How AI is driving a future of autonomous warfare | DW Analysis
It's late 2020 and war has broken out in a place the world had forgotten. A festering conflict has erupted into full-scale fighting. Ground zero is Nagorno Karabakh… a disputed region in the Caucasus mountains, fought over by two former Soviet republics: Armenia and Azerbaijan. This looks like a textbook regional war – over territory, over ethnic and national pride. Fought while the rest of the world is consumed by the pandemic, it doesn't get that much media coverage.
But for those who are paying attention, it is a glimpse of future wars. You can find it right here, in the propaganda pumping out from the start of the war. Azerbaijan's border patrol posts this video on its YouTube account just as the conflict begins.
The lyrics are a rush of jingoistic fever, with a mantra: "hate" for the enemy. But look carefully, and you'll see what makes this conflict a watershed in modern war. Watch out for these trucks in the background.
In this shot you can just about see what's inside. Then … a launch, in slow motion. What emerges is not a rocket or a missile: it has wings that are beginning to unfold just before the video cuts away. We can see enough to identify what this is.
It's what's called a "loitering munition" from Israel's state-owned defence manufacturer, IAI. Its model name: the "Harop." The company's promotional videos show what "loitering munitions" can do. Once launched, they fly – autonomously – to a target area, where they can wait, or "loiter" in the sky for hours, scanning for a target – typically, air defence systems. Once they find a target, they don't drop a bomb, but fly into it, to destroy it on impact.
It's earned them the nickname "kamikaze drones." In the war over Nagorno Karabakh, these weapons didn't just make for good propaganda. They made a real difference. Azerbaijan had spent years investing in loitering munitions.
Analysis by a US think tank showed that they had more than 200 units across four different models – all of them sophisticated Israeli designs. Armenia only had a single, domestically made model with a limited range. "The really important aspect of the conflict in Nagorno Karabakh, in my view, was the use of these loitering munitions, so-called kamikaze drones, these pretty autonomous systems." Ulrike Franke is one of Europe's leading experts on military drones. "They also had been used in some way or form before, but here, they really showed their usefulness, militarily speaking, of course.
It was shown how difficult it is to fight against these systems." As Azerbaijan celebrated victory, you could even call Nagorno Karabakh the first war that was won -- in part -- by autonomous weapons. Little wonder the Harop was on show that day. And other militaries were paying attention. "Since Nagorno Karabakh, since the conflict, you could definitely see a certain uptick in interest in loitering munitions. We have seen more armed forces around the world acquiring or wanting to acquire these loitering munitions." The Nagorno Karabakh war amounted to a showcase for autonomous weapons technology. With a clear message: this is the future.
It's a future that is coming at us fast. Ever more advanced models are coming onto the market… Designed to hit a wider range of targets… The manufacturer IAI even markets one of its models with the slogan… "fire and forget." Fire and forget… think about that. Already, today, autonomous weapons are being used to
find a target over long distances and destroy it without human intervention. And this revolution is just getting started – turbocharged by artificial intelligence. In the United States, a major report from a "national security commission" on artificial intelligence talks about AI enabling a "new paradigm in warfighting" – and urges massive amounts of investment in the field. This isn’t all about autonomous weapons – there are many other areas of the military which will be using artificial intelligence.
“One area where we see a lot of AI-enabled capabilities is in the realm of data analysis. So we get we are gathering all so much data in military operations. Another area, which I think is quite promising, but also still relatively removed from the battlefield, from the combat is logistics. AI can definitely help to make this more efficient, cheaper, better, easier, all of that.” And fuelling all of this is an intensifying global competition, which spans all the way from these more prosaic fields to the autonomous weapons we’re looking at today… "The Chinese and the Russians have made it very clear that they intend to pursue the development of autonomous weapons… Martijn Rasser, a former analyst at the CIA, covers emerging weapons technology at Washington's leading defence think tank "and they're already investing heavily in the research and development of those systems."
It's not just the superpowers piling in. Britain's new defence strategy also puts AI front and centre. And as we've already seen, Israel is a leader in the autonomous weapons field.
