The crystal ball of war: how AI predicts tactics months in advance
10:13 Share

ACAST powers the world's best podcasts. Here's a show that we recommend.

Hey, everybody. I'm Jen. I'm Jess. And we're Fat Mascara, the only beauty podcast you need in your life. We're beauty editors by day and podcasters by night. And we've got all the industry gossip for you, like insider product reviews and advice you're not going to find online. And the best part is interviews with the most sought-after experts in the beauty biz. Charlotte Tilbury. Jen Atkin. And makeup artist Sir John. That's just a taste of what you're going to get on the Fat Mascara podcast. So come hang with us. New episodes drop every Tuesday and Thursday.

ACAST helps creators launch, grow and monetize their podcasts everywhere. ACAST.com Welcome to the World in 10. In an increasingly uncertain world, this is the Times' daily podcast dedicated to global security. I'm Toby Gillis, joined today by Alex Dibble. Yesterday's World in 10 took us to the seabeds of Europe and NATO's battle to protect the continent's underwater internet cables and gas pipelines, which are being sabotaged by Western enemies.

Today, the tech theme continues as we dig into the potential future of artificial intelligence in warfare. The Times' Berlin correspondent, Oliver Moody, also has an Eastern Europe brief, and he has been in Tallinn, capital of Estonia, to see how a new programme which claims to predict enemy attacks before they even begin works. Could it be a NATO weapon of the future, or even now? Delighted to say Oliver joins us. So, talk us through it. Is

Is this really a war crystal ball? Well, what it is fundamentally is a battlefield or intelligence management system. These are super common. They've been around for a few decades now. And the idea is basically that the modern battlefield is absolutely teeming with millions of data points. And what we have increasingly is a

Radical profusion of sensors and other kinds of information scraped together from every imaginable source, whether that's kind of social media or surveillance of cell phone traffic and emails, human intelligence sources, spy satellite feeds, etc.

And these are programs that tie all of these things together and try and visualize it in a way that allows a human commander or analyst to work out what's going on as quickly as possible. And that allows you to allocate scarce resources to where they're needed most. And the logical extension of that is if you can analyze those trends in the present, why can't you extrapolate them into the future as well if you have a large enough

data set to train the algorithm on so this company censusq in estonia has now taken that step and what they claim is that at a tactical level you know if you're a unit commander out in the field and you're wondering when and where the next attack on your sector of the front line might be coming they can help you to anticipate that kind of event with a timeline of up to two days in the future

But on a strategic level, the kind of grand sweep of war planning, they say that they can anticipate trends or scenarios four to six months in advance, which is a very kind of impressive and striking claim to make. So Oliver, all of this data feeds in to the algorithm, makes these predictions, which if they come true, of course, would be incredibly useful. Has it been successful in the real world?

Well, they are extremely guarded in terms of what kind of predictions are being made, who they're making them for, and very specifically what kind of real-world scenarios this has been used in. For obvious reasons, if you're the FSB or the GRU or indeed Chinese intelligence, this kind of information would be extremely valuable.

I think you can judge it, though, by the fact that their clients, again, I can't say who they are for security reasons, but they do include a number of very serious Western security services and militaries. So I think it's reasonable to assume that they would not be making these claims and getting the contracts if they couldn't at least to some degree deliver on them.

And there's all of this data floating around from drones, sensors, wearable tech, the sort of stuff we all wear, actually. What's been the holdup, though? Why hasn't the promise of this tech come somewhat sooner? It's a really good question. And the answer is, first of all, that before the full-scale Russian invasion of Ukraine, the data sets were pretty poor.

heterogeneous and piecemeal, you're talking about a very, very broad range of conflicts that Western militaries were either involved in or kind of advising on. If you think about, you know, Iraq, Afghanistan, the Balkans, peacekeeping missions, it's very hard to tie those things together in a meaningful way.

