Hacker News Reader

My experience in the field is about 15 years old at this point, but here is my take on it. Your supposition is premised on one of two things: (1) drone AI that can out-think a human; or (2) perfect communications between drones and an operator on the ground.

These are unsolved, and very difficult problems. The state-of-the-art right now is that we have trouble getting back the video feeds from our reconnaissance drones due to limitations in the downlink system. Drones also create new attack vectors--you might be able to ground hundreds of drones by taking out a handful of satellites or communications hubs.

It is possible that these obstacles can be overcome. But waiting on technology to be where you need it to be can be a very iffy thing. (Cue: "we were supposed to by on Mars by 2025."). I certainly wouldn't gamble the national defense on it.

> drone AI that can out-think a human

In my mind they are always remotely controlled, not only because I think it's extremely dangerous to have an algorithm pulling the literal trigger, but also because of what you mentioned.

Hence the "or". Either they're smart enough, or you need to solve the communications problem.

This, by the way, is why all RAF pilots are officers. Mere enlisted are not to be trusted with nuclear weapons, back when we had tac nukes. In WW2 lots of pilots were sergeants.

I should be clear that I think this policy is classist nonsense. There’s no reason that posh boys from Eton should be considered more trustworthy than anyone - probably the opposite in fact

It seems like what you need is a manned fighter that stays in the air some distance behind a fleet of semi-autonomous drones doing the actual fighting/killing. The pilot acts as an officer, selecting targets and providing general orders, while his or her plane provides the short-range encrypted comms which are hard to intercept or interrupt.

I like this idea. Sort of a "Zerg swarm" approach.

If you are going with Starcraft analogies, what has been described is exactly the Protoss Carrier.

One thing I’ve often wondered: are there any video games that can be considered “classics” in the sense that people will make reference to them and expect that most of a broader-culture audience will pick up on them?

I ask because I really want to start using starcraft metaphors in discussing society. The planting of the psi emitter on Tarsonis is such a powerful example of a human being willing to defect hard against his fellow humans in a way that is seen in history but I don’t think I’ve seen in popular fiction.

Did you mean: air to air missiles

EDIT: active or semi-active guided

Probably more like flying missile-pods?

When does the recursion stop?

I believe this is actually something that the USAF is already investigating. https://www.google.com/amp/s/www.popularmechanics.com/milita...

Non AMPutated link https://www.popularmechanics.com/military/research/a26028/f-...

That's supposed to be one of the future features of the F-35 Joint Strike Fighter:

https://medium.com/war-is-boring/the-navys-new-drone-will-be...

https://www.military.com/daily-news/2015/05/19/air-forces-ne...

But the F-35's software still doesn't work right, let alone the F-35's other issues:

https://medium.com/war-is-boring/the-f-35-is-a-terrible-figh...

I speculate there may one day be a place for a high-flying, well defended drone pilot carrier, whose purpose is to get the pilots close enough to the drones to have guaranteed instantaneous remote control - not even a satellite delay.

There might be, say, 6 pilots in that ship in VR rigs controlling the drones. They might be controlling dozens of drones a reasonable distance away, which are acting sometimes autonomously and sometimes under remote control, as needed.

Carrier has arrived. /Protoss-voice

Also, until full AI control, it seems that operators not in the field have subpar performance if not opposing desire (see the drone operators going deeply depressed)

> drone AI that can out-think a human;

But doesn't this assume that there will be some sort of 1v1 matchup between the pilots and the drones? I don't see any reason why this should be true, as the drones should be able to be produced much cheaper. So sure, maybe I human pilot could always shoot down a single drone, but how would they go against a swarm of 50 drones all perfectly coordinating with each other.

A fighter jet is still a fighter jet and just getting rid of the pilot isn’t going to make them cost 50x less (just like future autonomous cars aren’t going to be cheaper than what we have now) Also I don’t think the point of op was necessarily to envision a 1v1 scenario but that the AI needs to be at the level of a human more so becouse it is carrying extremely lethal weapons that one would prefer to be used under human (or equivalent) judgement.

People have been predicting the obsolescence of the manned fighter aircraft since the '50s, when guided missiles went mainstream. It even sort of happened in the Soviet navy, which had no fleet carriers, but instead cruisers which bristled with enormous, sophisticated anti-aircraft and anti-ship missiles.

But it still hasn't happened for real. Maybe this time will be different. But maybe not.

