An interesting step towards the distant future. However, this was an imperfect combat simulation in that AI had flawless SA. Right now and into the near future, I don’t know anyone who can get that kind of information onto an aircraft.... manned or unmanned.

When this happens in real life it will represent a major leap forward in several fields, but that’s 20 years out at least, likely further. Until then it’s still just theory and I’m not too excited. A computer winning a computer game is not new.

I do AI work, mostly related to the financial markets.

Most AI systems don't really do anything spectacular. They take common ideas and concepts and execute them to perfection. No second guessing, no hesitation.

The models for these systems will be the best examples the programmers can find. So, you'll be competing with the best there is.

How do you beat them? You know the same things they do. There are always exceptions to strict rules - but you have to know the rules being used.

V.

quote:

Originally posted by x0225095:
As the title says....interesting story.

https://www.foxnews.com/tech/a...ts-human-clean-sweep

Edited to link original story ...

https://www.airforcemag.com/ar...ilot-in-darpa-trial/

Now do surface Navy. They seem to be the ones that need AI to drive safely...

quote:

Originally posted by sigcrazy7:
An interim solution would be a fighter swarm, where one pilot has two or three AI fighter escorts. The pilot would provide the judgement and overall control, while the AI planes would coordinate their attacks on the target. There’s at least a SciFi movie in there somewhere.

It's been done. Buck Rogers had an episode where there was a swarm of enemy fighters flown by barely-trained humans, under psychic control of the 'expert' pilot. Buck figured out real quick how to deal with this - he pushed through the swarm and killed the expert pilot. Once he was dead, the swarm fell apart and became instantly useless.

Same concept would apply. Find the human pilot and kill him, then the swarm becomes just another 'dumb' AI.

Drones/UAVs/AI have their uses, but I still have not seen it *demonstrated* where a UAV can be superior (or even equal) to a human in the aircraft. Maybe some day, but not today.

This was a technology demonstrator, showing what might be possible. I will be the first to admit that AI/machines have distinct advantages over human-operated equipment. Faster reflexes, higher G-tolerances, lighter weight (without all the considerations for keeping a human alive, such as ejection seat, canopy, oxygen/pressurization, human/machine interfaces, etc) (this is especially important for space-based applications), etc. They are truly 'expendable,' since there is no aircrew safety to worry about. Missions could be one-way, dramatically increasing their range.

In the late 1970/1980s, Rockwell came out with HiMAT (Highly Maneuverable Aircraft Technology) remote-operated aircraft. It was smaller than even the relatively small F-16, had wicked maneuverability (much higher G-force capability than any manned aircraft), but was severely limited by the flight control limitations of the day. Add a truly capable AI to this sort of airframe, and then we will see something wonderful.

Good post Hound Dog.

2 additions on higher G tolerance: first it remains theoretical that we could build a UCAV with the sensors & processing capacity that also withstands higher G forces. It’s not a given that we can cheaply (or at all) engineer something like that.

Secondly, even if we could... who cares? Are we expecting an advanced AI to get into a gunfight?

quote:

Originally posted by arabiancowboy:

Secondly, even if we could... who cares? Are we expecting an advanced AI to get into a gunfight?

I believe that, eventually, we WILL have AI gunfights. History has shown that we have a VERY poor track record of predicting what the 'next' war will look like. We keep trying to re-fight the 'last' war, and as a result, we always seem to have to scramble to adapt to the new/current fight.

Every major war saw significant technology/strategy/tactics changes. WWI introduced the tank, combat aircraft, and submarine as efficient weapons. WWII saw the battleship lose dominance to the aircraft carrier.

Back in the late 1950s/early 1960s, the US military thought the 'dogfight' was a thing of the past. All fighters would do is shoot radar-guided missiles at Soviet bombers beyond visual range over Canada. Therefore, when we got to Vietnam, we had an entire inventory of aircraft that were grossly unsuited to 'modern' air-to-air combat. Robin Olds lamented that one of our most numerous 'fighters' had a bomb bay (the F-105; which was really a tactical bomber more than a fighter). We had to learn how to 'dogfight' all over again, armed with crappy missiles that were entirely unsuited to fast-maneuvering fighters.

