Drone Wars
Drone Wars
The MoD has a new strategy for uncrewed platforms. Is it the right one?*
*No. Read on -
Image credit: Unamanned Systems Technology
Another week, another British military publication on the future of warfare. Hot on the heels of its AI playbook comes the Ministry of Defence’s approach to thinking about drones. The headline came from the Chief of the Air Staff, Rich Knighton, who described three tiers of drones – disposable, attritable, and Gucci. It’s certainly one way of thinking about the issue, whatever you think of the godawful word, ‘attritable’. Let's unpick the types.
At one end of the spectrum, tier 1, are small, cheap drones, bought from Ebay and Amazon, and very tactical. These are of the sort that are changing the battlefield in Ukraine. Someone recently showed me the motherboards – you can get them online for a few dollars. And they’re apparently proving devastating in close combat. This week, Ukraine seems to have lost its first M1-A1 Abrams tank – speculation online is that a drone strike was responsible.
Then, in the middle, a group of drones that, frankly, I don’t think exists yet. These, ACM Knighton offered, are drones that you’d ideally like to get back, but if you don’t well – so be it. Is there such a thing? Perhaps with a swarm, able to operate with a degree of range and some degree of robustness (for things like weather and terrain) this sort of thinking makes sense. Could Shield AI’s new V-Bat fit the bill? It’s mid-size, and so might be cheap enough to manufacture and operate in tremendous quantities, as per the DoD’s aspiration for its Replicator programme. Certainly Watchkeeper and Reaper don’t work here. Reaper especially is far too scarce and expensive that you'd not really mind whether or not it’s shot down. That’s a problem – in recent months it looks like the Houthis have developed a knack for bringing them down. The basic type is now rather venerable – Predators go back almost a quarter of a century – and was never intended for highly contested airspace. The problem is that capable air defence has evidently democratised in the interim.
And lastly, the exquisite, super expensive platforms that you can’t afford many of, and that you certainly don’t want falling into enemy hands. Their ruinous expense could derive from the exotic materials they’re made from, or from the hush hush sensors on board. Think of a platform like the F-35, but uncrewed. Or perhaps it’s like the ‘loyal wingman’ idea – something along the lines of the Kratos Valkyrie. This type of drone would offer greater performance - like range, persistence and firepower. Maybe it would boast a more sophisticated AI setup – with greater on-board computing power – ‘edge computing’ in the jargon. Who knows, it might even host a fancy quantum computer, of the sort that the Royal Navy is currently experimenting with.
I get the three-way split. It makes intuitive sense. And yet I can’t help thinking it’s industrial era thinking applied to the AI age. A bit like having a cockpit in the Tempest mock up: yesterday’s news tomorrow. For me, the better way of thinking about drones is software first – that’s the most important bit of the system. True autonomy (not remotely piloted) is where we are heading. And with no pilot to protect, there’s no reason for platforms to look much like they do now.
Maybe I’m being excessively optimistic about the progression of AI pilots, or about the extent to which algorithms can capture our intent ahead of deployment. But if I’m right, I think disposability will be a feature across the entire spectrum of drone types. The cost of developing an AI pilot sophisticated enough to deal with challenging anti-access environments might be fierce. But cloning that algorithm for less intense environments will be trivial. In short, AI favours distribution and mass across the spectrum of air power roles. I want a strike package that will overwhelm your defences with mass and its fancy flying. And I want air defences that will be able to defend against scale. Quantity, in short, has a quality all its own, across the full spectrum of air power – so much so that small production runs of exquisite jets are not my vision of the future.
There are plenty of other ways of categorising the drone force of the future. Perhaps by role, for example. The existing air power roles are as good a place as any to start. My hunch here is that we’re about to move back towards specialised platforms, away from multirole ones. In the coming swarm, there’s no need to have all the units doing the same thing. But the pilot itself will be a generalist, at least to some degree. With language models, like GPT, you can have a specialist, tailored model, but it’s still the case that the most powerful generalist beats them at their specialism. So, I think, will it be for AI pilots. Sure, some ‘fine tuning’ will be needed to turn the best dogfighter into the best ground attack pilot. But the best of the best will get there through general factors, like superior edge compute.
Another way of thinking about drones is the question of who operates which role. History suggests that cultural identity shapes how people see the utility of air power. We might reflect here on the cultural differences between the USAAF’s tactical school (who were into air supremacy) and its strategic force – with their long-range bombing role. Or consider the different approaches of Coastal and Bomber Command in the Second World War, and their tussle for resources. Well, what’s needed here, surely, is some organisation that brings to bear seamless experience of the air and land domains. Step forward the RAF Regiment – the force of the future. Don’t @ me!
This is my regular newsletter contribution for the Wavell Room’s Edge of Defence - you can, and really must, subscribe! www.wavellroom.com