What happened with missiles is that, on the one hand, they turned out not to be the universal weapon that some thought they would be, and on the other hand, they were co-opted by manned aircraft, which are now platforms for launching and controlling them. Perhaps something similar will happen with drones?

There's already an emerging air power strategy where your stealthy fifth-generation fighters infiltrate enemy airspace, while your fourth-generation fighters, with their large radars and impressive arsenals, hang about beyond the reach of enemy missiles, supplying information and lobbing in shots to be guided to their destination by the vanguard. Drones could replace manned aircraft in the rear echelon, and perhaps augment the front line; being more disposable, they can risk turning their radars on.

Just curious - what's the point of stealth aircraft infiltrating enemy airspace when they can't use radars or they would be shot within 30s? Seems to me like just bombers make sense - fly in, enable radar/weapon bay for a few seconds, drop the load, make yourself undetectable again, move out; once the air defense is down, send in 4th gen fighters.

Attacking enemy ground anti air positions can be a job for stealth fighters. It is a high risk assignment. But there are few alternatives given the range of mopst AA defenses. These missions often requires the attackers to make their presence known in a controlled way to get enemy targetting radars turned on so that these stations can be tracked and targetted. Once the targetting radar is down, the rest of the AA battery typically becomes pretty useless.

Low observable aircraft can use passive sensors for some purposes, and can also accept data links from other offboard radar platforms. The latest radars are also difficult to detect and can be turned on for a few seconds at a time.

https://en.wikipedia.org/wiki/Low-probability-of-intercept_r...

Future radars will probably be designed to emit signals that look like random wide spectrum noise, and then rely on advanced signal processing to assemble the reflected "noise" into a useful picture. It will be challenging for adversaries to detect those emissions in a useful way.

You could do a sort of reverse wild weasel attack, where instead of sending in an aircraft ahead of the attacker to light up air defenses, your attacking aircraft could already be there and seeing more of what's going on based on the air defenses interest in the 4th gen aircraft.

http://mistyvietnam.com/ https://en.wikipedia.org/wiki/Wild_Weasel

In the sixties, the gunless fighters had problems and there was a return to more maneuverable fighters with guns. But missiles today have quite different capabilities compared to those in the sixties and seventies.

Even in 1991 basically it was all missiles. https://en.wikipedia.org/wiki/Air_engagements_of_the_Gulf_Wa...

If you start a manned figher program now you might be able to retrofit autonomy later, otherwise you have a state-of-the art manned fighter.

If you don't and bet on drones, you are betting on uncertain developments in artificial intelligence and the hope that you'll develop a framework in which you can use weapons that make decisions for themselves and are left with your current, then out-of-date planes if you are wrong. And likely spending a bunch of billions in the process as well. Todays mostly remote-controlled drones with some autopilot features are a very different beast from what you'd need to generally replace human pilots.

Ok, fair points. The estimated cost of replacing the trident nuclear deterrent is somewhere north of £25bn so I guess £2bn is "cheap" anyway.

I take your point about uncertainty of AI, particularly in light of growing concerns that self driving cars are a lot further off than predicted.

But in regard to betting on drones being risky, surely betting on old fashioned manned planes is riskier?

"Old-fashioned manned planes" work. It's not riskier to bet on the sure thing that has an upgrade path (in the event that drone AI does make the jump, it's not like a modern fly-by-wire plane won't be able to be retrofitted) to the potential new thing.

I think the quality of AI necessary really depends on how you structure the force, and what you're fighting. It's not like they'd be out there doing Top Gun style dogfights. They'd most likely take the role of high-altitude bombing and launching missiles from over the horizon, guided by AWACS.

> They'd most likely take the role of high-altitude bombing and launching missiles from over the horizon

Not really a fighter role, which article and parent explicitly talk about (e.g. otherwise "turn harder" isn't a relevant benefit). In other roles partial remote-control is more viable, and less intelligence needed, on the other hand the benefits maybe not so great - maybe cost primarily, once it is figured out? I'd guess a bomber could operate in slightly more dangerous situations and wouldn't need a full crew being trained and paid, but you'd loose flexibility and the crew is likely not a big impact weight/aerodynamics wise.

> It seems like manned fighter bomber aircraft are fast becoming an outdated concept. Drones can be lighter, turn harder and crucially not endanger the life of the operator. Why is the MOD committing £2bn to building new ones? Fighter aircraft historically take decades to be developed.