In fact, wherever you look, countries of all sizes are jumping in. No wonder there's talk of this becoming an arms race. Germany's foreign minister Heiko Maas is clear that that arms race is already underway. "We're right in the middle of it. That’s the reality we have to deal with." If anything, this might go deeper than an arms race… "AI is here to stay. And there is a belief among the major powers that this could make a
difference on the battlefield in the future. So they are frenetically investing in it. Indian Diplomat Amandeep Singh Gill is the former chair of the UN government experts' group on lethal autonomous weapons "And this is a race, in a sense, which cuts across the military and the civilian fields, because there's also the sense that this is a multitrillion dollar question. It's about the future of resilient economies."
That is what sets this new era apart from arms races of the past. During the Cold War, the development of nuclear weapons was driven purely by governments and the defence industry. Beyond power generation, there wasn't much commercial use for nuclear technology. Today, AI is rapidly entering our everyday lives. It might even unlock the phone in your pocket when you hold it up to your face. This emerging ubiquity of AI important. Because it means that developments in AI cannot be contained – they are BOUND to bleed across between civilian and military fields -- whether we like it or not.
“AI is by definition dual use or multi use, it can be used in all kinds of ways. It really is an enabler more than the technology. There is a whole range of applications of artificial intelligence in the civilian realm, from health care to self-driving cars to all kinds of things.”
It means that something as innocuous as a new year's celebration in Edinburgh… or St Patrick's Day in Dublin… can be powered by similar swarming technology… …to what the Indian army showed off on its national day. In fact, swarming is one of the hottest areas of autonomous weapons development right now. The US Navy has released footage of early demonstrations. Here, fighter jets drop over 100 tiny drones in mid-flight.
Once they're out there, it's almost impossible for the human eye to keep track of them. The whine of their motors -- almost the only sign of the threat in the sky. Experts say they will make highly effective weapons. "You could take out an air defense system, for example, by -- just you throw so much mass at it as so many numbers that the system is overwhelmed. This, of course, has a lot of tactical benefits on a battlefield. And no surprise,
a lot of countries are very interested in pursuing these types of capabilities." Not least the head of the body advancing the US army's modernisation, as he explained in an online think tank forum. "Most likely drone swarms are something you're going to see on the battlefield – on a future battlefield. I don't think it's a matter of if – as a matter of fact, I think we're already seeing some of it – it's a matter of when we begin to see it." And feeding the momentum of this potential arms race -- in order to fight these weapons, you need these weapons. Humans don't have a chance.
"When you're defending against a drone swarm, a human may be required to make that first decision. But I'm just not sure that any human can keep up with a drone swarm." This issue of speed gets us to a critical emerging danger of autonomous weapons... The weapons we've seen so far are capable of a high degree of autonomy. But they wouldn't be IMPOSSIBLE for humans to control. Even a "fire and forget" weapon needs a human to fire it, and they're still operating in a way that we can pretty much grasp. Now let's think ahead, a decade or two into the future. That's a decade or two of rampant
technological development - and adoption - of increasingly autonomous weapons. "I think what is very likely that in 20 years' time we will have swarms of unmanned systems, not even necessarily just airborne drones -- it can also be ground systems, surface vessels, etc. So different units operating together and carrying out attacks together, which does indeed require quite a high level of AI-enabled autonomy." To fight these systems, you will need these systems. Because human beings are simply too slow. "This is what potentially may drive an arms race that -- some actors may be forced to adopt a certain level of autonomy, at least defensively, because human beings would not be able to deal with autonomous attacks as fast as would be necessary. So speed is definitely a big concern here." And that could have fateful consequences for how wars begin.
"We could find ourselves in a situation where because of this this problem of speed and autonomous systems having to be countered by other autonomous systems, we could find ourselves in a situation where these systems basically react to each other in a way that's not wanted…. In the literature we call this "flash wars" -- where you have an attack or even just that you think that there is an attack and autonomous system reacts to that. Another autonomous system by the opponent reacts to that attack. And you have this escalation potentially very fast. Hence the "flash" – "flash wars," where you basically have an accidental military conflict that you didn't want." We've already seen something like this on the financial markets.
The "flash crash" of 2010 wiped more than a trillion dollars off the US stock markets in just minutes. It was driven by trading algorithms feeding off each other in a dizzying spiral. How it happened is STILL not fully understood. In a flash crash, trading can be halted to prevent disaster. The risk with a "flash war" is that there might be no pulling back.
