News for hackers. Stuff that matters.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by akmarinov on Friday August 21, 2020 @07:36AM (#24233988)
Isn't it unfair to the human pilot, as the AI can pull G-Force maneuvers that'll otherwise kill the human pilot?
Re:AI slays top F-16 pilot in DARPA dogfight simulation by parsimo2010 on Friday August 21, 2020 @07:45AM (#24234036)
Warfare is only sometimes about good sportsmanship. During the part where you’re trying to kill each other, it’s totally fair game to pull more G if you can. It’s also fair to use a better jet with better missiles of you have them.

Once (if) the other guy has ejected and is under canopy, then it’s important to be a good sport again- no shooting at him while he’s defenseless, and if you capture him then you have to treat him with dignity as a POW.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by arethuza on Friday August 21, 2020 @07:54AM (#24234103)
And if he loses an artifical leg during the ejection process you may let a replacement be parachuted in:

https://www.rafbf.org/news-and-blogs/operation-leg-%E2%80%93...

Re:AI slays top F-16 pilot in DARPA dogfight simulation by ufmace on Friday August 21, 2020 @10:00AM (#24234946)
That's really cool. But I gotta wonder - why didn't the Germans just give the guy a new artificial leg instead of letting the Brits parachute drop one in? Seems like it'd be easier for everyone.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by philwelch on Friday August 21, 2020 @01:31PM (#24237002)
Those are usually custom fit to the individual and the Germans were probably happy not to take the effort of repeating all that work.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by xiaopanga on Friday August 21, 2020 @01:21PM (#24236905)
There may be specific fit issues for a new artificial leg. Most prosthesis require quite a bit of fit adjustment.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by WrtCdEvrydy on Friday August 21, 2020 @08:50AM (#24234436)
See, there are weird war stories that if we turned into a movie, people would think it's Hollywood embellishing.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by philwelch on Friday August 21, 2020 @01:46PM (#24237162)
This is true. A Bridge Too Far both includes incidents that actually happened but were deemed unrealistic by critics, as well as omits certain details specifically because they would be assumed to be unrealistic.

The one movie I would love to see is one about the battle for Castle Itter toward the end of the war, where Americans, German soldiers who had defected to the resistance, a deserting SS officer, and a group of prisoners including two former French prime ministers, two French generals, Charles de Gaulle’s sister, and a pre-war French tennis champion all teamed up against an SS force attempting to recapture the castle and the VIP prisoners within. Which sounds like a completely absurd movie but it actually happened.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by arethuza on Friday August 21, 2020 @09:06AM (#24234547)
There is a movie about Bader: Reach for the Sky:

https://en.wikipedia.org/wiki/Reach_for_the_Sky

I can't remember if it includes the bit about parachuting in a replacement artificial leg... I can't imagine there was much desire to portray Hermann Göring in a positive way in the mid 1950s!

Re:AI slays top F-16 pilot in DARPA dogfight simulation by gadders on Friday August 21, 2020 @09:05AM (#24234541)
They did make it into a movie, but not sure if it covers the parachuted legs incident: https://en.wikipedia.org/wiki/Reach_for_the_Sky
Re:AI slays top F-16 pilot in DARPA dogfight simulation by derivagral on Friday August 21, 2020 @10:42AM (#24235287)
Is that second part still a thing? I know ww2 got hairy, but as a regular civilian I don't know a ton of context about whether this was isolated or the start of a trend.

https://en.wikipedia.org/wiki/Laconia_incident

Re:AI slays top F-16 pilot in DARPA dogfight simulation by parsimo2010 on Friday August 21, 2020 @11:22AM (#24235631)
Some thoughts:

A) The "law of armed conflict" isn't a formal law, it's a collection of treaties recognized by most countries, and each country can have a slightly different interpretation. But most of the written stuff only goes back to the Geneva Conventions, and before that it was just tradition that kept everyone in line. Basically, if your side didn't follow tradition then there was a chance that you wouldn't be treated well if you got captured. So people mostly behaved themselves.

B) All (civilized) forces agree on the idea that it is unethical to attack non-combatants. There are grey areas, but everyone agrees that a pilot is no longer a combatant after his plane is disabled. If they land in your territory you can capture them and hold them prisoner, but you can't kill them, even if they were dropping bombs on you a few hours earlier. Similarly, if you sink a ship you are not supposed to kill the survivors.

C) Accidents happen and communication between enemies isn't totally reliable. Although you aren't supposed to kill people engaged in life-saving or rescue operations because they are non-combatants, what are you going to think if you see an enemy ship or plane approaching your defenseless friends? Sometimes people do bad stuff intentionally though, and hope that the lawyers can make a decent case if they get caught. Considering that evidence often gets destroyed and that dead men can't give testimony, they often get away with it.

D) A non-combatant can lose their protected status by engaging in combat. If a downed pilot pulls out his sidearm and starts shooting while you're trying to capture him, then you can shoot back. You can't shoot a clearly marked medic trying to save someone's life, but if that medic picks up a gun then you can shoot them. Paratroopers are practically defenseless while they are under canopy but they are considered combatants, so you can shoot them in the air.

E) AI probably won't be granted personhood until after they demonstrate sentience, so computer pilots probably won't get any of the courtesies discussed above.