That's a speculative assumption. Predictions about future technological developments are notoriously unreliable.

Unmanned planes, as they currently exist, are all remote-controlled for the most part. That leaves them far more vulnerable to attacks against the communications links than manned planes. If your adversary has a successful attack against the link, your fancy all-remote Air Force is grounded. Remote control is a massive Achilles heel.

Right now, AI controlled planes are no more than a speculative fantasy.

Given that reliable military defense is such a critical capability for a nation state, it makes the most sense to continue investing heavily in the state of the art of proven technology. Investment speculative technology is extremely risky, and therefore should only be done carefully, in moderation. You don't bet the farm on it.

Hang on. Has there ever been a high performance fighter drone? All the UAVs I've seen aren't particularly agile nor fast. A remotely controlled drone might be limited in its performance by the lag time in input and output (first relay sensor data to operator and then wait for controller input). It's one thing to remotely pilot a drone loitering over a battlefield at 100 to 200 miles per hour and quite another to do it when they're going twice or three times that speed (not to mention above Mach 1 and super-cruising). Therefore a high performance drone must by necessity be AI controlled at the local level rather than wait for operator input. I'm not convinced that we will have AIs that smart (in addition to being able to make various decisions that are literally life or death) before this plane is being deployed.

What you're proposing is basically trading one set of problems for another and 5th generation fighter planes have two, possibly three to four, working examples. A high performance fighter drone does not exist in reality as far as I know.

You're looking for the Boeing X-45 UCAV (test program completed mid-00s) and the Boeing Phantom Ray (current project).

Oh wow. That's super interesting. I see how they go around the problems I mentioned. Their AI don't need to do any crazy maneuvering because it's really stealthy so it's avoiding any sort of dog fights. All the actual fast maneuvering is done by the munitions, which already has enough intelligence to hit most targets. Brilliant. Totally not how I imagined the problem would be solved but my thinking simply removed the pilot and replaced it with an AI whereas Boeing moved the game to a different field with a different set of rules.

"said the jet could be used with either pilots or as a drone."

As for your brexit comment:- "He added that the UK, currently excluded from the latest fighter programme underway between France and Germany"

So hardly fair to say its posturing by Theresa May when you factor that in.

From other sources, it does not appear that UK is actually excluded from this programme. On the other hand, Dassault are the lead in the Franco-German project - that could be an issue, I don't know.

>the jet could be used with either pilots or as a drone.

Seems they are going dual purpose. Fighters weigh about 30 tons so whether you have 100kg of pilot of not won't make a huge difference.

...cockpit, displays, controls, ejector seat. Designing it so the G-Force won't squash the pilot, strengthening the weak and vulnerable cockpit.

Supporting both seems wrong to me.

Though given we don't have full pilot replacing AI yet it might be an acceptable compromise like the way current self driving systems are mostly bolt ons to regular vehicles.

The first sentence addresses this:

  "...said the jet could be used with either pilots or as a drone."

It looks like they're going for optionality at this point, and have both bases covered.

Missiles have improved so much that being able to turn a little harder is of marginal benefit. Even if you remove the human pilot, the airframe is still subject to load limits.

£2bn is just for initial design work and won't go very far. The full design will be much more expensive.

I'd think not needing to engineer in a pilot might allow for an airframe engineered for higher load limits, though.

Sure but that just doesn't help much since the missiles are so much faster and able to turn even tighter. And anyway, most aircraft are now destroyed on the ground rather than in the air so it seems questionable to spend a lot of extra money to build stronger airframes for drones to address what's essentially an edge case. Better to focus on low observability to avoid being detected and shot at in the first place; if you're trying to avoid a modern missile coming at you then you've probably already lost.

> Pouring £2bn into a fighter in 2018 seems like a completely political move, a Brexit bluff and mere posturing from Theresa May.

That is exactly what it sounds like to me. Some general handwaving about as-yet to be invented or developed technology, references to the glorious past, a dig about Germany and France excluding the UK from a current project, and a non-functioning mock-up of what it could possibly look like. But other countries are welcome to join the effort. Whilst vigorously pointing at all that smoke and mirrors, mentioning that what they are really going to be doing is a series of upgrades to the Typhoon (that they would have planned for years ago already).

This is Brexit politics.

