Hosted by Christina Ruffini and Sir Richard Dearlove
Guest
Episode Summary
In this week’s One Decision In Brief, hosts Kate McCann and Sir Richard Dearlove, former MI6 Chief, discuss how Israel is deploying AI on the battlefield and how the technology is intercepting calls, mapping behavior, and even generating target lists. They examine the two most common AI models employed, called "Lavender" and “Gospel,” how they’re driving innovation and whether international law can keep pace with this new era of machine-driven warfare. Is this the future of war? What happens when algorithms, not humans, are shaping life-and-death decisions?
Episode produced by Situation Room Studios. Original music composed and produced by Leo Sidran.
Listen & Subscribe
- Watch on YouTube
- Listen on Apple Podcasts
- Listen on Spotify
Transcript
INTRODUCTION
Mccann: Hello and welcome to One Decision in Brief. I'm Kate McCann, and I'm joined by my co-host, Sir Richard Dearlove. Hi Richard.
Dearlove: Hi Kate.
Mccann: I suppose broadly, the heart of this discussion is about the fact that war drives innovation. And we have spoken before, haven't we, about Ukraine and especially drone warfare and how much that is changing the way that war happens and the frontline and the way that, you know, armies engage with one another. But there's a really interesting case of this in another conflict, which is in the news, which is of course, Israel and Gaza. And this is AI. And I know a lot of people will be familiar maybe with the idea of AI being used for targeting, for airstrikes, for sometimes even facial recognition. But what Israel is doing is also about gathering intelligence and working out patterns of behavior. And that is about large language models. It's about facial recognition. Yes, it's about chatbots and gathering information on phone calls, text messages, looking for patterns of language and behavior and running them through systems which can almost start to predict who or what might be a threat to Israel. Now, for lots of people listening, you might think, well, that sounds like a surveillance state. And I suppose that's exactly what we are talking about here. But Israel is using this really effectively, particularly they are using it in the West Bank. And some of the sources from this elite group that is working in the Israeli military are saying this is the future of conflict. What is this Unit 8200? What is this unit, and what do they do?
ANALYSIS WITH SIR RICHARD DEARLOVE
Dearlove: It's all sorts of techniques to collect data on every aspect of the battlefield in Gaza. But I mean, the battlefield, of course, includes the civilian population, and the enemy is embedded in the civilian population. So ultimately, they are gathering massive amounts of data and applying AI models to analysis of it. And I think everything that moves, speaks, everything there is susceptible to collection in one way or another. I don't think people necessarily realize, I mean, even going back many years, the importance of intercept and the importance of analyzing the data from intercept. Just think of all the Palestinians with their mobile phones phoning each other, whether it's a family matter or whether it's a shopping list or whatever. All of that is relevant to the battlefield, and I think that what the Israelis have almost certainly managed to do is to crack the issue of collection and crack the issue of analysis and putting all of this material together. And that leads on, of course, to the issue of targeting and how you then deploy your resources militarily to counter the threats. So it's pretty damned important, and I'm sure the scientific area of analysis is being changed and catalyzed massively by the Israeli experience.
Mccann: Let's talk about two of the AI models that the Israelis are using, because they are interesting and they are quite specific. One of them is called Gospel, one of them is called Lavender. Now Lavender is essentially a kill list. It's essentially a group of targets or suspected combatants. It's used by Israel. It has been used at the start of the war in Gaza to help choose attack targets. And Gospel is this sort of AI-guided tool that is cross-referencing information that they've gathered from different sources. So that could be things like voice intercepts that we've just been talking about there. It might be facial recognition. It could be different sources to suggest, you know, looking at maps and intelligence to look at where people are moving and pinpointing that using data from mobile phones. Those two AI models together, which are essentially behavior modeling, patterns modeling, interacting with this kill list. And then another sort of tool which allows them to track people's movements through their phones. It's making it very easy. This is what experts are saying for Israel to target. But it also throws up this question, especially when we're talking about a war zone like Gaza, of the legality of using this kind of information, first off, and the collateral damage, second, because we've seen it firsthand, haven't we? We've seen Israel target — pinpoint target in some cases, people, but there are huge civilian casualties around that. Where do we stand on the legality of this kind of tool used in this kind of way?