The Sidewinder and Sparrow missiles had horrible reliability rates. I did a bunch of research on the Vietnam air war, and IIRC the Sparrow had about a 30% hit rate. These missiles were tested against slow un-maneuvering bomber-sized targets at high altitude over the Western deserts. Then, they were surprised they did not work over a humid jungle at low altitude against small violently-maneuvering targets. The Falcon missile was a joke - it had a liquid-cooled heat-seeker head that had to be activated a full minute before it could be fired. That is an eternity in a dog fight. Oh, and it did not have a proximity fuse; at least one account I read told about one Falcon missile that flew over a Vietnamese MiG about a foot above the enemy's canopy.

The early F-4 had no gun at all; nobody 'needed' them since all the pilot had to do was push a button and let the missile do all the work. Later in the war, American fighter units added gun pods, and later they finally got an internal cannon on the F-4. The Navy began training its crews for dogfighting, and the USAF followed later on, dramatically improving survival rates and enemy kill scores. It took the US YEARS of war to finally adapt to the new reality of air combat.

Point being, it is only a matter of time before an enemy starts fielding AI fighter aircraft. If we have nothing comparable, we risk falling behind. The US has, for the past 70+ years, placed its emphasis on superior equipment and superior training to compensate for larger numbers of less-trained enemy aircraft. If China or Russia (or even North Korea) were to develop an effective AI fighter platform, and we didn't, the US technology lead would greatly diminish. I don't know what that would look like, but I'm sure we don't want to risk it.

It seems to be something AFRL is looking into: https://en.wikipedia.org/wiki/Kratos_XQ-58_Valkyrie

quote:

Originally posted by BBMW:
That's more a sensor issue. Early UAVs were equipped with high magnification day and infrared cameras that had very limited fields of view. This is why I brought up the idea of equipping a UCAV with continuous 360x360 degree sensors. It would essentially have infinite peripheral vision. This would be backed up with a trainable high magnification sensor.

quote:

Originally posted by sns3guppy:
Something we frequently ran into in Iraq was that unmanned equipment would lose targets, especially mobile targets. The unmanned equipment, even when being operated by a man, had distinct disadvantages, not the least of which was a lack of peripheral vision. I've been assigned to find targets that the unmanned assets lost. I've also seen cases, quite a few times, in which unmanned equipment missed basic stuff; an explosion or flash, for example, which was picked up by pilot peripheral, but not by the UAV.

The UAV can have a decision tree, but can't have judgement. The pilot gets paid for judgement.

Skynet is fine if you hard wire in off switches.

And I would point out that The Terminator is a work of fiction.

quote:

Originally posted by limblessbiff:

quote:

Originally posted by BBMW:
I assume this was in a simulator and both the real and AI pilots were flying equivalent simulated aircraft.

I've thought for a while that the AF should take th basic current F-16 design, remove the cockpit beef up the airframe to take 12-15G, add 360x360 multispectral EO sensors, and set it as an antonymous UCAV. This is especially true since it seems like they now have the software working.

Do you want Sky Net? Because this is how we get Sky Net

quote:

Originally posted by BBMW:

And I would point out that The Terminator is a work of fiction.

Science Fiction often (but not always) becomes science fact.

Just a few examples:
HG Wells wrote about atomic warfare, back in the late 1800s.
Jules Verne predicted atomic submarines that could travel at incredible speeds almost indefinitely.
Star Trek predicted auto-opening doors, 'speaking' computers (like Alexa), iPads, iPods, directed energy weapons, and a wide variety of other cool technologies.

I think it is self-evident that any true AI just MIGHT start to think for itself. We aren't there today (that I know of), but no man knows what tomorrow may bring. . .

quote:

Originally posted by 18DAI:
They should have used a Navy pilot. A hung over Navy pilot would be even better. Would have kicked that drones AI ass then. Regards 18DAI

Nah... The Navy aviators would be too distracted looking forward to the shirtless volleyball game after their sortie.