Source: this was made very clear to to me at SERE school and is made clear to everyone at any SERE school regardless of branch. The exact phrase my instructor used was, "don't start bopping people on the head if you get caught behind enemy lines, but if it looks like the enemy isn't playing fair then you might as well get to bopping."

Re:AI slays top F-16 pilot in DARPA dogfight simulation by ourmandave on Friday August 21, 2020 @10:12AM (#24235022)
Once (if) the other guy has ejected and is under canopy, then it’s important to be a good sport again- no shooting at him while he’s defenseless, and if you capture him then you have to treat him with dignity as a POW.

But what if the simulation is a Spielberg movie and if you let the AI go, later on he'll kill Tom Hanks?

It's like the Trolley Problem but weighted by how much you liked Joe Versus the Volcano.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by jvzr on Friday August 21, 2020 @07:54AM (#24234099)
Beer missiles are the best!
Re:AI slays top F-16 pilot in DARPA dogfight simulation by parsimo2010 on Friday August 21, 2020 @08:16AM (#24234221)
Thanks for the catch. I edited my typo so now everyone will be super confused when they read your comment.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by ethagknight on Friday August 21, 2020 @07:40AM (#24234005)
From the article, the pilot was sitting in a simulator. It’s also unfair to the pilot that he or she was not given all appropriate measures to avoid a gunslinger dogfight using long range weapons, teamwork, and stealthiness, but that wasn’t the point of the simulation.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by throwaway0a5e on Friday August 21, 2020 @07:46AM (#24234043)
Exactly. If two fighters are in a dogfight in 2020 it's because an entire network of systems has failed.

This exercise simply isn't reflective of modern tactics.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by willvarfar on Friday August 21, 2020 @08:39AM (#24234372)
Open warfare is rare. Most fighters patrolling near potential adversaries have complicated rules of engagement that give up their bvr advantages.

So you’re in your fancy fighter, patrolling a no fly zone, and you’re vectored to do a viz id on something that turns out to be a helicopter. Now you’re sweating it in case what could be an aid flight is actually going to fire an rpg at you as you pass, or if the whole thing is bait to bring you over aa or an ambush by some outclassed fighter that’s gonna magically arrive on your six...

Re:AI slays top F-16 pilot in DARPA dogfight simulation by throwaway0a5e on Friday August 21, 2020 @08:46AM (#24234414)
When war starts someone always gets the jump on someone. Look at the strategic picture. That's one airframe. It's acceptable losses in exchange for not shooting at things you can't see prior to open hostilities.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by thret on Friday August 21, 2020 @08:02AM (#24234151)
Then why do we train for it constantly in coalition exercises?
Re:AI slays top F-16 pilot in DARPA dogfight simulation by afarrell on Friday August 21, 2020 @08:06AM (#24234172)
Sometimes entire networks of systems fail.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by throwaway0a5e on Friday August 21, 2020 @08:14AM (#24234214)
Because having the training is preferable to nothing in the same way that having poorly equipped, poorly trained fresh recruits manning your trenches is better than nothing.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by KineticLensman on Friday August 21, 2020 @08:40AM (#24234378)
Yes. He was also wearing an operationally unrealistic VR helmet. So, he wasn't flying a real platform for which he would have had good muscle memory for real flight controls.

This is actually a big deal. Very small differences between a sim and the real platform could be showstoppers for pilots (e.g. the contrived example of the 'fire' and 'radio' buttons being swapped in the sims vs. the real planes). Sims that are used for real combat training have to be exceptionally high-fidelity in all aspects that affect a given training task.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by dpflan on Friday August 21, 2020 @08:26AM (#24234283)
Indeed. Reminds me of the Machine People and Hesperus in House of Suns discussing how they can push the limits of the spaceships that humans physically could not survive.

- House of Suns: https://en.wikipedia.org/wiki/House_of_Suns

Re:AI slays top F-16 pilot in DARPA dogfight simulation by rangibaby on Friday August 21, 2020 @07:41AM (#24234008)
Wouldn’t it work that way IRL too?
Re:AI slays top F-16 pilot in DARPA dogfight simulation by TuringNYC on Friday August 21, 2020 @08:26AM (#24234281)
I saw the page and assumed the AI had boundaries on maneuvers that reflected the real-world boundaries on those aircraft.

Otherwise, the AI can easily win -- fly at light speed to behind the opponent, fight, fly at light speed away. The only thing to do with a simulation like this would be to put boundaries on the simulator.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by tiku on Friday August 21, 2020 @08:28AM (#24234299)
The boundaries are on the human body, not on the AI. So it's more about speed and certain turns etc.
Re:AI slays top F-16 pilot in DARPA dogfight simulation by TuringNYC on Friday August 21, 2020 @08:46AM (#24234413)
There are definitely also boundaries on any aircraft. There are speed limits, acceleration limits, etc.

The AI is just controlling the aircraft, so any simulation where the AI isnt bound by the aircraft's limitations would not be a good setup.

This isn't different from simulation based training of autonomous vehicles. Imagine if in an AV simulation, the AV controller could opt to accelerate 0-60 in 1 second, or if it could suddenly go 300mph to skip an up coming collision, etc.

Re:AI slays top F-16 pilot in DARPA dogfight simulation by jmnicolas on Friday August 21, 2020 @08:55AM (#24234471)
All is fair in love and war.