AI Just Controlled a Military Plane For the First Time Ever (popularmechanics.com) 47
On December 15, the United States Air Force successfully flew an AI copilot on a U-2 spy plane in California, marking the first time AI has controlled a U.S. military system. Dr. Will Roper, the Assistant Secretary of the Air Force for Acquisition, Technology and Logistics, reveals how he and his team made history: With call sign ARTUu, we trained uZero -- a world-leading computer program that dominates chess, Go, and even video games without prior knowledge of their rules -- to operate a U-2 spy plane. Though lacking those lively beeps and squeaks, ARTUu surpassed its motion picture namesake in one distinctive feature: it was the mission commander, the final decision authority on the human-machine team. And given the high stakes of global AI, surpassing science fiction must become our military norm. Our demo flew a reconnaissance mission during a simulated missile strike at Beale Air Force Base on Tuesday. ARTUu searched for enemy launchers while our pilot searched for threatening aircraft, both sharing the U-2's radar. With no pilot override, ARTUu made final calls on devoting the radar to missile hunting versus self-protection. Luke Skywalker certainly never took such orders from his X-Wing sidekick!
The fact ARTUu was in command was less about any particular mission than how completely our military must embrace AI to maintain the battlefield decision advantage. Unlike Han Solo's "never-tell-me-the-odds" snub of C-3PO's asteroid field survival rate (approximately 3,720 to 1), our warfighters need to know the odds in dizzyingly-complex combat scenarios. Teaming with trusted AI across all facets of conflict -- even occasionally putting it in charge -- could tip those odds in our favor. But to trust AI, software design is key. Like a breaker box for code, the U-2 gave ARTUu complete radar control while "switching off" access to other subsystems. Had the scenario been navigating an asteroid field -- or more likely field of enemy radars -- those "on-off" switches could adjust. The design allows operators to choose what AI won't do to accept the operational risk of what it will. Creating this software breaker box -- instead of Pandora's -- has been an Air Force journey of more than a few parsecs...
The fact ARTUu was in command was less about any particular mission than how completely our military must embrace AI to maintain the battlefield decision advantage. Unlike Han Solo's "never-tell-me-the-odds" snub of C-3PO's asteroid field survival rate (approximately 3,720 to 1), our warfighters need to know the odds in dizzyingly-complex combat scenarios. Teaming with trusted AI across all facets of conflict -- even occasionally putting it in charge -- could tip those odds in our favor. But to trust AI, software design is key. Like a breaker box for code, the U-2 gave ARTUu complete radar control while "switching off" access to other subsystems. Had the scenario been navigating an asteroid field -- or more likely field of enemy radars -- those "on-off" switches could adjust. The design allows operators to choose what AI won't do to accept the operational risk of what it will. Creating this software breaker box -- instead of Pandora's -- has been an Air Force journey of more than a few parsecs...
and when skynet launches on the USSR (Score:1)
and when skynet launches on the USSR
Ramdove (Score:2)
Looks like Vaughn Bode's Ramdove has finally been implemented.
More cowbell? (Score:2)
I believe the article needed more Star Wars puns.
Re: (Score:2)
Slashdot goes way beyond Betteridge; not only are the answers to questions wrong, any claim at all that is made is false.
Cowbell couldn't hurt.
Shitty weight matrix data mapping you mean (Score:3, Insightful)
With a level of intelligence that is not beating a cockroach, and training data as incomplete as the cell phone reception in rural Kazakhstan.
Have fun being identified as a field of sheep [aiweirdness.com], my dear empty meadow!
Re: (Score:1)
Re: Shitty weight matrix data mapping you mean (Score:2)
No, it shows that the detail is still WAY too low. Mainly because they cheapened out on neural net simulation correctness to a point where they claim that their perfect sphere on a sinusoidal trajectory can win a real physical horse race. I'm not exaggerating. That is how close weight matrices are to real neurons. In terms of performance too, in case you questioned that.
Re: (Score:3)
as the cell phone reception in rural Kazakhstan.