But also you run up against computational problems. You're trying to integrate lots of very, very different kinds of data, some of it visual, some of it audio or radar or text. You're trying to evaluate, are these text sources reliable? Is this crap that's being fed to me deliberately by the other side? And weaving all of that together quickly, then getting those insights appropriately

approved by human analysts and then delivered to the people in the field in the time that they need to be able to act on them has been really tough. But what we have seen is that the full scale invasion of Ukraine and the very, very close involvement, particularly of British forces,

And American military intelligence in helping the Ukrainians has absolutely turbocharged these developments at the same time as AI has just become a lot more powerful and economical. So you're starting to get to a point where you can actually get these technologies out into the field in a useful way.

Oliver, these developers are in Estonia, which also borders Russia. What can we take from this development happening in this tiny country rather than somewhere better known for tech like the US or maybe even Israel? Well, it's important, first of all, to point out that it's not just Estonia, right? One of the biggest European players in this sector, which has been around since the 80s, is Systematic, and they're based in Aarhus.

In Denmark. So because this is essentially software, the kind of barriers to entry are number one, are you good at software? And number two, can you get your hands on relevant data? And the answer is basically the Estonians are exceedingly good at software. They have one of the most dynamic networks.

Tech sectors and talent pools in Europe. You think about like Bolt or Skype was developed by Estonian engineers. And also they have very, very good intelligence services and a kind of small but extremely well-drilled military. And they've been taking part in every single Western coalition deployment that they could possibly

get into for the past two decades because they want to demonstrate their worth to NATO and to deepen those alliances, particularly with the US, on which the existence of their country depends. So in this kind of context, there's every reason why the Australians would be good at this stuff.

There's a horrible phrase, isn't there, that militaries fight the next war with the tactics and the techniques that they've learned from the last one. What are the military commanders looking at? What do they hope for with this technology as it develops? Well, if you're Estonia, you have basically one single scenario in mind, right? Which is you are next to a neighbour that has invaded your territory recently

at least 40 times over the past millennium, if you go by what Sunnistonian historians tell you. And there is every reason for them to fear that it might do so again in the near future. Their intelligence services estimate that it could be sort of three to five years after some kind of ceasefire or at least temporary halt to proceedings in Ukraine.

So they have a pretty good idea of Russian military doctrine. They know what Russia's capabilities are. They know the kind of assets that NATO has on the ground in the Baltic states, including in Estonia. They know the kind of assets that NATO can get out there to the front line if it becomes a front line and how quickly, more or less, they're going to get there.

So in terms of kind of planning for how this invasion might play out, they have most of the kind of starting preconditions that you would want in order to simulate these kind of scenarios with the software we're talking about. Oliver, one final thought. If this can predict future war movements, couldn't it predict a myriad of other things? How else could it be used, in other words?

Right. I mean, if you think about this as a kind of more general machine learning algorithm, it's spotting relationships between data points, suggesting...

Complex patterns in that data that a human analyst might not be able to see and then extrapolating them into the future. And the battlefield is the most kind of dramatic application for that kind of technology. But CensusQ does also use it for things like detecting tax evasion.

or smugglers. It's not just military intelligence that's interested in this kind of software. You could also see other, even financial intelligence agencies using it. Okay, Oliver, thank you. That is Oliver Moody, The Times correspondent in Berlin. And that's it from us. Thank you for taking 10 minutes to stay on top of the world with the help of The Times. We'll see you tomorrow. ACAS powers the world's best podcasts. Here's the show that we recommend.

This season on The Dream. Supplies are being provided by nurses who run out in the middle of the night and purchase diapers. But the hospital is still charging as if they still have these items. We are digging into every topic we've ever wanted to cover on this show. It's a spinning plate analogy. The second that you stop spinning those plates, that crashes. So you can never stop working. The Dream season four comes at you weekly starting Monday, January 20th.

Acast helps creators launch, grow, and monetize their podcasts everywhere. Acast.com.