Dearlove: Well, I don't think that the law, or let's say international law, which is very evolutionary and controversial, remember, has advanced as fast as the technology has. So the sort of lock-in of how we view battlefields and civilian casualties and what's actually happening, I mean, in a way the two have become separated. My own view is that there has to be some rethink. I mean, bear in mind that one of the problems in Gaza, and this is right from the start of the conflict, is that Hamas prepared a conflict if and when it should happen, and of course it did happen, in which they would make sure that the civilian population was a shield. They couldn't be attacked without massive damage to the civilian population. So there is no front line in this warfare. There is a sort of complete mix of civilian and military capability. This is one of the reasons why, of course, Hamas is characterized not as an army but as a terrorist movement because of these tactics. I'm sort of, in my own mind, really confused. And of course, if you go back to World War II, let's say the last six months of the conflict in the early months of 1945, there was no question that anyone worried too much about the civilian population in Berlin as both the Russians and the Allies advanced on the city for the final defeat of Hitler. And in fact, what's happened in Gaza is pretty catastrophic because every time the Israelis attack a Hamas target, they're making a calculation, as far as I can see, based on AI from what we now know, what level of civilian casualties is acceptable to them, I'm not saying to the international community, to kill a Hamas kingpin. They're gonna kill 50 people to kill one Hamas kingpin. And I think this is what's probably happening with the AI calculations.
Mccann: And there are suggestions that targeting officers in the Israeli army have been given numbers in the sense that they've been told, you can kill X number of non-combatants, you can kill X number of civilians, depending on how senior the person is that you are targeting. I suppose the bigger question for Israel and for all of us is Israel is essentially the front line of utilizing this kind of technology at the moment. And whatever you think about the conflict in Gaza, you know, this is technology that is going to be used more broadly than just in this one instance. The IDF, the Israeli Defense Force, the argument there is that these tools are assisting humans and that humans will always and do still maintain the final sign-off. And so the laws of war apply in the same way that they do at the moment. It's just a precision instrument. But others argue if AI is so sophisticated and if it's able to run through all of these different models and work out, okay, this is the person we're looking for, we think they're going to be here at this time. And the AI is almost building the entire argument for a specific target. Where does the AI stop and the human come in? And if it is the AI that's making these decisions, because we do know that there is capability for AI to make these decisions without human interaction, well, how do you prosecute an AI for a war crime? You can't.
Dearlove: Well, you don't, really. I guess you accuse the regime using it of general violation of international law. But I mean, you can imagine a situation, let's say with Gaza, where you've got a fleet of loitering weapons, drones of some sort or predators that are continuously flying around this city, being fed analytical data by AI and making instant targeting decisions based on the data that they're being supplied with. So one can see very, very easily how this evolves into a sort of robot-driven battlefield. And we are getting into some really complex ethical areas. I mean, I have to admit that in terms of the way a lot of people apply and think about international law, which has very honorable motivations, it's just incredibly outdated. It doesn't sort of take account of realities on the ground at all. And I think we're moving in that direction. And bear in mind, the interpretation of international law, for example, in the United States amongst certain American lawyers, is quite different from the way some Europeans view it. And I'm sure the way Israelis view it as well, they probably fall between these two versions and they have their own views.
Mccann: I suppose that the other question for us to grapple with, as if that one isn't big enough, is the role of tech companies here, because tech companies obviously are partnering with countries like Israel to develop this software, this data, these tools. There have been employees of some of these big software companies who have suggested that even the staff in those companies do not have access to or cannot access the data that is held in these big AI models. I mean, Eman Hasim, former Google software engineer, who was fired after a sit-in at Google at protesting the war in Gaza, said that she had worked on a team that helps test the reliability of a sovereign cloud for Israel. So that's a secure system of servers. But it was kept so separate, according to this employee, from the rest of Google's global cloud infrastructure that even the company couldn't access or track the data it stores. So if there's no ability from within a company to track, and there's no ability for another country to track and there's no oversight in international law at the moment because these new systems are advancing so rapidly. Well, where is the oversight and what's to stop? But some humanitarian organizations are suggesting journalists are being targeted, that you could easily feed into these models. These are the people we're looking for, or these are the people we wouldn't mind taking out, you know, and they're journalists and that is illegal. You're not allowed to do that.