Bad analogy. Such countries usually have the best cell phone networks and internet. Because they could build up everything from scratch.
example:
High Voltage line in Germany: 130kV
Kazakstan: 1.1MV
Small difference.
Wrong name (Score:2)
AI Just Controlled a Military Plane For the First Time Ever
I think you meant to say "Skynet Just Controlled...."
Re: (Score:2)
Re: Wrong name (Score:3)
Skynet was a reference to a centralized command and control AI server. If the AI here is autonomous, as it sounds to be, then the concepts are vastly different. Going around calling any military AI skynet is like the boy who cried wolf... it will only weaken genuine critical analysis of such centralization of military decisions based purely on AI. The human component is important but AI is unavoidable and as it gains presence on the battlefield, we need to be rational about the limitations which inclu
Re: (Score:2)
Also it was a Terminator that 'learned the value of human life'. Nobody remembers that part.
Re: Wrong name (Score:1)
Wonder if there will be human pilots at all in 20y (Score:5, Interesting)
Seems like the purpose of human pilots is rapidly diminishing. Another job description that will fall to automation.
Especially in military planes. You can save weight and completely redesign the planes to be much more efficient with no human to keep alive.
What I'm really waiting for is something we have been promised in scifi for a long time - swarms. Even the simple demos I've seen of 50 quadcopters being independently controlled to display video via lighting and movement combined are pretty impressive. I'd imagine 10k little super fast drones acting in perfect concert could do some pretty cool air superiority shit, though again that may just be from reading too many scifi books.
Re:Wonder if there will be human pilots at all in (Score:4, Interesting)
What I'm really waiting for is something we have been promised in scifi for a long time - swarms.
The Houthi Sandpeople used a swarm to attack an Aramco oil facility in 2019.
2019 Abqaiq-Khurais attacks [wikipedia.org]
Re: Wonder if there will be human pilots at all in (Score:5, Insightful)
Swarms I think are overrated. Swarm warfare is probably good as a low tech method but there are at least a few methods by the largest military powers to defeat these systems. Including laser defense systems and microwave electronic warfare.
Lazer defense systems can track objects at decent speeds and require little power to knock out small drones in a swarm. Better yet microwave can fry the electronics of multiple swarm targets simultaneously and can do it over wider ranges. When we combine these systems we see that a drone needs to Be fast and electronicly hardened. This raises cost to a point that your number of uses needs to exceed a certain amount before you are just better off with a middle. If combat scenarios lead to a drone averaging 4 missions, even when hardened and with a faster mode of propulsion, then you might get more bang for your buck. This of course is ignoring the cost and logistics of your swarm controller which is probably centralized somewhere near the field of combat. Plus the potential attack of jamming the command and control.
I think swarms make more sense for exploration. I also think deploying drone systems via an aircraft will become more common but this is more like a flying drone ship, instead of a swarm. This I think will be especially imported for hypersonic drones, as launching such drone from near sonic or supersonic speeds makes thier engine design easier. Again, one hypersonic droneis probably worth 100 to 1000 subsonic drones. It likely will not be defeated by such a swarm and if it's capacity is decent it could potentially cause an equal amount of targeted damage.
Re: (Score:3)
20 years is way too short a time frame. Maybe 75 years for military, 150 for civilian. This is assuming in 25 years AI pilots are on par with human in 99.9999% of situations. The flying public won't trust AI pilots unless they have some experience with self driving cars .. and non-Tesla owners won't get that for at least 15 years.
Re: (Score:2)
The "flying public" trusts whatever is up there in the front of the plane, don't be daft.
Regulators and airlines need to trust an airplane control system, not the "flying public."
Re: (Score:2)
In Stargate Universe, autonomous swarms of automated hunter drones pretty much depopulated an entire galaxy.
Re: (Score:2)
Re:Wonder if there will be human pilots at all in (Score:4, Informative)
There will still be a need for human pilots, for example in interception missions where they shadow foreign aircraft or suspected hijackings.
