Mutual assured dronestruction

Western airstrikes on the Middle East: déjà vu all over again. Twenty years ago, the USA attacked Sudan and Afghanistan with Tomahawk cruise missiles. Two days ago, the USA attacked Syria with … Tomahawk cruise missiles. Aside from the (de)merits of each attack, isn’t it a bit surprising that technology hasn’t really changed small-scale strategic warfare in that time?

Just you wait. In the next decade, that strategic calculus will change a lot, and probably not in a good way. Consider this sharp one-liner from Kelsey Atherton last week:

Of course cheap drones are already being used on the battlefield in small-scale ways: by Daesh, by Hezbollah, by Hamas, by drug cartels, and of course by traditional nation-state militaries worldwide. But those are piloted drones, used in short-range, often improvisational ways; interesting but not really strategically significant.

Meanwhile, across the world, we are in the midst of a Cambrian explosion of artificial intelligence and automation technology. Consider, the startup that began as a literal one-man self-driving-car project. Consider the truly remarkable Skydio, a self-flying drone that can follow you wherever you go, avoiding obstacles enroute.

…Do you see where we’re going here? Right now only major powers can toss a few explosives at a faraway enemy to drive home a political point. But imagine a flock of bigger Skydios, reprogrammed to fly to certain GPS locations, or certain visual landmarks, or to track certain license plates … while packed full of explosives.

A  Tomahawk costs $1.87 million. It seems to me that we are not far at all from the point where a capable and wealthy non-state actor like Daesh, Hezbollah, Hamas, the Sinaloa cartel … or any unsavory group willing to be used for plausible deniability by a nation-state … could build a flock of self-flying targeted kamikaze drones, then smuggle them into the Western destination of their choice, for a lot less than the price of a single Tomahawk. The self-flying and targeting software / AI models won’t need to be nearly as perfect as that of a self-driving car. A 50% failure rate is more than acceptable if you only want to show force and sow panic.

It’s chillingly easy to envision a future of mutual assured terror, a multipolar world in which nations and terror cells and drug cartels and starry-eyed cults alike have the capability to inflict faraway havoc on thousands and constant dread on millions, a smoldering kaleidoscopic landscape of dozens of factions enmeshed in tit-for-tat vengeance and vendettas — ceaseless cycles of sporadic attacks which rarely kill more than a hundred, but send entire populations into perpetual fear and fury. Fury which will be very hard to direct. Like hacking, autonomous drone attacks will be extremely difficult to attribute.

You may call this science-fiction scaremongering, and you may have a point. It’s true that nothing like this has happened yet — though the existing adoption of commercial drones for warfare is a distinct warning sign. It’s true that it would be wrongheaded and ridiculously preemptive to try to slam the barn doors before any drone horses arrive. I’m definitely not suggesting that the West should start thinking about restricting research, or trying to control either hardware or software. (Even if that worked, which it wouldn’t, it would be pointless; drone hardware is cheap, and R&D is global.)

But it’s not too early to start thinking about how we will cope if and when self-flying kamikaze drones do make asymmetric strategic warfare possible. And it’s definitely not too early to try to minimize such warfare before it happens, ideally by actually trying to deal with the root causes of the conflicts burning around the world, rather than lobbing a few cruise missiles their way every time we feel the need to seem particularly outraged. Because one day, not long away at all, that approach will begin to rebound on us disastrously.