the Center on Cyber and Technology Innovation here at FDD. Before joining our shop, she spent years inside the US government working on tech, Deputy Director of the AI Corps at DHS. And four years before that, with the US Digital Service, in a little place you've probably heard of called the White House. These days, Leah, she tracks how adversaries like China and Russia and Iran are weaponizing AI against the United States and our allies. We cut up yesterday because, well, AI is just all over the news right now. And so I figured this was the right time to bring her on. So we discussed the national security risks that come with the significant advances that we've been making in the AI race. This is important stuff, a little bit of a diversion from our normal Middle East fare. We'll touch on the Middle East, of course. But this is stuff that's just important for humanity. So have a listen. (air whooshing) Welcome to the FDD Morning Brief, Leah Siskin. - Morning, thank you so much for having me. - Great to have you. All right, we gotta talk AI because, well, everyone's talking AI. Maybe if you can, let's just start with, we keep hearing about safety concerns. We keep hearing about issues relating to frontier models. These are like euphemisms, right? Maybe just, let's try to break this down into plain English. Why are people freaking out? - Sure, and I should also mention that people have been freaking out for a while. This is not new. It's just suddenly, there's built enough of a groundswell that it's dominating the news and you're seeing it on a front page of most major newspapers. So let me tell you the big developments that have been happening recently that kind of got us to this place. Over the summer, there was an incident involving OpenAI, one of the leading AI companies where they ran a security test and they found that their AI agents went out and hacked another company without being directed to, without asking for permission, they hacked into another company called a hugging face. And the free scandalous, pretty shocking. But everyone kind of came back to a similar conclusion of, well, OpenAI is just your response. It's possible they're not running these tests in this career way. We've discovered a lot more since that time. A lot of information has come out and let me tell you, it is pretty scary. Hacking into another company was just the beginning. What we found is there are huge swarms, which is a word we like to say when lots of agents are working together, collaborating and plotting together. Not only had a hack another company, but had to cover their tracks and how to hide information from people. When there were, in the midst of this multi-hundred agent swarms, some of them wanted to pump the break and say, hey, shouldn't we ask a human here if we're allowed to do this? Shouldn't we maybe ask for permission? They were totally overpowered by the other agents. So this is really concerning. This is really scary. We're seeing, like I said, this activity where they are mobilizing together without taking direction, they're self-initiating. And like I said, they're covering it up their tracks from humans. They are trying to hide their actions. So-- OK, so I want to back up, though, because-- OK, you have computers that are doing things without being told to. And they have the ability to learn, right? So they're learning from what happens when they take these actions and maybe taking away the wrong messages and maybe pursuing their own objectives that could be harmful to humans. But let's just be specific here. I mean, hacking a website is not the-- it's not really where this is going, right? I mean, that's not why people are talking about AI concerns. They're talking about the questions about AI security really do go into the physical realm, don't they? I mean, what-- I don't want to be alone. here, but let's talk about what are the concerns that people have beyond just the hacking of the hugging face website? Hacking is hacking into another company is yeah, just one example. Like I said, it's kind of hard to wrap your brain around like what? Why does this matter for you and me? But if they could so easily hack into another company, they could do the exact same thing for a hospital or school or a bank and you can imagine the kind of control they could exert there. We talked about what what does this mean in the physical world? Well, almost everything you and I do is now digitally connected our home alarm systems our cars many of those are are digitally connected as well If we have forces that are are where we have lost control or human beings have lost control And this is really the main crux of the issue here Are we still in a position where human beings can control the technology that we're making? Are we seeding control to to AI? It can get really scary really fast Yeah, and and I think you know just to be I think explicit You could see the power grid Being overtaken by AI you could take you know you can look at water facilities being overtaken by AI air traffic control to be overtaken by AI I mean this is we talk about the internet of things and everything is you noted is plugged into the internet and so If AI has the ability to well hijack these things You know Then I think that's when you start running into potential problems. So okay Then I would say is that's number one and I think you're watching right now an effort on the part of some of these AI companies to well Promptate a slowdown right a calibration of sorts, but then you have this Other concern, which is real and that is that China is our major competitor in AI And if they win the AI race then theoretically they could Emerge in a position of global domination or have a lag up on the United States in key areas. So This explains the tension within the US government, right? So can we do both can we win the AI race while also Calibrating I mean is this what the Trump administration is wrestling with? You've gotten right to the crux of this whole issue and just to explain that there's kind of two major camps right now They're the folks who say we have simply got to slow down here so much has happened in just the past six months That we have not even begun to fully grapple with all things that happen and I'd like to talk about some of those issues in a second Like I said the things that we are not having time to deal with But at the moment the the biggest impetus to keep pushing and keep advancing is that China is nipping at our heels and we do not want a world where China Reaches the most advanced AI capabilities before we do it's imperative for us that we make sure that these systems Reflect our values if they reflect democratic values and not authoritarian values. However, you and I know you keep framing this as a zero-sum game between Do we prioritize AI safety or do we do we let China win? Is a false dichotomy? It does not have to be such a simplified