"If the beginning is bad enough, it may not even matter anymore that the original event wasn't supposed to be an attack in the first place. But you could have a situation where the counterattack is so bad that you end up in a war." Now, think back to Nagorno Karabakh -- a regional war where autonomous weapons may have tipped the balance. In a future world with the risk of "flash war," places like this could face even more instability, even more conflict. "We are moving in the world into a world where systems will be more autonomous. But we need to make sure that we minimize the risk of unwanted escalation, of lethality decided by machines without any human control."
But how do we do that? How do we prevent the worst? As we’re about to find out… the world is struggling to find a way… "My fear is that there will be more unpredictability in how we get to armed conflict, so the pathways to getting to the battlefield won't be clear to policymakers. So they will not understand fully the risks of certain actions or certain happenings, and that will make the whole world a more dangerous place." Amandeep Singh Gill was at the centre of United Nations efforts to try to get a grip on autonomous weapons… a process that critics say is on the brink of failure. This is where it all happens… The UN buildings in Geneva. It’s here that delegates from UN member states gather with experts and NGOs to talk about the future of autonomous warfare.
This process is part of what's called the UN Convention on Certain Conventional Weapons. A diplomatic tongue-twister launched in the 1980s to try to regulate non-nuclear weapons that were deemed so dangerous that they need special attention. Things like land mines and blinding lasers. In 2014, lethal autonomous weapons made it onto the agenda.
It has been very slow going. The process has yielded a set of "guiding principles" – saying that autonomous weapons should be subject to human rights law, and that humans must have ultimate responsibility for their use. But these "guiding principles" have no force… they're just a basis for more discussions. For campaigners calling for a ban, that's not good enough.
"We do get frustrated by the delays that have happened and the delay in moving from discussions to actual negotiations of a new treaty. The main problem with this forum is that it operates by consensus. So meaning any one state can block progress and block that shift from discussions and negotiations." Bonnie Docherty lectures on human rights at Harvard Law School - and is also a spokeswoman for the “Campaign to Stop Killer Robots” – a high-profile coalition of NGOs. She has mapped out principles for an international treaty.
"The overarching obligation of the treaty should be to maintain meaningful human control over the use of force, and where it should be a treaty that governs all weapons operating with autonomy that choose targets and fire on them based on sensor's inputs rather than human inputs." That idea of keeping "meaningful human control" is broadly echoed by many countries, but only 30 states support the campaign. They're mostly smaller nations but include one giant in the form of China. But Beijing's true position is blurred.
"China has called for a ban on, or expressed support for a ban on USE, but has not, to my knowledge, expressed support for a ban on development and production. We believe that you need to prohibit development as well as use of these inherently problematic systems, because once things are developed, the genie is out of the bottle." And the other great military powers aren't keen at all on those sorts of limitations either. Russia is accused by many of taking any opportunity to thwart the Geneva talks. But there are plenty of other objectors too. "Russia has been particularly vehement in its objections… Some of the other states developing autonomous weapon systems such as Israel, the US, UK and others have certainly been unsupportive of a new treaty and have expressed varying degrees of support for actually continuing discussions. So those are some of the roadblocks that we face."
As things stand, the US is highly unlikely to support a ban. Rather, it has set out its own principles, which include human involvement. "A ban on autonomous weapons systems is essentially infeasible just because the technology is out there. The Department of Defense has been very clear about its commitment
to ethical uses of these technologies, where right now the position is that a human being has to be on the loop or in the loop when those weapons are used so that it won't be fully autonomous in the sense that there won't be any human interaction with these weapons systems." But the reality is that the US, China and Russia are competing so intensely in all areas of AI technology that it’s questionable whether any of them would sign up to a treaty that significantly limits what they can do. “The large powers will have will always have agendas. They want freedom of manoeuvre. They think that they need to have agency over technology development. And sometimes they've been very sceptical of the role of international organizations, multilateral forums in understanding and regulating technology.” Aside from the lack of interest from crucial players… the challenge of tackling an intangible technology like AI with the traditional tools of “arms control” is genuinely difficult.
"A lot of the old ways of arms control and arms control treaties don't work anymore and don't apply anymore to these systems, because we are to put it bluntly, we're talking about software rather than hardware. So a lot of arms control systems in the past basically were about, you know, allocating a certain number of systems. You are allowed one hundred warheads of this type and you were allowed one hundred heads of this type. And we're basically counting. You can't do this with the A.I. enabled weapon systems that we were talking about, because it doesn't matter what it looks like from the outside. But what's in there."