Yes, see some of this dramatic footage:

* Refuling drones that land and take off of aircraft carriers:

https://www.washingtonpost.com/news/the-switch/wp/2017/12/29...

* X47B killer drone take off and landing on aircraft carrier:

https://youtu.be/DOzUrHZWmus

Hm, what do you want to say, with your videos? Those are remotely controlled drones. Nothing to do with AI.

Current drones have sophisticated flying capabilities already. The X-47B was the first done to fly to and land on a carrier. It did so five years ago. Unlike the other drones in use by the military, the X-47B is not piloted, but is described as semi-autonomous and is even capable of autonomous in flight refueling. It is much larger than other military drones and more comparable to a manned fighter. Only the most skilled pilots can land on an aircraft carrier (the time I was flown out to a carrier it took the pilot several passes to catch the arresting wires) so the fact that a drone can do so with only minimal human direction indicates the very impressive computer managed flight control that has already been achieved by drones.

The X-47B is a major accomplishment in done aeronautics, from [1]:

> In March 2014, the X-47B won the 57th Annual Laureate Award for “extraordinary achievements” in aeronautics and propulsion hosted by Aviation Week. On 9 April 2014, the National Aeronautic Association selected Northrop Grumman, the United States Navy, and the X-47B's development team as the joint recipients of the 2013 Collier Trophy for excellence in aeronautic technology.

From [2], “The X-47B is the world's first autonomous warplane. From takeoff through landing, it flies with little or no direct control from human handlers.”

[1] https://en.m.wikipedia.org/wiki/Northrop_Grumman_X-47B

[2] https://www.popsci.com/technology/article/2012-07/i-am-warpl...

Communications keeps being the problem with drone fighters. And situational awareness has always been the problem with air combat in general.

Maybe the next military aircraft project should be a drone AWACS? Design a radar/comms package that can be mass produced at low unit cost, and a drone to carry it, and spam enough of them over the battlefield that if a few of them get shot down, the rest will still maintain surveillance and network coverage. There may be little experience with autonomous fighter planes, but there is plenty of experience with autonomous network nodes.

Drones are a good fit when an adversary doesn't have the capability to disrupt the communications infrastructure. Manned aircraft are a deterrent to actors with this capability.

Semi or fully automated flying vehicles, which can maneuver at maximum speeds and G-forces, without needing to support a human pilot?

We already have those: they are called missiles.

I think we should be realistic about who or what the threat is: and these days, for Britain, there is no threat, and no empire to protect or trade lanes to keep open anymore.

If they want to be lobbing bombs against less-advanced enemies, prop-driven modern CAS is a good investment. Long loiter times and cheap to maintain.

Remote controlled drones have another vulnerability that can be exploited: their communication channels.

If you figure out a way to jam or overload their communication channels, you just neutralized them. Unless they figure out a way to make a wireless signal unjammable or have the AI completely negate the need of an operator, the need of an operator inside those fighters is still relevant.

That's assuming the primary mission of an aircraft is combat.

But how true is it that the primary mission is combat?

The other problem is latency.... nothing you can do about physics when it comes to remote control. Remote control versus fully autonomous raises a bunch of other questions.

You're wrong because "drones" of general utility are mostly a marketing meme.

Any 5th generation Western fighter is probably going to be an improvement over the F35, so there is a market gap to be filled.

Pouring £2bn into a fighter in 2018 seems like a completely political move, a Brexit bluff and mere posturing from Theresa May.

I doubt it is related at all, Italy is a partner. Unless you are commenting on Five Star?

What happens when you’re communications with the drone is jammed? Or satellite relay system destroyed?

The communications issue is a very common theme in this thread, but I don't understand why it's an issue. Once the UAV is in the air, why do you need to communicate with it? You can't communicate with a bullet once you've pulled the trigger.

Drones aren't used like bullets. They're not fire-and-forget.

The timing isn't just to do with the Farnborough Air Show. When you have MPs launching the project rather than the partners announcing they have backing then you can see where the impetus is coming from. BAE have always got a next generation plane/boat/tank/missile on the drawing board. There is just a hand-wavy pledge by government to spend on the development, they have not ordered any planes or coughed up hard cash.

Post Brexit Airbus could pack up and leave the UK. Interestingly the Broughton factory in Wales used to be in England but the Conservatives arranged for the border to be re-drawn enabling Airbus to get an EU regional development grant for their factory.