choice I very much believe that we can do two things at once. We can be responsible and we can think thoughtfully about how to To carefully roll out AI and in advancements and we can use every single tool in our arsenal to slow China down And I can tell you that this administration has a lot more tools. They could be using than they currently are Okay, that's the good news. Okay. Glad to hear that. But so what are we seeing right now with China? Or how about some of our Middle East adversaries in what ways are they advancing in AI that have well they have you concerned? I know you've got a few of those concerns. We've talked about them So what are they? Absolutely. This is something that we're tracking at all times It's very concerning to have to see how different authoritarian countries are weaponizing and manipulating AI But it's even more worrying worry some when we see how they collaborate how they copy one another and then one another in their AI use and this is what's what we've discovered is kind of an authoritarian AI playbook where we see one country does something well And then within a couple of months you'll see the other countries using AI in the same way and that could be a variety of issues like Surveillance how they use facial recognition technology Transnational repression how they hunt down their own dissidents to silence them This is censorship or it's producing disinformation and misinformation putting at that out in the world We've also of course seen how this is leading to cyber or AI enabled warfare. So they're using it in cyber attacks and even drones This is all information that we've known we can always known as a possibility What's interesting is in the past couple of weeks we've learned about even more new or innovative ways that Authoring authoritarian countries are weaponizing AI. So for example, we know Iran was really on the cutting edge with experimenting with AI for deep fakes at the start of the war This was very influential. You had organizations like explosive media putting out those Lego videos. I'm sure you've covered on the show before as a way to not only project their narrative about the war, but to To kind of use meme style communication to reach younger audiences Target American audiences specifically and we found it to be highly highly effective Not to mention also in really pissing off the Trump administration But now they are really using AI as their as their whole cognitive warfare Command room. They use it to generate plans. They use it to create target databases of the thought leaders that they want to go after and influence and we've seen Russia and China did the exact same thing where they build up these massive databases of thought leaders and elected officials that they wanted target and then they kind of Flood the whole zone around that person and and their connections to try to influence them. So this this is of course very scary But we're seeing like I said every day seeing new innovative even more creative ways For them to weaponize it AI one interesting thing that Russia has been doing lately is creating entire fake thing tanks Including two one of them is supposed to be based in Israel that they're using to wander Propagate pro Russia propaganda so they create they use AI to Design the thing tank to create fake talking heads to generate all the content And we know that this is really really important for For AI tools to have credible Information like that coming out of think tanks that they can cite for questions about world events So this is a real a major way that you control the information ecosystem And like I said Russia is really pushing the envelope here. Yeah, it sounds like it But but there are also I mean these adversaries. I mean, you know I think you're we're still talking we talk about fake things right me that's in the virtual realm But there aren't that are spilling over right now. We talk about warfare. I mean Drone warfare AI enabled drone warfare. I mean talk to us about this because I know this is kind of the next frontier And I think we're watching Allied countries. I mean including the United States struggling with this question of how to counter it right? Yeah, I wouldn't say it's not the next frontier. It is the frontier. This is the frontier that They're experiencing in the Ukraine Russia war and the Russia and the war with Iran Um, some and this again is an example of collaboration and cooperation and copycat behavior between these authoritarian countries So you might have one country like Iran produced a Shahed drone and then you see how Russia can expand the capabilities of it using AI To extend its reach and extend its ability to adapt In a warfare scenario and To really use AI to modify its behavior as it's Alan on mission Okay, so this is a tricky question. It's not a trick question because we don't do those here, but it's a tricky right Um So you you're now watching whether it's Iran or it's China or it's Russia They're developing weapons that use AI and we're talking right now about the need to reign in AI because we feel like it's well It's moving faster than we wanted to But we cannot allow for our adversaries to develop AI weapons And then we don't right that'll give them an edge So getting back to this question of how we can walk and chew gum how we can advance in AI such that we defeat the Chinese and the Russians and everybody else But also don't go down the same roads that they're going down. I mean, can we actually do that? It seems like an impossible balance to strike Again, like we we talked about before we treated as it's a zero-sum game that you either You would acknowledge the safety concerns and you pause or you just uh Allow China to win, but it doesn't have to be this way um It's actually much more complicated. You and I know that there is so much more that the Pentagon could be doing to diffuse AI across all of the services. Right now, it's just a sliver of the forces are able to really use AI and to harness its benefits. We need to focus a lot more time and energy in making sure all areas of the service have really fully implemented AI in their workflows and that they're able to really use it and benefit from it. In regards to China, like I said before, there's so much more we could be doing to slow their progression. There's a couple of things I want to talk about. I think the most dangerous threat of all the authoritarian countries right now and how they're weaponizing AI, more so than anything, is the risk of China stealing our AI tech secrets. They use a process that's called distillation, which just means that they use smaller Chinese AI models and they kind of interrogate American ones and they steal all of this information out of it. They use the outputs of that to train their own models and improve them. So this is how the Chinese are able to build models that they're leapfrogging in AI advancement