Since we had this discussion about UAVs and F22s etc...

[Tom Ligon]

I'm in the UAV controls biz.

One of our cheap ones is capable of flying a sub-scale FA-18 with most of one wing missing. It figures out how to compensate for battle damage on its own. The box in question probably costs about the same as a day or two of flight training for a real pilot.

UAV control systems have to pass the same EMP requirements as the avionics on manned fighters, which are fly-by-wire. EMP requirements are actually not that tough to meet. Good luck killing one with EMP.

You might take one out with gamma or x-rays, but the level would melt a human pilot.

And they are autonomous.

As I said, I just love flying combat simuators, and I can whup the butt of the typical AI pilot in a typical PC flight simulator, but the days of dogfighting between human pilots are pretty much over.

[Betruger]

How do you compare with military grade AI pilots? Consumer PC AI doesn't seem like a fair comparison.

[kunkmiester]

Isomers are radioactive, so forget about that.

EMP is overrated in many ways. I think a lot of people will be surprised about what is and isn't taken out by a HEMP. Non-nuclear lower altitude devices are a different can of worms however, but have their own limits. Suffice it to say, the military has been spending a lot of money ever since te Corona(?) shot to protect their gear. The military, and a substantial amount of civilian infrastructure I imagine, will be "quite operational when your forces arrive"

DEW have a good ways to go before they get used more often. There are systems in use, but they generally target sensitive parts--overloading IR sensors on SAMs, or example--rather than destroying in general.

[Tom Ligon]

Exactly. I know I can outfight a PC when the planes are no more maneuverable than my plane and the pilots are dumber than stumps, and don't learn.

It is not entirely clear to me that I could out-fight something more maneuverable, harder to hit, and an idiot savant at air combat maneuvers.

The US will wish to maintain absolute air superiority, and will develop weapons systems to do that. For the forseeable future, that mix will include the multi-purpose smaller aircraft we call "fighters", with human crews. It will, however, increasingly mix UAVs into the team, and they will handle the dangerous work. Plus, the crewed fighters will have UAV capabilities in case the crew is debilitated.

[Betruger]

So I misunderstood - You meant that UAV AI isn't that good, but paired with its UAV performance, it definitely beats human pilot and plane in any dogfight.

[ladajo]

Isomers are radioactive, so forget about that.

EMP is overrated in many ways. I think a lot of people will be surprised about what is and isn't taken out by a HEMP. Non-nuclear lower altitude devices are a different can of worms however, but have their own limits. Suffice it to say, the military has been spending a lot of money ever since te Corona(?) shot to protect their gear. The military, and a substantial amount of civilian infrastructure I imagine, will be "quite operational when your forces arrive"

DEW have a good ways to go before they get used more often. There are systems in use, but they generally target sensitive parts--overloading IR sensors on SAMs, or example--rather than destroying in general.

The larger issue in the whole EMP/HEMP/SREMP effects is the lack of actual test data. Small scale and sim data does not speak to large area infrastructure vulnerabilities. Unfortunately, non-military and non-continuity of government infrastructure is not actively protected. A continental scale event has a high chance of catastrophic disruption to civilian infrastructure. Having a functioning military to fight is cool, but the potential millions of no food/no water/no power civilians after one week is a big ugly mess.
I agree that shielding for effects is not that hard, but unfortunately, there is no standard dictated to be met in the civilian sector. The really is no equivalent for TEMPEST in the civilian arena.
But bottom line is, until some angry weapon onwer pops off a megaton device over a someone elses country, we really will not know the actual depth of damage to the infrastructure. I don't know about you, but turning off the majority of power and clean water to New York City for a week or two gives me the willies.
The following link is a good public access doc on EMP hardening. Been around for I think about 15 or so years.
http://www.fas.org/nuke/intro/nuke/emp/toc.htm

[Tom Ligon]

Mil Std 461 is available on-line. It covers all sorts of electromagnetic performance issues, including EMP.

Yes, it is a military standard, but every time I go down to a lab to have these tests run, all sorts of consumer products are going thru the same tests. Among other things, products are being certified to not cause EMI/EMC problems on civilian aircraft. But you would be surprised at the number of civilian products that can handle a wicked whack of EMP.

I'm transmitting this on a fiber optic network, on a home with underground power, double surge protection on the house panels, isolated ground power to my computers, running on a protected UPS and an Iso-Bar. At the cabin I can run entirely off the grid.

