The news about the A.I.-flown fighters won’t be too welcome in some quarters. Fly-by-wire is state of the art, but the art is changing, too. You need to read the whole story to see how this announcement by DoD will work, but a lot of new cards have just entered the playing deck.
DARPA is operating the program, called Air Combat Evolution. The system was developed by Heron Systems. The A.I. has already done well in dogfight simulations, (see video) hence the next step of developing live A.I.-flown fighters.
Nuts and bolts
This isn’t just some tech fad, or cost-cutting exercise. A.I. flying has been coming for a while. There are plenty of positives from a military perspective:
• A.I. systems mean that fighters can access a lot more space onboard which would otherwise be used to support human pilots.
• Streamlining systems to exclude the human interface also saves space and processing power.
• Plane design no longer needs to be configured to have a human pilot.
• A.I. autonomous planes wouldn’t need telemetry like drones.
• A.I. can learn and develop skills and coordinate in combat.
• A.I. combat system reporting is likely to be very efficient in real time.
• A.I. can resist higher G forces than human pilots, allowing for some pretty savage in-flight capabilities.
• From a purely cost perspective, A.I. planes could be a lot cheaper, particularly if A.I. helps in design and construction.
The game changing role
The major role of air power is air superiority. Performance dictates superiority. A.I. can do things people can’t. It’s more than theoretically possible that A.I. warplanes can outperform human pilots in many specific environments.
When hypersonic systems enter into war environments, war in the air is likely to be fought at those incredible speeds. Humans can’t do that. They would be at a serious disadvantage to anything that can operate at those speeds. So it’d be suicide for human pilots to try to match the A.I. on its own terms.
There are likely to be a lot of other things flying around that are as bad or worse, too, like anti-air missiles, laser beams, etc. It’d be a waste of people.
Important, note: This doesn’t mean there’s no role for combat-trained pilots. Tactics, target selection, and similar issues need to be done by humans. Setting priorities for use of air assets is no joke. Air power is expected to do a lot more than ever before, and someone has to make the call about how that power is used.
So the weapons are now flying themselves…? Maybe not, at least, not yet.
Autonomous A.I. or semi-autonomous planes are a mixed bag of possibilities. Autonomous planes have to deal with issues on their own. Semi-autonomous planes could be “semi-managed” by remote pilots to the extent of avoiding hazards, recognizing risks, and setting programs for operations.
Another major issue is that as everyone knows, algorithms and codes can do so much. The unfamiliar environments of unpredictable combat might be quite a challenge for any A.I. “Go there and do that” isn’t quite enough. Risk assessments are part of combat. Dumb ideas are obvious to humans, but not necessarily to an A.I. which may not have any frame of reference.
For example – There’s a big floating thing over the target. Nobody knows what it is. The A.I. charges in, as instructed. The big floating thing turns out to fire a lot of area weapons which don’t do airframes a lot of good. A human pilot would distrust the big floating thing on sight. A.I. pilots might not. The same applies to any unfamiliar element in a war zone. Airborne booby traps or psychotic starlings, it all has to be learned.
The big recommendation for A.I. pilots is that they’ll save a lot of lives. They can test planes safely, too. The big non-recommendation for now is that all tech has to go through an incredibly thorough learning process. A.I. fighters are coming, and they’ll go into space, too. Just allow some room for thought.
