The future isn't one operator flying ten drones. It's one commander giving ten intelligent machines a mission – and being able to trust what happens next.Robert Merryweather
- Not one artificial intelligence trying to make every decision, but different kinds of intelligence doing different jobs:
- React Instantly, Think Fast and Think Deep – all operating within a separate system that governs what it is actually allowed to do
- That’s the concept, but how would this play out in a real mission?
Imagine a brigade preparing to advance through contested territory. Somewhere ahead, mobile air-defence and long-range fires systems threaten its route. The command given to an autonomous team could be quite simple:
Find those threats. Maintain awareness of them. Be ready to support action against them – but do not engage without human authority. Let’s say we have three drones to do this. Nobody tells them exactly where to fly, but instead, Think Deep plans out the mission and stays ready to adapt.
One drone moves farther ahead carrying an electromagnetic-warfare payload. Another has EO/IR sensors. A third keeps a wider watch over the area. The system balances coverage against endurance because finding the first interesting thing isn't much use if every drone has exhausted its battery getting there.
For a while, they search. Then the first drone picks up an RF emission from close to a treeline. Now there is something to investigate.
Think Fast shifts the team around it. The EO/IR drone changes course to get a look. The third doesn't follow. It keeps searching elsewhere, because the commander asked the team to understand the threat across an area, not to send every sensor towards the first possible target. The second drone finds vehicles concealed near the source of the transmission. One sensor has the electromagnetic picture. Another is building the visual one. Their information is combined and confidence increases that this is one of the systems the brigade is looking for.
Meanwhile, a drone gets too close to terrain. Two flight paths begin to converge. Navigation becomes unreliable for a moment – React Instantly deals with those things. No mission planner needs to deliberate over collision avoidance and no human operator should have to intervene every time a drone needs to make a small correction.
The team now has a target of interest, but it doesn't attack yet. Think Deep looks at what the team as a whole should do next. One drone should maintain surveillance. A strike drone is available, however artillery elsewhere in the force may be a better option. What else still needs to be found? Which assets will be needed later? What is available now? Does using one option close off another?
The autonomous system can help assemble those choices, but the decision needs to be made by the authorised human.
The reconnaissance drone keeps watching, while the information moves into the wider fires network. Then another sensor reports something new: unexpected friendly forces moving into the area around the target.
The engagement is paused by our fourth layer: the Trust Governor. Conditions have changed so it can’t now allow the strike to go ahead, but Think Fast maintains surveillance and Think Deep tries to find another option to achieve the commander’s intent, remaining aware of the developing situation.
The friendly force eventually clears the area, so the system gives the commander a new artillery strike recommendation, which is approved and executed. Before the mission is finished however, one drone carries out a battle damage assessment while another returns to the broader search.
Now let’s look back at what the humans actually did. They didn't fly three drones. They didn't manually move one sensor after another into position. They didn't decide every route change or deconflict every drone.
They gave the system an outcome to achieve and boundaries in which to operate. That, for me, is where military autonomy becomes genuinely useful.
The aim isn't to take every decision away from people. It is to take away the thousands of routine tasks involved in positioning, sensing, coordinating and adapting so that people can concentrate on the decisions that really demand their judgement. That is why the four layers I described in my previous article matter.
- Think Deep keeps sight of the mission.
- Think Fast deals with the problem that has just appeared.
- React Instantly handles what cannot wait.
- And the Trust Governor makes sure that, however capable the other three become, intelligence is never confused with authority.
The future isn't one operator flying ten drones. It's one commander giving ten intelligent machines a mission – and being able to trust what happens next.
Read my previous article: Why military autonomy should think like a brain, not a black box