[DeltaV]

First, show me a robotic lawn mower that can reliably mow a hilly, gopher-holed, in-spots-muddy, tree-spattered, leaf-strewn yard without shredding the dog, garden hose, bird feeder and car fenders, in all lighting, wind and traction conditions, and maybe then I'll start believing you about adaptive flight control algorithms that can handle everything a human opponent can throw at a UAV in air-to-air combat, in addition to the varying environment.

That sort of UAV autonomy is a long way off (about the time F-22s reach end of service life, in my estimation). AI research claims to the contrary notwithstanding. Jiggling the remaining flight control surfaces to estimate dynamic parameters of a battle-damaged UAV is a far, far cry from winning against an unpredictable human pilot, fighting for his life, who knows the UAVs weak points. Human pilots have been dodging SAMs for a long time. Algorithmically, an air-to-air combat UAV (currently nonexistent) would just be a hopped-up SAM, with less speed.

Flying killer robots, like the autonomous ones I once worked on, are good for fixed targets or big, slow-moving targets, like ships. I haven't seen much AI improvement in the last 25 years, mainly sensor, data link and materials improvements. The current man-in-loop UAVs don't impress me, either.

[Betruger]

He did specify dogfight. First paragraph in your above post sounds like the running (or latest at least) DARPA autonomous pathfinding challenge.

Dogfighting AI UAVs: I haven't seen ("flown" against) anything but consumer level dogfighting AI, so I can't comment. I don't think it's so far fetched as you say, though. No real AI improvements in 25 years sounds incredible.

[alexjrgreen]

No real AI improvements in 25 years sounds incredible.

The Cray supercomputer just installed at Oak Ridge National Laboratory is arguably as complex as a human brain. But not intelligent.

We have lots of party tricks, but no credible AI. Even MIT is going back to basics to try and find out where we went wrong:

Science goes back to basics on AI.

[Skipjack]

ladajo, you are talking about an EMP caused by a nuclear explosion. That is something very different to what we were talking about.
An EMP like that would fry any system, including those in the F22...
As Tom said, I dont see any reason, why UAVs would be THAT much inferior. Sure you might be able to jam them, you might be able to use an EMP on them. You might...
But then that would also render your F22 pretty useless, since it is relying to a large extent on the same systems a UAV relys on.
And again, if you can have 10 UAVs for one manned fighter, the odds are tilted a lot towards the UAVs. They are also smaller, making them harder to spot on rader and IR. If a UAV is lost, so be it, it is a small price compared to the loss of a pilot.
The pilot that was flying the UAV will simply walk over to the next desk to command the next UAV.
No pilots that need to be rescued from captivity or to be picked up behind enemy lines. It makes a lot things much easier.
You can also be much more ruthless with your UAVs. Lets sacrifice a couple of 100 UAVs in a decoy- attack here and then hit them with the full attack elsewhere.
Some generals might be willing to do that with human pilots, but it would not only be ethically difficult but also difficult for other reasons (politics, etc).

[Skipjack]

Alex, dont put to much weight on that AI- article.
This is a very different AI, they are talking about (more like AL).
There has been AI in flightsimulators and other games for decades now and it is getting much better all the time. That does not mean much, yes, but there has been a lot of improvement.

[Betruger]

Yeah, I do mean strictly dogfighting AI as Tom Ligon was arguing.

[kunkmiester]

So I misunderstood - You meant that UAV AI isn't that good, but paired with its UAV performance, it definitely beats human pilot and plane in any dogfight.

First, show me a robotic lawn mower that can reliably mow a hilly, gopher-holed, in-spots-muddy, tree-spattered, leaf-strewn yard without shredding the dog, garden hose, bird feeder and car fenders, in all lighting, wind and traction conditions, and maybe then I'll start believing you about adaptive flight control algorithms that can handle everything a human opponent can throw at a UAV in air-to-air combat, in addition to the varying environment.

Let's see here, an OODA loop measured in milliseconds, rather than seconds, plus the performance, yes, it should work.

Air is actually a much easier environment to operate in. Radar data is already effectively the ones and zeros computers like--you don't have to do much to make it identify and classify an object. You don't have to worry about gopher holes--there are none. Tactics for air-to-air combat are well known, and again can be rendered to ones and zeros.

This would be a ruthless analytical killing machine, not a flying Johnny-five.

[alexjrgreen]

Alex, dont put to much weight on that AI- article.
This is a very different AI, they are talking about (more like AL).
There has been AI in flightsimulators and other games for decades now and it is getting much better all the time. That does not mean much, yes, but there has been a lot of improvement.

The Victorians had clever mechanical automatons, and round the world yachtsmen had mechanical automatic pilots in the '60s.

Electronics is quicker and lighter. Not intelligent.

Even a single celled organism is intelligent, so we must have missed something.