Germany has been quite active in trying to navigate around these problems… its foreign minister says that the world has to find a way… “Just like we managed to do with nuclear weapons over many decades, we have to forge international treaties on new weapons technologies…” Heiko Maas is a member of Germany’s social democrats and has been a vocal advocate of arms control. “They need to make clear that we agree that some developments that are technically possible are not acceptable and must be prohibited globally.” In fact the German government has laid out its intention - in the document that underpins the current coalition. It says… "We reject autonomous weapons systems
that are outside human control. We want to prohibit them worldwide." That sounds pretty clear. But even this is complicated. Germany for instance does NOT support the Campaign to Stop Killer Robots. It says there’s a better way. “We don’t reject it in substance – we’re just saying that we want others to be included the global controls that we would need to ensure that autonomous weapons systems don’t come into use… So military powers that are technologically in a position not just to develop autonomous weapons but also to use them. We need to include them.”
So this isn't just a debate about the rights and wrongs of autonomous weapons. It's also a debate about PROCESS. On the one hand, Germany says an agreement it only worth anything if the big countries are on board - they want that elusive consensus in the Geneva process. On the other, the Campaign to Stop Killer Robots says the matter is too urgent to wait. They say there's just time for one more round in Geneva. "We feel that if states don't take action by that point, that they should consider strongly they should move outside of the Convention on Conventional Weapons and look at other options. So they could go to the UN General Assembly to negotiate a treaty. They could
start an independent process, basically a forum that is not bound by consensus, but is guided by states that actually are serious about this issue and willing to develop strong standards to regulate these weapon systems." There’s precedent for this… with land mines, for example. In the 1990s, the Geneva process couldn't find consensus. Instead, more than 100 countries broke away to create a ban called the "Ottawa Convention." But the great powers didn't sign. And more than 20 years later, the US, Russia and China still haven't joined the Ottawa Convention.
"It's a dilemma, isn't it? So you can do away with the rule of consensus and then you can have results quickly, but they will not have near universal support at the very least, they will not have support from the countries that are developing these capabilities. But through the rule of consensus, you force those countries to engage. So I think it's a choice that the international community makes in these forums." So the world doesn't agree on what to do about autonomous weapons. And it can’t even agree on HOW to agree on what to do about them. In this situation, is there ANY prospect of a solution?
"In the end we may end up with rules or norms or indeed agreements that are more focused on specific uses and use cases rather than specific systems or technology. So you where you basically agree, for example, to use certain capabilities only in a defensive way or only against machines rather than humans or only in certain contexts. But as you can imagine, implementing and, first of all, agreeing to that and then implementing that is just much harder than some of the old arms control agreements." Compounding this is the rock-bottom level of trust between the major powers right now. US-Chinese talks in Alaska in early 2021 descended into a bitter round of accusations.
"When there is lack of trust, you tend to attribute all kinds of intentions to the other party and you tend to overestimate what they might be doing and overshoot in your own response. Today, frankly, the developments on the technology front are actually adding to the mistrust." As the US, China and Russia slip deeper into an era of "great power competition," the challenge will be to carve out areas like this -- where they can put mutual interest above the visceral drive to be on top. THAT is the spirit of "arms control."
"You don't make arms control agreements with your best friends and allies. You always, by definition, you know, try to negotiate them with your enemies. And this isn't exactly new… I don't I don't think it's impossible that these players, which are already opponents and may eventually become even more adversarial, can come together and agree on certain minimum requirements simply because it is in everyone's interests."
For Germany's foreign minister, the whole world has responsibility here. “The world must take an interest in the fact that we’re moving towards a situation with cyber or autonomous weapons where everyone can do as they please. We don’t want that.” Climate change serves as an ominous warning of what can happen when humanity sees a common threat on the horizon but FAILS to act in time to stop it. The Rio Summit kicked off the UN’s process of talks to tackle climate change way back in 1992… It took 23 years to get to the Paris Agreement… And it’s clear even THAT wasn’t enough… It's already too late to prevent much of the devastation that scientists predicted from the start. With the scenarios we've just seen -- the warning signs are just as clear, and if anything even more urgent.