Dearlove: I agree, but once you are into, let's say, a conflict which one could characterize as total warfare, I don't think any of these legal concepts are very easily applied. I mean, we apply them in terms of looking in from outside, but if you're inside and you're part of that conflict, I think ultimately, whether you're a lawyer or a journalist or whatever, or a medic in a hospital, or a child. I mean, all of these people we know quite well from Gaza now are extremely vulnerable in the sort of conflict that has developed, which is a version in my book of total warfare. But shouldn't we try to police it? Well, I think the general moral feeling is yes, but when there's no distinction between who's a soldier and who's a civilian, this becomes incredibly difficult.
Mccann: How do you think if these technologies that we're talking about, these AI models, these facial recognition models, the chatbots, the tracking, I mean, some of it would have been available during your time as the boss of MI6, but with the advances that we've made, would it completely change the game in your world?
Dearlove: To an extent, it does change the game because the way to good intelligence is along a data road, okay? But it still doesn't, as it were, push aside the importance of human sources. In the intelligence world, there's a combination of these assets. I think what's happened is that the data-driven analysis, the data-driven intelligence, has become much, much, much more important. And I would say that the well-placed human source is still very important, but the others have risen up to an almost equal standing. It's a world which is in tremendous flux. I mean, it's being driven by the technology revolution. You know, if you talk to some of the big names in Silicon Valley, I mean, they're imagining a world which in a way is almost run by tech companies and not by governments. The speed of tech advancement and the people who are the brains behind it are more important in the future world than any government, and that ultimately their decisions will displace a lot of government decision making. I mean, I don't subscribe to that view myself, but it's definitely being written about and theorized about at the moment.
Mccann: Do you think the world is just more dangerous now? Or do you think the UK should be developing the same kind of technologies that Israel is working on? Is that something that would make us safer?
Dearlove: Yes, I mean, I think we, in the UK, to make our expenditure on defense effective, we have to conceptually digitize completely the future battlefield, and in a way that makes us safer. I mean, you could argue that that's an essential part of an effective defense policy. The irony is that the UK has been working on this slowly — the keyword is slowly — for a very long time so it hasn't made very fast progress because it hasn't had to. Whereas if you look at the experience in Ukraine, the Ukrainians within a year of the conflict have digitized their battlefield almost completely. And the irony now is probably the European leaders in this whole area are Ukrainians. So what's happening, despite the difficulty of the conflict in Ukraine, they're beginning to export their technology to their partner countries, one of whom is the UK. So we'll derive great benefit from what they've done, and I'm sure there'll be Ukrainian advisers in the UK working with the British Army.
Mccann: It becomes philosophical, but I can't lie to you, it concerns me deeply, the idea that in some ways it's a very human thing to want to outsource very, very difficult decisions which are made in war, which are not desirable and which I would not be capable of making myself. So therefore I can't judge as someone who doesn't make them. But it is very natural for us to think if an AI model can make those decisions for us, then the burden is taken off our shoulders. But it worries me. The idea of that worries me because the human emotion, while it may well be responsible for wars, it feels like it's also a stopper, an ultimate stopper in the end. You know, we can't do it anymore. We don't want any more destruction. It's too much. And AI doesn't have that. A computer doesn't have that. It has an end goal, which may well mean that it would continue until it felt it had won, and that worries me. There's no moral restraint. There's no consciousness.
Dearlove: There is just the fact of the capability, it is operating without reflection.
Mccann: Which is why we need humans in charge. Let's end on a positive note. We need humans in charge. That's the point. We do for all their faults and excesses. There you go. Rejecting AI. Brilliant. Richard, it's been nice to catch up with you.
Dearlove: Very nice talking to you, Kate.
Mccann: Well, that's it for this week's One Decision in Brief, and do stay tuned to our regular scheduled episodes. As I said, you can find those on our YouTube page, on onedecision.com or wherever you get your podcasts from. I'm Kate McCann. Thanks very much for listening.
