The SAAF Forum

Apparently the Europeans are trying to develop a combat drone: http://www.brecorder.com/top-news/1-fro ... light.html

It's interesting to note that even Spain and Greece, two countries with deep seated economic problems, are participating in the project together with France, who seems to be the leading partner. Also interesting is the absence of Germany and England.

Always wondered about pilotless fighter planes. Can't think how you can successfully replace a human pilot. Reminds me of the movie Stealth with Jamie Fox and that other fox called Jessica Biel! . One just cannot program a machine to apply common sense and conviction.

_________________
Gil en spring!!!!!

Denels latest UAV's can take off, do a mission and return completely automatically, so it is possible.

_________________
The first thing you do when you find yourself in a hole is you stop digging!

http://www.facebook.com/#!/group.php?gi ... 2681351417

The Book Worm: http://novelbooks.weebly.com/index.html

Ja, but can they make life and death decisions? Can they decide when to fire and when not to without human interference? I understand currently drones are oprated much like R/C planes and that most cricial decision making still require human in put, but isn't it better to have the human there? And how will drones react to engagement by enemy aircraft?

I can surely see the benefits of recon drones, target painting drones and the like, but fighter drones? I don't think so.

_________________
Gil en spring!!!!!

Is it really that much different from AI opponents in fighter games? As far as current tech is concerned, I think we'll be able to field a drone that can engage enemy fighters on its own. Everything else, especially things where the target cannot be easily identified, or places where intuition and judgement is critical (close air support, intercepting wayward aircraft), would require a human to make decisions for it.

However, before we get even that far, we'll need to have that debate about giving AI the capability to shoot to kill of their own accord. For a long time I think, we'll still have a pilot pulling the trigger.

My point exactly! But even in fighter games all enemy aircraft follow a series of pre-planned programmed routes routines and movements. I've been playing video games, pc. PS2, PS3, XBox 360, Wii, Kinect, for many years and I can tell you ANY adversary can be beaten once you know his next move, and if you play the game long enough you'll figure it out.

Even AI has limited capability as far as judgement etc. You cannot program intuition, as you mentioned, but also no other abstract feeling and perceptions unique to Homo Sapiens Sapiens.

_________________
Gil en spring!!!!!

You guys might want to do a little research on Asimov's Laws of Robotics.

Hmmm, some midnite reading for me them. Let's check it out and then get back to this topic.

_________________
Gil en spring!!!!!

We send pilots to flight school to do what exactly? Apart from giving them experience in how the aircraft handles, and its limitations, I'm fairly sure we teach them how to respond particular situations or emergencies.

You'll also find that AI is advancing quickly, and also note that game AI is generally optimised to run quickly on home computers. Military-grade AI on military-grade hardware will run very differently.



This is true of all warfare, and life in general.



A weakness of a game that doesn't evolve, but I can assure you that fighter AI would evolve. Another issue is that you have to play the game long enough. How many times do you think you can be shot down before it kills you, on average? You won't be able to identify every potential move of an AI in a single aerial engagement, and the required time gets longer and longer as the system gets more complex. Remember that AI exists that can adapt to a player's strategy (and this was RTS AI from a few years back, god knows what they're up to now), so there's a lot that an AI can do to stop you reading its moves.

Tell me, what's the latest high-quality air combat game you've played?


And no, you can't program an AI to be truly paranoid, but there are a lot of things you can program into it, as long as the person programming thinks of it, which is why pilots are guaranteed a role there.

There's other things to consider, like a pilot having to have a radar return parsed and converted into something human-readable, while an AI can use the raw data. There's no delay between information returned and information noticed, reaction time can be significantly faster. There's no G-force limits on the pilot, physical fatigue etc. Only the limits of the aircraft.

You do become very vulnerable to EMPs and jamming (depending on your setup) I guess, but there's shielding and such.

As mentioned, I believe it's the ethics, not the technology, that pose the biggest hurdle for fully automated killing machines.