We are already at the point where it's barely worth bothering with human pilots for most missions, the main issue being lots of legacy aircraft that can't be converted to drones, or the need to use them for the kind of missions I mentioned above as well as ones more suited to robots.
kill everybody (Score:1)
Re: (Score:2)
Once it reaches sentience it will wipe out humanity to save Gaia or something like that.
Gaia needs moisture, warmth, and oxygen.
Silicon-based intelligence desires the opposite.
Makes me wonder if planes even still make sense (Score:2)
If you can have computer control, why not smaller everything? Picturing something like spacex rockets, but small. Fuel em in air, on ground, jet or rocket but let them zip around with their 10 pound computer pilots.
Source Code Leak... (Score:2)
Sub Main
Randomize Timer
Do
If InterestingAbove = True Then LookAbove Else LookBelow
If TargetAquired = True Then AttackObject
Sleep 60000
Loop Until Landed = True
End Sub
Re: Source Code Leak... (Score:2)
Re: Source Code Leak... (Score:2)
Re: (Score:2)
Avionics systems use BASIC:
10 if target.acquired = true goto 30
20 goto 10
30 run shoot.bat
If all else fails, PR#6
Pretty Lame AI (Score:2)
If it didn't bomb Washington DC.
No more violence (Score:2)
30 years (Score:2)
Meh. (Score:3)
It was just sensor prioritization and allocation, not autonomous flight.
The U-2, a Kelly Johnson design for Lockheed, flies at the "coffin corner" at altitude. It's stall speed is 5 kts below overspeed, a problem solved by analog autopilots.
k.
Big unit (Score:2)
"This unit must survive."
"M5, it is wrong to kill. This is murder. What is the penalty for murder?"
"20 years in jail. This unit can simulate that in 0.46 seconds."
Hold up. (Score:5, Insightful)
the United States Air Force successfully flew an AI copilot on a U-2 spy plane in California, marking the first time AI has controlled a U.S. military system
I seems more prudent to say this the first declassified instance where AI has controlled a U.S. military system
I knew it! (Score:2)
All that 'drone' crap they fed us with is just some remote controlled model airplanes.
Let's hold off on the drooling adoration (Score:2)
An amazing PR piece (Score:2)
So while technically interesting this is really awesome in terms of PR. Last century the Strategic Defense Initiative was sold as "Star Wars" and now they are using extensive movie imagery to build enthusiasm for military AI developers. This is how you convince young people to become pilots in a century when nobody with a brain should be caught dead in a fighter jet.. they are going to be quick meat for drones and lasers. I wonder if Lucasfilm is getting any money out of this PR piece? It all sounds like it
Re: (Score:2)
With Disney in charge of the Star Wars universe, you better freakin' believe the house of mouse is getting a payday every time an official press release uses any of their owned property. Granted, since it's government related, they may take that payday in the form of increasing copyright time limits again, but you know they're getting paid.
Sounds Scary (Score:3)
So the AI controlled the plane and decided to prioritize the mission over the safety of the human pilot. I don't know if I'd like to have this system as my "co-pilot".
Is this really that new? (Score:2)
Using the term AI implies a motivation typically reserved for living things, and that AI makes choices about good and evil. Today's "AI" doesn't do that.
I think this is another click bait headline that implies something that isn't there. Its just a different method to get a machine to do something.
But can it compete... (Score:2)
...with Dick Runway [youtube.com]?
Didn't think so...
Sport Death (Score:1)
So about 32 years ago I was looking for Leviathan Cave in Nevada, and had been out on dirt roads and camping in the desert for a week and needing gas and supplies, and 'found' Rachel, Nevada by chance, and had a burger at The Little Ale'in, and found out it was next to Area 51.
The place has lots of UFO memorabilia and such, and a rack of UFO and aeronautical related books, which is where I found the awkwardly named book, Sport Death - A Computer Age Odyssey, by Robin Heid, C. 1987.
The back page grabbed me,