Its not nearly so simple as AI beats human in simulation therefore its better in actual combat than a pilot. Former F16/18 and current Red Air pilot’s take below:


https://m.youtube.com/watch?v=ziCQqmEllZo

quote:

Originally posted by Hound Dog:

Science Fiction often (but not always) becomes science fact.

It really doesn't. Simply because a figment of imagination and a fancy of flight got the guess right, doesn't mean that science fiction informs the future.

This is a widely had debate in writers circles, particularly in the science fiction writers crowd, where it's nearly heresy to suggest that it's just fiction.

I had some of that debate with a woman writer some years ago, who had done a lot of work with Star Trek. She insisted that Motorola had patterned the flip phone after star trek, and therefore the writing was a direct influence into the future. Indeed, the thin tablets used for writing seemed ridiculous when they were first broadcast, but later with the iPad and other mimics, it doesn't seem so far fetched, now.

The fact is, Motorola did model their phone after the communicators on Star Trek. This was a sales gimmick, a way to appeal to the masses and sell phones, not fiction writing the future.

The six million dollar man was fiction, and yes, I did have a Steve Austin lunchbox when I was kid (and an evil kinevil lunch box, too)...but it's just that, fiction. Any semblance with the future is coincidence, or sales. Follow the money. Imaginary guesswork cum pop literature does not translate to technology, nor advance. Sorry, fiction fans.

I enjoy my writing, but am not foolish enough to believe it creates more than words on the page, or a conjured image in the readers head. Beyond that, of course, it's not real. Thank god. Fiction doens't become fact.

There's a point not being discussed here. It isn't so much whether the AI pilot is better than a human pilot as far as decision making, tactics, etc.. It's that it gets the human out of the plane, and removes the need to have the human still doing stick and rudder remotely.

By doing that they remove all the negatives of humans. The human brain will crush itself in the skull at anything more than 9G, IIRC. So this becomes a limitation on the maneuverability of jet fighters. Get the human out of the cockpit, and they designers can instantly turn up the G the plane can handle. Also humans have limited sensory perception, and limited ability to process sensory input, but in quantity and speed. Using artificial sensors and feeding that to human pilot compensates for that to some degree, but again that's limited. A computer can process much more data and at higher speed. Computers don't get tired. If refueled, an autonomous UCAV could stay airborne, and pin sharp, for days. Each human pilot has to be individually trained. This is a multiyear, seven figure cost project for every pilot. And eventually the pilot will leave the service, or can get killed in action. Developing the AI piloting system will take longer and cost more. But then it can be simply copied into the computer of each aircraft, no matter how many there are. If an aircraft is destroyed, build a new one, and copy the AI software into it. I'm sure there are other advantages I've forgotten or haven't thought of.

So even if the AI just gets close to the ability of a human pilot, it starts with advantages already baked in.

quote:

Originally posted by erj_pilot:
Oh Jeezus..

Yep, here we go again....

It's like sendec reborn

quote:

The human brain will crush itself in the skull at anything more than 9G

FYI - in 1954, John Stapp took 46g in a rocket sled.

Sounds like the movie Stealth, which was on last night, a 2005 failure at the box office,

EDI- UCAV is Extreme Deep Invader a complete AI plane... vs three top pilots in F/A-37 Talon, a single-seat fighter-bomber with amazing payload, range, speed, and stealth capabilities.





It's on Showtime 8/28 at 12:45 AM and 3:45 am in case you missed it. Still interesting it's out again while this discussion is going on...

You don't build AI piloted aircraft to deploy weapons, you build AI weapons that can fly and loiter as far or farther. Building many more weapons than a single aircraft for same money.

In WWII, aircraft were used to intercept Buzzbombs, now Iron Dome, Patriot batteries can effectively defend against many simultaneous and varied missiles and rockets at once. Scale that sort of thinking up for both defensive and offensive. Rather than limited numbers of high dollar aircraft to bring weapons to a theater or distant threat.

quote:

Originally posted by sns3guppy:

quote:

Originally posted by Hound Dog:

Science Fiction often (but not always) becomes science fact.