This amused me though:

Let me read up on all of this a bit before I respond, guys. I made a statement and even though M4rek made some very valid AND interesting remarks I owe to us all to explore my point of view in an attempt to justify it.

In the meantime: Carry on, doctor!

_________________
Gil en spring!!!!!

A few links to get you started:
http://www.economist.com/node/21556234
http://www.usnwc.edu/getattachment/851e ... view-Essay)---
http://ethics.calpoly.edu/ONR_report.pdf
http://www.stlr.org/html/volume12/marchant.pdf

Roger, we're talking Fighter Combat Drones here, right? Asimov's first law broken already, so I strongly doubt if Asimov's laws is relevant here.

A new set of rules need to be defined to address all the unknowns and possibilities of automated unmanned vehicles (robots?) capable of killing humans (with bombs, missiles, guns etc.)

I'm still working through all the other reading you supplied (thanks by the way) and to be quite honest I cannot see fully automated unmanned fighter capable vehicles any time soon.

Yes we use robots, remotely controlled by humans, to perform diffcult and often dangerous activities, so that we can prevent human losses, such as those used to disarm bombs, surveillance drones, etc, etc. as already mentioned.

But what we are talking about now is a fighter ready unmmanned automated aerial vehicle capable of killing another person without the intervention of a second human.

Much like setting your highly trained protection dog loose on another human being, and then turning your back to see if the dog can make his own decisions, after you have given him an order. Can you expect your dog to NOT harm anyone else when trying to subdue your perp? Can your dog make rational moral decisions?

So now we expect a piece of metal and electronics, that cannot love, hate, dislike or reason, to make such moral rational abstract decisions.

I believe unmanned fighters, remotely controlled by humans from a strategic position is quite possible, but presents its own set of challenges, such as reaction time, and sensory input of the remote pilot. Would the drone get that sixth sense many fighter pilots report on before an imminent engagement? But that is a discussion for another thread.

Automated Unmanned fighter capable vehicles IMHO opinion is irresponsible. Its one thing sending an automated drone to take photos of enemy positions, record enemy force counts and even to paint targets, but its a totally different story to tell this same drone to go into "hot" environments with a view to defend itself and in effect kill another human being.

The Geneva Convention places a premium on the rights of innocents as well as the rights of those combatants in contact situations: when I stop firing at you and "put up my hands" (figuratively speaking) and pose you no more threat you cannot kill me. Would a drone be able to distinguish this scenario?

Look, I agree I have no idea what current developments in this field entail and what the near future drones/robots will be capable of, Balerit, Roger and so forth seems to be much better placed to enlighten us.

I just cannot comprehend a machine making life and death decisions without the intervention of a human. I don't think its morally ethical and I don't think it should even be considered. Much like it would be immoral to allow cloning of people ...

However, having said that, I am still reading and enlightening myself to the world of Robotics and UAV's and such, so I might still change my mind ...

_________________
Gil en spring!!!!!

Sorry M4rek, for some odd reason I did not see this comment from you:



We're actually agreeing here. I too believe the technology would probably be able to advance that far, or very close to that far, but on the ethical front we need a lot ,more discussion and development.

_________________
Gil en spring!!!!!

It is however an important starting point; most importantly perhaps in showing just how difficult it is to give a logic engine even the simplest instructions without it killing someone. I've exaggerated that of course, but at the end of the day, you have to be careful what you tell an automated machine what to do, especially in context with the logic that has been programmed into it.




Which is the fun part =/



We should be wary of a difference between general laws of robotics, general laws of combat robotics, ground-based/ground-attack robots, and air-only robots.

A drone to enforce a no-fly zone is much easier to deal with than all of combat, because as long as it has working radar, GPS and weapons, the result of giving it a (true) no-fly zone to enforce will always be the same. If it flies, it dies. It doesn't really need to make decisions.