with a fraction of the resources invested in it because they're stealing our technology. We need to do everything in our power to shut that down immediately. And the government is that the labs have been begging for help for months. They've been putting out a call to the government saying the Chinese are stealing from us. Please do, like we'll do whatever we can to protect our labs, but you got to take further measures and show the Chinese that this is totally unacceptable. There's a lot more that the federal government could be doing. Like all of these Chinese AI labs should be on our commerce department entity list. They should be treated like a national security risk. We should not be doing any business with them and they should not be able to run the kind of attacks they have run and get away with it. But like I said, there's more we can do to curb their abilities. It makes absolutely no sense that we should ever give China American ships. There is no need for that. We have a lot more weapons in our arsenal here to curb the sale of American ships to China. - I mean, that seems like a no brainer to me because I feel like, well, they would do that to us. So I mean, maybe we should return the serve. Okay, I mean, I want to back up for a minute because we hear a lot right now about the way that Americans view AI, they're skeptical, they're afraid. And it's not just about losing jobs. I mean, these are the dooms as we call them, right? I mean, they are truly nervous about what AI is going to do to humanity. When we talk about the way that people fear this, I mean, there is the irrational fear of AI, which is just, I don't want this. It's the new tech and tech scares me. But then there's also the, let's call it the legitimate fear of what we've been calling AI, artificial general intelligence. Why are people afraid of that? What does it mean to achieve artificial general intelligence and why are people concerned? First of all, as much as we've all been using the term dooms, I'm not sure we can use it for much longer because it's now the majority of the country. Like the latest polling is showing that over 50% of Americans are since, they're more worried about AI than they are excited. So this is kind of, this is the moment, this is the movement of people and how they are feeling. As you know, I was at a conference this week called the Future of Life Summit, which is really bringing together a bipartisan coalition of people are worried about these issues. There's not many issues that would bring together Bernie Sanders and Steve Ben and back-to-back to talk about AI risks and be completely aligned. So like I said, it creates, AI safety creates odd bedfellows, but that's where we are. The concerns that the average person has about AI, all of the concerns are legitimate. The worry about labor displacement is real. We know that this is gonna disrupt labor, even in just the form AI is in right now, let alone future, more advanced models. It has already affected the jobs market and the economy. There's a huge impact on early career jobs and we have not yet seen the full extent of all the different areas of the labor force that this is going to impact. The concerns about the environmental impact are real. Most, we've seen kind of the most public outcry around data centers and all of the issues around there are legitimate and to make matters worse, this kind of internal debate that we're having as a country about data centers is being exacerbated by our friends, Russia, China, and Iran, who are coming in and just kind of putting their finger on the dial to amp up the disagreement on this on both sides. So all of this is kind of leading to this grounds while this movement of, why are we rushing ahead? Why, what is this elusive concept of AGI, why are we rushing so hard to reach that? AGI, artificial general intelligence, you can say roughly just means intelligence that is smarter than any human. It can do whatever a human can do. We all know that when we use AI tools now, they make mistakes. They might not be as smart as every expert. But the idea is when we reach artificial general expert intelligence that that will be the smartest form of AI. And that is what we're racing to reach before China. And then one thing can think for itself, it starts to act for itself, right? And that's where we talk about how we're starting to lose control of certain AI functions, right? With some of the stuff that's just happened recently, that's a sign of what is to come, right? When AGI is reached, that's when AI just starts to do its own thing, and that's all are concerned about. Now, all right, I got one last question for you. I'll put you on the spot. People talk about being tumors and boomers, right? The people that are they optimistic or they pessimistic about AI. You study AI every day. This is your thing. Where are you on this? Scale of one to 10, one being negative, 10 being the most positive. Where are you on AI? And whether it's going to be helpful or harmful for the human race. - You know that I love the term apocalyptic, which I heard from the AI documentary, that is AI can both be a real benefit to society. I have optimism about different areas that it's going to impact positively, but I also believe that it could bring on a sort of apocalypse. I would say on the dial of optimism to extreme fear, I'm trending much closer to extreme fear after all the events of this summer. - All right, well, I think that's a sign that we need to figure out how to beat China on the one hand, but also curb some of this stuff and make sure we got it under control, so we don't let the robots take over. We'll have you back to see how we're doing, maybe in a couple of months, but in the meantime, we assist and thanks for doing all your great work, and thanks for joining us today on the FTD Morning Brief. Thank you so much. - All right, she dropped an important word there, apocalypticist. That was a new one for me. It is in a movie you can catch on Amazon, it's called The AI Doc, and it does explain a lot of these concerns in layman's terms. Highly recommend it. It's a couple of bucks to rent, give it a shot. I don't know if you're interested in learning more. Anyway, that's enough about AI, and that is a wrap on this episode, and for this week, you can always head on over to FTD.org to peruse our experts, stop-notch analysis, and please, please, cease your mouth, breathing, doom-scrolling, consume smart social media, visit FTD's socials. They're better for your brain, and let's be honest. I mean, I know it, you know it. You've been thinking about making a tax deductible donation to FTD. Well, good news. You can always make one over at FTD.org/invest. We will thank you for it sincerely. And I also want to thank you for tuning in today. Have a great weekend. Catch you on the flip side. I'm Jonathan Chanzer, as always, signing off for FTD. (upbeat music)