AI Fighter Pilot Wins In Combat Simulation

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
Does anyone remember who the game developer was that said, and I'm paraphrasing, "developing AI that can beat you every time is easy, dumbing AI down so that it misses sometimes is what makes it more realistic." I think that applies in this situation.

An artificially intelligent fighter pilot system has defeated two attacking jets in a combat simulation. The AI, known as Alpha, used four virtual jets to successfully defend a coastline against two attacking aircraft - and did not suffer any losses. Alpha, which was developed by a US team, also triumphed in simulation against a retired human fighter pilot.
 
I'm really excited to see autonomous fighter drones in the future. Even if they were 80% as effective, the sheer number would be able to overwhelm. That and the ability to maintain 100% uptime over a theater. Never get tired. Never make mistakes.
 
how about getting things a bit more even? 1 AI with 2 virtual jets vs 2 comparably armed human controlled simulated jets
 
On the other hand, I feel like playing Command and Conquer Generals after reading this story.
 
Well Chuck Yeager said "It's The Man Not The Machine" in his video game

Chuck%20Yeagers%20Air%20Combat_1.png



Humans can't be hacked like drones already have.

A human can decide not to drop ordinance on a target based on moral/ethical challenges.

That being said I'm all for a Protoss Carrier pulling in.

 
So what, damn Japs have been killing me in 1942 since the 80's.
 
Humans are so damn arrogant, our species will create it's own demise eventually without a shadow of a doubt.
 
  • Like
Reactions: Liver
like this
"We're not going to make it, are we.... Humans, I mean"

"It's in your nature to destroy yourselves"

:D :) :D
 
Plus wouldn't the AI have the advantage of not having to worry about G forces which would cause a human pilot to blackout ?
 
They still would have to (planes would fly apart otherwise), but the breaking point is a lot later than a normal human.
 
I'm really excited to see autonomous fighter drones in the future. Even if they were 80% as effective, the sheer number would be able to overwhelm. That and the ability to maintain 100% uptime over a theater. Never get tired. Never make mistakes.
From Acceleration of Blackout in Fighter Pilots the number of g-forces a pilot can withstand are very limited. Autonomous drones have no such restriction. I expect them to dominate in the near future as long as they are hack and spoof resistant.
 
OK, so 2x AI vs. 1 human and they win? Wonder what would have happened if it was 2x human vs. 1 AI?

2 vs.1 almost always comes out with the 2 being victorious.

Also, how did the AI keep track of the human? Was it limited to radar and visual using the same parameters as what a human would use?

Not enough information given for me to believe that this was a valid test or not.
 
Well Chuck Yeager said "It's The Man Not The Machine" in his video game

Humans can't be hacked like drones already have.

A human can decide not to drop ordinance on a target based on moral/ethical challenges.

That being said I'm all for a Protoss Carrier pulling in.



It's only hackable if it's remotely controllable, which means there's a human on the ground somewhere that can decide to abort the attack.

That being said, if you're worrying about moral/ethical dilemmas when you're dropping bombs on targets far below you, then you're probably in the wrong line of work.
 
Automating military equipment / robotic soldiers is a worrying prospect. While I'm generally all for preventing soldier injury and death I like the fact that much of our military would have the good sense to defect if our govt wanted to try some serious bs.
 
  • Like
Reactions: Liver
like this
Im skeptical that an ai won't fail terribly somehow. The deeper the layers of strategy, the harder it is for an ai to keep up.

Is an ai useful? Hell yes. But you have to be able to compensate for its weaknesses.
 
do pilots still dogfight much though? I thought it was all stand-off aa missiles.

Yeah basically its who has got the best radar and guidance systems. The platform it sits on, not so important. They only make the stuff look sexy for the recruiters.
 
do pilots still dogfight much though? I thought it was all stand-off aa missiles.

I thought I had read somewhere that most (and I really mean all) of the active pilots in service right now had not experienced a real life dog fight, only those in simulation. When was the last time we even fought someone with a air force?? Gulf War possibly? I don't think we ever had real air to air combat with the Russians, so now your going back to Vietnam.

When the B2 bomber first came out, it used to have an escort of F-22's. But they stopped doing that once they realized there was no threat to it.
 
Give me two AI with two virtual jets with the "Exact Same" specifications and loadout. In other words no AI cheating like they usually do. And pit me against them and I'll win more often than not. The problem with using a Real pilot and the AI is they are bound to whatever rules they are trained/programmed into believing. In a virtual world there aren't any physics to worry about and as such strategies that wouldn't work in reality work. A real pilot is unlikely to succeed due to years of ingrained training. The AI isn't going to win because of program limitations. A gamer on the other hand has none of those restrictions and if good enough is going to beat the other two every time.
 