It really doesn't. Simply because a figment of imagination and a fancy of flight got the guess right, doesn't mean that science fiction informs the future.

Yes, it really does.

My point is, that scifi concepts such as atomic weapons, nuclear-powered submarines, and AI systems like Skynet (to quote myself) often (but not always) become science fact. That is an important and indisputable distinction. I never said "all" scifi elements come to pass; just many of them. So what if they are 'lucky guesses?' The fact remains that Wells and Verne wrote about things decades or even a century before they came to pass.

Heck, look at Wonder Woman's invisible jet. That was comic-book fantasy in the 1960s; yet, we have had these sci-fi 'invisible aircraft' flying since the 1980s with the F-117 and B-2 bombers (and who knows what other black projects we are operating outside the public's view?).

I never implied or stated that Wells' writings directly birthed the Manhattan Project or Verne the Nuclear submarine - just that MANY things in scifi end up becoming reality. The guy who invented the iPod did so after watching a Star Trek/Next Generation episode, marveling at how they could push just a couple buttons and play a song from a vast library. Heck, space flight was considered science fantasy well before Yuri Gagarin proved it could be done by humans. I am not saying that EVERY scifi element will come to pass; just that a great many of them (more than I can count) already HAVE become reality.


Even the concept of AI itself used to be 'science fiction.' Isaac Asimov came up with his famous "Three Laws of Robotics" in 1942 for his works of fiction.

To quote Wikipedia: https://en.wikipedia.org/wiki/Three_Laws_of_Robotics

quote:

The Three Laws, and the zeroth, have pervaded science fiction and are referred to in many books, films, and other media. They have impacted thought on ethics of artificial intelligence as well.

See, we here have another example of science fiction being used to inform the future of yet another cutting-edge technology. AI was impossible in 1942 when Asimov wrote this. Yet, here we are today discussing how an AI pilot did against a flesh and blood pilot. This could not have happened even 10 years ago.

I missed quite a few of the fine points made due to not understanding much of the jargon/acronyms but pilots may find Lemoine's take interesting.

There's a not-bad book called "Day of the Cheetah" by Dale Brown loosely based on this concept - a prototype AI fighter jet goes rogue and an experimental neural interface fighter with a human pilot has to try to take it down.

The concept of AI in the book has nothing to do with reality, though.

quote:

Originally posted by Hound Dog:


Yes, it really does.

My point is, that scifi concepts such as atomic weapons, nuclear-powered submarines, and AI systems like Skynet (to quote myself) often (but not always) become science fact. That is an important and indisputable distinction.

Hence the eternal debate in writing circles: does science fiction writing inform the future. The answer is a resounding "no."

While certainly someone may write about something that does not exist, which is later invented, there is no connection between the imaginary fictional events, characters, and things of the short story or novel, and what's later developed by science, as technology.

Science may yet develop Scotty's transparent aluminum (already has, actually), and Wonder Woman's glass airplane (a bit of a stretch to refer to the BDSM cartoon "Wonderwoman" as "science," but so be it). If Science does develop a glass jet, for reasons one can hardly fathom, it won't be the result of a 1940's cartoon. It will simply be the logical progression of science.

Galaxy Quest was a satire on the concept of developing technology based on fiction. While humorous, it's a dumb idea, the irony of which was the focus of the "Omega" machine. Not even the characters knew what it was, and as a result, no one knew what it did, and when they tried to develop it technologically from the fiction, it didn't do anything...because it was never explained in the fiction.

We may yet get phasers that are settable to stun...but it won't be a result of Gene Roddenberry's creativity, or for that matter, anyone trying to design what Roddenberry wrote. Just technology marching on.

In the meantime, we have flight control systems with varying degrees of automation that can in nearly every case, fly the aircraft more precisely than the pilot. And we use them extensively. What they can't do is replace the pilot, nor should one try.