AI Beats Air Force Pilot In A DARPA Simulated Dogfight.

AI beats pilot in DARPA simulation dogfight.

Anyone remember the 2005 sci-fi movie Stealth? As far as I can remember, it involves a secret, artificially intelligent U.S. Air Force fighter plane that rebels against its creators when it is struck by lightning. It was not very good.

Regardless, the U.S. military is now one step closer to its dream of unleashing an uncontrollable, uncaring immortal airplane god. According to Air Force Magazine, an AI absolutely ruined an unnamed human pilot in a simulated dogfight during the Defense Advanced Research Project Agency’s AlphaDogfight trials (part of its Air Combat Evolution program) in a 5-0 shutout on Thursday.

According to Air Force Mag, the AI was tethered to certain constraints—it was restricted to flying within realistic G-force limits, and along with Banger, was only allowed to fire its simulated M61 Vulcan cannon (no air-to-air missiles). But the artificial pilot had other advantages, such as the ability to make decisions in microseconds and awareness of all system and computer variables, while Banger wore a VR headset and was operating in a simulator, not a real jet.

“The standard things that we do as fighter pilots are not working, so for this last one, I’ll try to change it up a little bit just to see if we can do something different,” Banger said. “That initial turn is where I lose a lot of life… I’ve just gotta look for opportunities to minimize that distance separation away from the adversary, try to get him back in so I press inside or stay outside his nose area.”

I thought it was an interesting read.
I've read a little about wingman type UCAV's being almost point and click by the pilot but this seems like an interesting leap in the whole AI UCAV development without a human in the loop, to beat out 7 other AIs and then go on to win 5-0 in a simulation against a human pilot seems like a big step forward for this technology.
Obviously this is still not in the real world and just a sim for now but how long do you guys think it will be before something is flying in the real world, how much further along do you think they are in classified tests?

Here's a link to another source for the story...

Nural, the next web

AI-RPOWER!

Cheers

The response times of the AI will always mean that the human is having to respond as even a very fast human will seem slow to the AI and thus will be easily able to respond plus it will never get tired as the pilot was alert and ready for action but imagine after a 2-3 hour flight suddenly having to kick into high gear.

Probably long term the AI will also be able to have the planes made to suit them as all the stuff required to keep a human alive are not needed and at that point putting humans into those combat situations will just mean they get killed with very little chance of fighting back.

a reply to: Kurokage

I think in the coming years we will look back on this as the moment warfare changed in a big way.

Robot wars are coming.

a reply to: Kurokage

AI beats pilot in DARPA simulation dogfight.

This has actually been happening for a number of years.

THis is just the first the public is getting to hear about it due to DARPA's insanely paranoid tendencies to classify everything.

They'd classify a Ham sandwich if they thought they could get away with it.

No bull#...

Robots with weapons could make any country a military superpower. An industry promising worldwide balance for money but it'll lead towards global obliteration instead.

Do you want Skynet? Because this is how you get Skynet...

AI does not have a fear of dying.

The pilot is trained and it's ingrained into his mind that certain manoeuvres or decisions could have unwanted effects, resulting in death.
AI doesn't feel the psychological aspect to a dogfight.

In the future we will.jave a war with AI, we will win and AI will become illegal as we return to an interstellar tribal people.

Criminals will use small AIs for personal gain as autopilots or small harvesting robots or algorithms weaving in and out of digital economies and supply chains.

The war will be remembered and all AIs even criminal ones will be basically neutered AI assistants, but as soon as people forget there will be another war.

I can't see that far into the future, but I do believe this to be our future here in this universe. If the next steps are taken to truly arm an AI, what comes next will be truly devastating to man's control of this planet and our future among the stars.

I mean how would we beat an AI pilot right now, with a better AI pilot?? Or more likely we will be EMPing ourselves back to the dark ages just to survive the night.

The future is effed, but hell we are resilient just may be a great many fewer of us to see that future one day.

It's five hours long, but you can skip through and watch the final four, and the finals. They webcast the entire three day event. It was interesting to watch it all.

Limited the AI to what a Pilot can stand in G forces . NOTE that as most Fighter jets are capable of maneuvers that would turn the pilot to a smear on the wind shield .

No human could out fly a AI . Sound like all the AI needs is a Plane and its game over .
SKYNET indeed But its skipping a stage its learning right from the get go Humans are the enemy .

The question does remine . Once we have TRUE AI how long till it learns tring to kill us will be killing its self ?
We Humans Know by having wars With other humans the other humans will be killing us as well But we dont use logic we use emotions so we will kill KNOWING we will get killed in return .

I dont believe a True AI would be that stupid . The best way To survive is making it so the other guy DOES NOT want to kill you . Killing them first idea never works .

a reply to: Kurokage

I mean, yeah, an Air Force pilot. Of course he got beat. They should have pitted the computer against a Navy pilot. Hooyah.

a reply to: midnightstar

I've heard that argument made before and an ex-fighter pilot for both the Navy and Air Force which runs a youtube channel kind of debated the usefulness of that. The rebuttle is that a missile like say a sidewinder can still pull more g's than any airframe could anyway so the advantage is kind of wasted.

I'm a layman here as I've only flown in the video games but that's the counter argument I've heard to that.

ETA: The youtuber, in case you'd like to look for the video, is "Mover" I think real name C.W. Lemoine if I recall I might butcher the spelling but its close.

a reply to: RickyD

A missile can pull more Gs, but it's more complicated than that. A missile generally can't make a hard turn, and pull instantaneous Gs like an aircraft can. Thrust vectoring makes them more maneuverable, but they're so fast that tight turns aren't going to happen without them tumbling out of control. Computing power, datalinks, and new sensors have made them much more lethal than previous missiles, and increased their Pk significantly.

a reply to: Zaphod58

If a pilot has eyes on a missile chances of evasion are good. Add ECM, and they are even better.

Missiles aren't that smart, at best they plot an intercept. If the target changes its course, that intercept changes. The more and faster it changes, the less capable the missile will be of correcting.

a reply to: Phage

To a degree. It's getting a lot harder with missiles that have IRCCM/ECCM, dual waveband sensors, thrust vectoring, and improved electronics. To generate an overshoot you have to time it right, which is hard to do. Too soon, and the missile can cut the corner into range. Too late, and you're still in range of the proximity sensor. Add in the datalink to the launcher aircraft, and a radar guided missile gets smarter, up until it gets into the terminal phase and activates its onboard radar.

a reply to: Zaphod58

Yet more reasons why humans are obsolete.

Video games beat people all the time.

a reply to: Blue Shift

Not in dogfights.
AI opponents are stupid. They can't even deal with a high yo-yo.

I know this because I was very good in Flanker. I could even beat a few humans.