Give me two AI with two virtual jets with the "Exact Same" specifications and loadout. In other words no AI cheating like they usually do. And pit me against them and I'll win more often than not. The problem with using a Real pilot and the AI is they are bound to whatever rules they are trained/programmed into believing. In a virtual world there aren't any physics to worry about and as such strategies that wouldn't work in reality work. A real pilot is unlikely to succeed due to years of ingrained training. The AI isn't going to win because of program limitations. A gamer on the other hand has none of those restrictions and if good enough is going to beat the other two every time.
AI's in games have perfect timing and aiming. Programmers have to go out of their ways to create flaws in the AIs for humans to take advantage of. Also, I'm sure the programmers have taken the actual physics of flight into consideration for their simulations.
 
Give me two AI with two virtual jets with the "Exact Same" specifications and loadout. In other words no AI cheating like they usually do. And pit me against them and I'll win more often than not. The problem with using a Real pilot and the AI is they are bound to whatever rules they are trained/programmed into believing. In a virtual world there aren't any physics to worry about and as such strategies that wouldn't work in reality work. A real pilot is unlikely to succeed due to years of ingrained training. The AI isn't going to win because of program limitations. A gamer on the other hand has none of those restrictions and if good enough is going to beat the other two every time.

Yeaahhh....I'm guessing the the AI routines created for an advanced military fighter project are going to be a bit more robust than hard on Air Combat.
 
AI's in games have perfect timing and aiming. Programmers have to go out of their ways to create flaws in the AIs for humans to take advantage of. Also, I'm sure the programmers have taken the actual physics of flight into consideration for their simulations.

Perfect timing and aiming don't matter one bit if they don't know how to react to a situation. AI at least at this point cannot think outside of the box. This means it knows how to handle anything expected. However it cannot deal with someone who is operating outside the constraints of that. No AI is perfect or anywhere close in a virtual environment yet.

Now bear in mind I'm talking virtual only right now. Obviously aircraft constrained by real physics and without the ability to "glitch" the AI, the AI would win. That is obviously the point of this. I'm just pointing out that a top end gamer would easily outperform a real pilot in a simulation only setting because they are not restricted by the thought of "can my aircraft handle this without crashing".
 
Perfect timing and aiming don't matter one bit if they don't know how to react to a situation. AI at least at this point cannot think outside of the box. This means it knows how to handle anything expected. However it cannot deal with someone who is operating outside the constraints of that. No AI is perfect or anywhere close in a virtual environment yet.
You're right as far as coming up with the only real strategy to fight these robots goes. We will need you in the upcoming battles when these planes turn on us.
 
Well Chuck Yeager said "It's The Man Not The Machine" in his video game

Chuck%20Yeagers%20Air%20Combat_1.png



Humans can't be hacked like drones already have.

A human can decide not to drop ordinance on a target based on moral/ethical challenges.

That being said I'm all for a Protoss Carrier pulling in.



Used to love me some yeager! F-86, F-4 or mig-21 (think it was the 21) against a bunch of prop bombers. Mmmmmm, turkey shoot.

Another good reference is Day of the Cheetah. AI/thought control plane against more traditional weapons. Loved that book back in High School. Pretty cool stuff in concept would be interesting to see how close it plays out in reality.

I think the drone/swarm concept is where this is headed. Cheap easily replaceable hardware is the way to go.
 
Well Chuck Yeager said "It's The Man Not The Machine" in his video game

Chuck%20Yeagers%20Air%20Combat_1.png

Machines have been the deciding factor in war since World War II, perhaps WWI. Unfortunately during WWII ours sucked until towards the end, but fortunately we could produce more than they could blow up. So in that case it was our factories that won the war.

AI just removes all doubt from what now is almost a 100 year romantic delusion about it being the man.
 
do pilots still dogfight much though? I thought it was all stand-off aa missiles.

In real world engagements, no, since in most of the current theaters air superiority is already given to one side. All American fighter pilots go through air combat maneuvering and dissimilar air combat training to learn how to dogfight or engage at a short range.

Real world use of beyond visual range weapons found the have a high failure rate. Newer versions of the AIM-120/AIM-7 have made BVR engagements more successful but there is still a major issue, positive identification of the enemy. Separating friend vs foe, especially with so many of our NATO partners flying Soviet block fighters, has been a challenge which is why the AWACS platforms keep getting upgraded. Last thing we want is downing a NATO aircraft.

So in many cases visual identification is used resulting in short range engagements. Thus crews going to Nellis/Fallon or Key West get to dogfight instructors flying US and Soviet block aircraft to keep their skills up. There are a handful of Mig 29's, among other aircraft (L39, TA-4, F5) being used by private aggressor companies that provide air combat training to the military.

Dogfighting/short range engagement will always be a skill pilots are taught. Plus when you run out of missiles all you have left is a gunfight. Better know how to use it.
 
Back
Top