Hacker News Reader

Comment by hailwren | Hacker News Reader

The fifth generation of fighter aircraft marked a significant departure from the idea of a fighter jet being anything like what we think of as a fighter jet.

They're now designed and employed as fighter 'platforms'. Instead of zipping around the battlefield guns and missiles ablaze in 1-on-1 combat, they're low observable long-range systems designed to understand and disrupt the battlefield by employing their electronic and long-range missile systems.

These AI are a natural extension of that. I imagine the 6th generation of fighter platforms will be commanding swarms of fighter drones to do the fighter part of their role.

View on HN

The US Air Force is already looking at this. I believe the plan wouldn't be to tie it to a specific "generation" of fighter, and will allow fifth, and maybe even fourth generation fighters to lead a "swarm" of fighters.

https://afresearchlab.com/technology/vanguards/successstorie...

The British Navy were already looking at fully automated dronecraft carriers back in 2015-16 too. Using smaller drones as armaments instead of missiles and the like.

FTA

"The overarching ACE concept is aimed at allowing the pilot to shift “from single platform operator to mission commander” in charge not just of flying their own aircraft but managing teams of drones slaved to their fighter jet."

Such a concept is well under development as the Air Force Loyal Wingman program. Boeing unveiled protoypes not long ago [1].

https://www.boeing.com/features/2020/05/boeing-rolls-out-fir...

Sounds like a Protoss carrier fleet from StarCraft.

Let's go back even further https://www.youtube.com/watch?v=BeZNoEVycOc&feature=youtu.be...

(Ur-Quan Dreadnought from Star Control II)

So, Ender’s Game?

This is likely to cause several interesting problems in my mind. For instance you may be about to sacrifice a lot of current talent to actually "fight" these planes, since you're now mainly focusing on the ability to coordinate and micromanage the battlefield instead of being physically/mentally able to assume the physiological role as primary control unit of the aircraft.

Also, to be honest, these AI controlled fighter platforms scare the shit out of me because there is now potentially fewer human decision points in the system.

Like it or not, you can end up with many times the destructive power in the air orchestrated by 1 guy without having the requisite sanity check of "Excuse the hell out of me, sir, but you want me to bomb WHAT?"

The capability to look at a situation and decide to call off is a feature of warfare that I think is frankly underappreciated.

We may very well be working unintentionally toward creating a world where a small

Holy hell, it'd be nice if I finished my thought. We're creating a world where fewer and fewer people are capable of marshaling more and more destructive power.

This does not bode well in terms of the law of large numbers being able to temper the extreme characters that setup may invite.

In my opinion this system is exactly the scenario that they were talking about when a bunch of ML researchers and stuff signed that letter saying it was necessary to ban autonomous AI weapons.

And it is bizarre that no one seems to be recognizing this.

In the extreme scenario where an entire mission involving 1000 aircraft is lead by a single commander it is entirely possible that the lone commander was a spy and turns against his own country.

But why would that human commander have to be in the air, putting his life on the line more than when they would be further away and on the ground/in a ship?

The only answers I can think of are a) that having human eyes in the sky still has advantages, or b) that long range communication is too unreliable (either as is, or because of possible enemy interference).

You may need an AWACS to direct the battle, but would it have to have human in board?

But is either true? If so, what’s the reason? Is there a c) “we can’t tell the air force yet that pilots who actually fly are a thing from the past”?

> But why would that human commander have to be in the air, putting his life on the line more than when they would be further away and on the ground/in a ship?

Latency + Bandwidth.

John Boyd must be rolling over in his grave

Boyd wasn't just a pilot but also a strategist. One of his biggest adherent is the USMC. I don't imagine him rejecting this outright. I would love to know what he would have thought of AI because the potential speed with which AI can digest observations and make a decision could be an order of magnitude greater than a human. Boyd's EM theory was heavily influenced by analysis he and his collaborators did using early computers. I can imagine him being awed by AI and embracing it.

I can't wait for the first time an AI squadron gets a sortie phoned in and it turns out no one double checked to make sure the info was good.

Are there any metrics on pilot's making abort calls and not attacking anywhere?

I'd imagine one of the (forget the name, the big planes with radar that buzz around the battle space), could handle a swarm of drones from a long distance. Not sure that it'd make sense to have a speedy jet out there doing that.

AWACS is the role, E-3 Sentry is the current plane. It's a 707 full of radar, but it's also a giant flying target. If you're going to link a bunch of drones to one platform, the drone controller is going to need a lot more survivability.

The problem is going to be keeping an AWACS close enough to command and control the fighters while not risking getting it shot down. An F-22 or F-35 (or even an F-15) is going to be a lot more survivable than an AWACS.

> is going to be a lot more survivable than an AWACS.

AWACS have vastly superior radar and electronics. It's going to see threats long before they are in range. And it's going to direct interceptors to deal with those threats long before they are in range.

I don't believe an AWACS has ever been shot down.

This lists a few times when AWACS detected and directed the destruction of threats https://en.wikipedia.org/wiki/List_of_aircraft_shootdowns

Many advanced nations now have "AWACS killer" missiles with 400+km ranges. I think this is why the sensor suites on 5th generation aircraft are so good.

An AWACS has never been flown in an area where it is likely to get shot down. That RADAR is a nice juice target for a HARM, and the maneuvers you can do in an AWACS is not nearly what you'd be able to do in any fighter. I don't understand what you are arguing. Are you suggesting an AWACS would be more likely to survive an attack than a fighter?

I believe you're thinking JSTARs? https://en.wikipedia.org/wiki/Northrop_Grumman_E-8_Joint_STA...

Or maybe the E-3 Sentry aka AWACS, but the JSTARs matches the "big plane" better and has a giant synthetic aperature radar for imaging large swaths of the battlespace. We deployed with JSTARs when I went to Iraq as a Shadow 200 TUAV Pilot myself.

>but the JSTARs matches the "big plane" better

Sounds like you would know better than me, but aren't both the E-3 and E-8 based on a 707 platform?

And don't you need both? If you're doing an air superiority mission you need an AWACS. If you're doing CAS you need JSTARS, right?

Yes, you're actually correct here. I interacted often with the JSTARs crew (we were on the same base in Mosul for a few months) and not as often with the AWACS crews. JSTARs was the first thing I thought of when you said big plane with radar though as it is literally a big plane with the sole purpose of SAR. The AWACS is more or less a similar thing, but with additional EW (electronic warfare) kit and sweeps the sky.

They're complimentary although the AWACS is more relevant today than the JSTARs is. Most of the hardware in the JSTARs hasn't been updated in decades and it is starting to show its age. We had trouble integrating our realtime video feeds from our UAV with the telemetry feed from the JSTARs because it is just clunky and old.

The longer the distance, the greater the risk that the control data links will be disrupted by the adversary's EW.

Slightly off topic, but how do they disrupt gyroscopes ?

Gremlins concept, deployed from C-130?

Reply on news.ycombinator.com