US Air Force secretary so confident in AI-controlled F-16s, he'll fly in one The US Air Force is rapidly scaling up its plans to automate some of its fleet, and the civilian boss of the service says he's planning to fly in one of the robo-planes this northern spring. Last week the USAF delivered three F-16 fighters to Eglin Air Force Base, for conversion to full AI control as part of the Viper …
Flying a jet fighter through a wide open sky is significantly less complex an AI problem than piloting a car though a city street with people walking, cyclists and other cars.
This is why I'm sure self-driving cars are decades away, there are a lot of steps along the way that don't exist. We don't even have semi trucks that are self-driving only on freeways yet, and there is a lot of money in that.
As down as I am on self-driving cars and trucks, the Airbus model of computer assistance/control, and anything I'm riding in being drive/fly-by-wire, I think this would be a very useful development .. if I were flying in a combat situation, I'd want one or more autonomous outriders.
The development of such outriders will be costly, in human lives, as well as treasure. Consider this scenario: outrider off on a recce, picks up an enemy heat-seeking missile. Outrider jinks and janks, keeps the missile at bay (no human pilot == outrider can stand much-higher G forces). Human pilot on the mothership, unaware of this, issues a recall order to the outrider. Outrider responds, missile "in tow". Outrider (and enemy heat-seeking missile) near the mothership, and the missile, detecting the much-stronger heat signature of the human-piloted mothership, targets the mothership. Scratch one human pilot and one mothership, much as in the "dog and dynamite" urban legend (idiot goes ice-fishing, fuses and lights a stick of dynamite to blow a hole in the ice and kill the fish, throws the lit dynamite stick ... and his dog retrieves it and brings it back to him).
More such potential mishap scenarios for motherships/outriders exist than I can think of, and that the design team (and people above them in the command chain) can think of, either.
1. The current model American AIM-9 has a usable flight time of 60 seconds, and an operational range of up to 22 miles. Then, as you wrote, it's coasting.
2. A minute is enough time for the scenario I described above to occur.
3. My scenario was an example, and there are many more which did not occur to me.
4. The designers and managers -- said managers making product functionality (-limiting) and time/financial decisions -- are not omniscient, and frequently insufficiently-imaginative to imagine how things can go wrong.
To take the aviation simile all the way: take-offs and landings, as well as taxiing is done by hand, as a general rule. As long as verbal instructions are given, someone human responds. Automation takes over once the filed flight plan route is reached.
Or should be - there have now been studies by FAA, ICAO and the rest of the alphabet soup on increasing over-reliance on automation/autopilot and decline of piloting skills, but I digress.And yes, there are auto-land systems for things like CAT III, etc. Without naming countries or airlines, some of the jet jocks have airmanship skills that leaves them at a loss when the brown stuff is hitting your turbofan intakes.
The point and analogy being that, driving in dense city traffic, the driver has controls; once you're on the freeway, let the automation take over.
Here in Australia you can flunk your instrument proficiency check by over-relying on the a/p, but you are definitely allowed and expected to use it to assist with single-pilot IFR. It's a fine line.
"Flying a jet fighter through a wide open sky is significantly less complex an AI problem than piloting a car though a city street...."
Flying a jet fighter is considerably more complex in cases where there are multiple fighters, some friend some foe, dueling in a tight area while watching out for enemy fire and SAMs.
Indeed, and a computer program reading from thirty different sensors at the same time, looking more than one direction in multiple spectrums is more capable than a human at taking all those sensors in.
The complexity of teaching an AI to fight, ID targets and so on is a challenge, but not insurmountable. Training a jet jock takes two to three years. Train the computer and you have that in perpetuity with no lead time. That’s a big advantage. Drones that you can expose to wild weasel while doing the job is a whole lot more expendable than any aircrew.
Also a little sad as sitting up front one of those machines is waaay up there as human experiences can ever go. I guess it’s the Pitts S2 for aviation fun…
"Flying a jet fighter through a wide open sky is significantly less complex an AI problem than piloting a car though a city street with people walking, cyclists and other cars."
No, just no. Robo-cars can pull over and stop in case of issues. An aircraft in comparison is hyper-complex. It needs to move forward at 100 knots+ continuously or it falls out of the sky. Weather, instrument failure, jamming, ATC considerations, lightning strike, bird strike, traffic proximity, all of which can change things in a heartbeat. Last a fighter, to be effective in it's role, needs to be armed with things like missiles, cluster bombs and guns. You want AI in control of those as well while flying over domestic bumper to bumper traffic?
Not to mention that the accuracy rate of "AI" today is around 60-70%. You want to be a passenger on those odds?
As Ukraine is showing, the jet fighter has an increasingly limited role to play.
Also given how the US is struggling to build the current generation of jet fights in a significant quantity, I suggest defence monies would better spent on making existing tech more reliable, eg. having missiles that can handle being flown more than a couple of times before they need their gyroscopes etc. recalibrated (biggest cause of sidewinder failures).
