Wow, so they actually let AI fly an F-16 fighter jet Pentagon boffins have for the first time used AI algorithms to automatically control a real F-16 fighter jet mid-flight. Well, OK, at least for the first time they can talk about. The aircraft flown in the experiment, dubbed X-62A or VISTA, was modified and equipped with the right hardware components to run the software …
We're getting to actually having an ACE Combat game become a reality... We've had shooting balloons, shooting satellites, ghost of ***** for an ace pilot, short range ground to air missile runs, and now we've got robot/AI controlled planes...
Life imitates art in weird ways sometimes.
(and yes I know I know it's a weird series but the plots in some can be really fun - AC:5 is a notible example as is AC:7 - we don't talk about 6..... Or Assault Horizon. Though that's drawing parallels as well)
Modern aircraft are 'fly-by-wire' meaning the pilot commands are translated to electronic signals and the actual mechanical movements of the plane are controlled through the electronics. On the other hand they have highly accurate simulators which log far more missions than real flight hours, but with close to identical electronic systems. So it stands to reason that
a) fitting an aircraft to accept commands directly from the electronics is presumably quite easy technically.
b) they have tons and tons of training data that is both very clean, and also harvested from their top pilots (unlike other models trained on random datasets trawled from the interwebs.
So its completely unsurprising not only that they did this but that the AI could beat a real pilot in a dogfight.
Question is though, if they don't have anything in their training set for unexpected stuff happening or shit going wrong, how ill the AI respond.
That's quite apart from the generally bad idea of letting a computer autonomously control highly dangerous weapons
That's quite apart from the generally bad idea of letting a computer autonomously control highly dangerous weapons
That's rather situational, really.
It would of course be rather dangerous to allow a computer to fully control a fully armed and fuelled tank within driving distance of a population centre of civilians in peacetime in case it exhibits unwanted behaviour like shooting everything that moves.
On the front line in Ukraine however, personally i'd be quite happy having a computer control an obsolete tank sat stationary with instructions to kill everything entering it's weapon range from particular directions, on the basis that it'd make at least some use of obsolete scrap that would be a deathtrap in which it'd be tantamount to murder putting people inside on a modern battlefield.
Militaries have rather simple, robust and direct (non software) solutions for dealing with things that don't work, and on a wartime battlefield "too dangerous" is a somewhat different concept to a peacetime training field.
DARPA had an AI trial that was pretty much what you described (they didn't have an actual tank, but they did have a system that was supposed to identify threats in the near vicinity).
That system was easily defeated by the marines tasked with getting close to it. In one case, two marines snuck up on the system by hiding under a big cardboard box. According to the article I read, observers of the experiment could hear the marines giggling the entire time.
the generally bad idea of letting a computer autonomously control highly dangerous weapons
I know what you mean, but a better way of phrasing it would be "letting a computer launch weapons".
Modern missiles and bombs are already 'highly dangerous weapons controlled by a computer', but currently there's a human in the loop who decides to launch them. On a modern military jet, there's very few controls that the pilot can use that don't pass through a computer.
That's quite apart from the generally bad idea of letting a computer autonomously control highly dangerous weapons
For SF fans, I'd recommend Peter Watt's short story Malak [Warning: PDF]. Whether giving the AI a nuke in the story is a good or bad idea depends on your ethical stance.
Skynet is the name of the UK's military satellite communications system. No really! It was named in the 1960's, so really it was the film that stole the name.
Limitations of the pilot are probably more of an issue, in that even with a pilots G suit then when pulling ~+9G you'll black out, or you'll red out after ~-1G.
I suspect that the hardware is physically capable of considerably more than the meatbags are at this point, making the meatbag a bottleneck that probably needs to be removed.
Not only G factor. You must also consider life support equipment (oxygen, heating, pressurization etc.) as well as displays, switches, control column, pedals, ejection seat and son on. Yeah, it would become like a drone so I don't understand why they bothered with an F-16 in the first place.
many reasons: they're there (parked). They're cheap (money already spent). And, ultimately, a vision of, say, 300 multi-ton kamikaze drones heading for Moscow (where else? ;) must give many males a solid... ehm... shiver ;)
And once you master the skies what's to stop you from doing the same to the good old tankie? Now, this pesky auto-loader...
I don't understand why they bothered with an F-16 in the first place
Assuming it isnt already I wonder how long until the AI can learn to dogfight without the limitations of a human onboard. I expect a fair amount of weight can be removed for not supporting a human or two and manoeuvres would be less restricted by what the human body can handle and instead what the airframe can tolerate.
I expect the logic/algorithms for actually conducting a dogfight are relatively simple (and probably already exist to a greater or lesser degree in the world of video gaming). The key factor here will be the ability to identify and track the other party in the dogfight - as autopiloted cars have shown, this can be quite tricky at normal road speeds on a flat highway, never mind at high speed with movement in all 3 dimensions.
@Hans Neeson-Bumpsadese
"this can be quite tricky at normal road speeds on a flat highway, never mind at high speed with movement in all 3 dimensions"
Not sure why anyone would downvote you for that. I wouldnt be surprised if it required some sort of combination tracking system. Maybe ground radar and AWAC's. Of course you are very right about the speeds if you have gone with understatement :)
@AC
"Someone been watching 'Stealth' over the weekend, Hmm?"
As much as the film gets slammed (I get why) I dont consider it a bad film if you are willing to accept it as the kind of daft thing it is. But no the last one was Maverick which did mention the issues of G forces on the body.
Weight matters a lot on aircraft landing on carriers (was this a Typhoon issue years back?) where the payload had to be dropped to perform a landing? Probably got some of the details wrong on that one. Might even help with fuel consumption too if the pilot and 'life support' is removed. From my armchair position of never flying one and only ever reading about such aircraft.
@AC
" The navy came up with SRVL to deal with any potential F35(B) loaded landing issues on the UK's (not-really) aircraft carriers."
I havnt been keeping up with it. I thought they had to dump payload before conducting that manoeuvre but as I said I have not been following it. I do wonder what a difference might be made without the life support aspects of the aircraft.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