"eg. having missiles that can handle being flown more than a couple of times before they need their gyroscopes etc. recalibrated (biggest cause of sidewinder failures)."
Don't missiles normally only fly once? I've not seen a missile with landing gear, other than possibly an Acme missile fired by Wile E. Coyote :-)
The Arab-Israeli wars of the 60's and 70's exhibited the full spectrum of aerial warfare. In the early days, the Israelis had absolute air superiority; brought about by precision and massed surprise strikes.
In the war-of-attrition period, Soviet SA2, 3 and various other SAMS shored up the Egyptian defences to the point that the Israelis could not operate so freely; with both sides air power largely negated and the war evolving into the near-statemate seen on the ground.
ECM and ARM capabilities of adequate scale would allow for close air support and more importantly, the interdiction missions to operate at medium altitudes; out of reach of MANPADs and light flak, to disrupt the all-important supply runs. This capability offer a chance to break the stalemate on the ground. It's also why you either air superiority/interceptor capability - to both enable and defend against this specific mission (that's not currently possible).
Both sides extensive use of drones in Ukraine, besides the local tactical value; are part of efforts to "disable" SAM systems - force enough engagements and you run out of expensive SAMs to counter those CAS and Interdiction operations that are currently unable to fly without risking enormous losses.
Talk about a waste of time and money!
They need to design a totally new platform for that, which will be FAR cheaper since it won't need a cockpit and won't have concerns about protecting a fragile meatbag sitting in it. There will be several versions, with varying degrees of cost and capability but with the goal of having 10-50x more aircraft in a combat area than they would currently have human piloted aircraft at the same program cost.
China WILL be doing this, if we're stuck trying to manage with human piloted aircraft in 10 years they will be completely overwhelmed by the numbers thrown at them in a real combat situation. It doesn't matter how much better human pilots are than AI piloted drones, they can't possibly fight off dozens of aircraft attacking them simultaneously - some of which will use kamikaze tactics and try to simply crash into the enemy.
I suppose it is good that the Air Force is beginning to see the reality of what faces them, but there will continue to be a ton of pressure against these programs from Air Force generals who were formerly fighter jocks - they'll do everything in their power to keep pilots in seats, long past the time they no longer belong.
If they design "a totally new platform" right now, then the first malfunction of the systems that they are developing will be a big setback to the program.
As it is, they can test with a rated pilot on board and move forward to an optimized platform later.
HOWEVER, this cost optimized platform may consist of loading the myriad of old F16s stored in boneyards with new software, removing a lot of the cockpits, including ejection seats, oxygen tanks and similar, and fairing over where the canopy used to be. If the airframe is only needed for a low number of flights into contested airspace, then the little lifetime left in an old F16 may just be what is needed.
I think I read somewhere that China may be (or may be feared to) doing the same with old Mig 21s.
What Ukraine has demonstrated (although it was already known), is that even the strongest anti-air system can be overwhelmed by masses of cheap unmanned systems. And so, suddenly a boneyard becomes a valuable resource
They need to design a totally new platform for that
Did Lockheed Martin or Boeing write this? Because I'm sure they're already working on their proposals to print money provide the USAF with aircraft that will definitely come in on budget. /s
These are only test aircraft, so it makes sense to use F16s, because the USAF has over a thousand of them sitting in boneyards.
Not sure such a contract would be limited to them. The DoD could/should entertain bids from smaller companies, since the consequences for failure would be a lot less for unmanned craft. There are still limited numbers of suppliers for stuff like jet engines or weapons systems, but the drone airframe and more importantly its AI software could come from anywhere.
You'll have to talk me down. I've got damage.
- OK, Topper. Ease her in.
Landing gear's frozen.
- Lookin' good.
Lost my radar.
- A little more power now.
I'm out of fuel.
- Right for lineup.
Lost a wing.
- Doin' fine.
There goes the other one.
- OK, Topper. Call the ball.
Touchin' down.
"They need to design a totally new platform for that, which will be FAR cheaper since it won't need a cockpit..."
So you missed the part of the article where it says the F-16 is only a development platform for the software? I'm pretty sure that yes, they will eventually stick this software is in a pilotless design hardware.
AI recognising and reacting to its surroundings?
Hope they have really good means of testing against (radar etc image variants of) Adversarial Image Attacks.
Pilot: Bogie sighted; weird paint job. Engaging.
AI: Aww, look at the widdle kitty; whoops, better switch off that nasty tone, don't want to scare him.
"Are you referring to the retired Lieutenant Colonel who used to teach engineering before entering politics?"
"Hoist with his own petard" still works even in our High Tech world !!!
So comforting !!!
Update:
Possibly found holding a 'Petard' my self !!!
:)
This post has been deleted by its author
This post has been deleted by its author
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