Problems start arising mostly when you have it making decisions, especially about whether or not to engage, and what constitutes a target. There's been enough cases where humans have screwed this up, so a robot can't be expected to fare that much better in all circumstances. In some, results may be better, as when an airliner overlapped with a parked military jet and a ship's radar controller marked one as the other. The airliner flew normally, but the operator got so worked up that his mind tricked him into thinking it was on an attack run against the ship, reported it to the unfortunate fellow in charge who had no choice but to blow it out of the sky. Protocol has changed as a result, as a second person checking the screen could have prevented a significant loss of life, but it is this sort of situation where an automated system would have saved lives, and I am sure there are other similar situations where human judgement can fail.



A dog is trained, and that training defines its decision-making process. A computer does the same, except the decision-making process is programmed directly. As such, it's much easier to predict the outcome of a computer's logic, because you know exactly what it's thinking. With a dog, you're missing that step; you don't know what it's learned from what you taught it.

Its instincts and what life has taught it might be beneficial in some situations, but is unpredictable.



I think it's unfair to say it can't reason. Reasoning is the application of logic to a process, and a machine is nothing more than a logic engine.

As noted above, we don't necessarily expect it to make moral decisions. It all depends on what you program it to do. You might program it with a way to tell friend from foe, shoot at foes, and ask for instructions if it can't tell which it is. Being fired at is a good way to identify foes, and can probably be programmed as a condition.



I'd go with something similar to the mars rover as far as control goes. Because of the time delay between Earth and Mars, a solution was found that gives the drone a set of pre-programmed behaviours, which are activated by command and with parameters from Earth. A fighter jet could work the very same way. Give it the programming and means to destroy a target from the word go, and look after itself in a non-hostile manner until the word go, and then have a human giving the command.

Some things are fairly black and white; when someone locks onto you and then fires a missile, you're well within your rights to shoot them down. A human would do exactly that unless given explicit orders (and even then they might do it anyway), a robot can be programmed to do that, unless given explicit orders, and will always comply with them.

I don't think we'll see robots firing at unidentifiable targets anytime soon, and I'm not sure if it's something we need or want to discuss right now. I don't believe it's something that is truly relevant (to the original post, as relevant as it might be to the current discussion), and could get a thread of its own.

I will touch on it regardless. At the end of the day, to fire on a target, an AI needs to identify it as such. The more difficult that is, the more awkward the conversation about ethics. That's why it is easier to talk about pure-air drones, as opposed to ground-attack or ground-based drones; because telling aircraft apart is much easier than identifying what is and is not a threatening humanoid. There are only so many ways of identifying one, and a lot of them are visual. You can use various other sensors to identify a humanoid, such as heat (though technically still largely a visual identification), but to identify if it's hostile is another matter. The easiest way is to know where a person is and is not supposed to be, such as inside a facility. Automated internal gun systems are perfectly feasible in certain types of area, and in fact one could make one using a handgun you might already own, and an auto-sentry system that can be bought online. Then just shut it down when you need to pass through the area. It's a very basic way of doing it, and results in a "secure corridor", but would be vulnerable to a breach while you're in it. Not very effective, but demonstrates the easiest instruction to give a drone: Kill all, and kill none. The ethics are about as complicated as a CCTV installation or a minefield on private land.

Telling a friendly and hostile human apart becomes more difficult. If CoD is to be believed, an IR strobe can be used to make this task easier for humans (referring to the mission where you get to play gunner on the AC-130). But one thing that a human can do that a machine cannot do (or at least not easily) is decide who is shooting at who, something that a gunner can do. It's easy enough to identify a person that is shooting to be shooting at you and thus hostile, and as a robot you can afford to wait long enough for this to happen. When people are shooting at each other, one cannot at this time expect a robot to intervene on that information alone. A human would struggle just as much. The best call there is to not shoot, unless you can positively identify a foe, without killing anyone you cannot identify as a foe.



Surely an automated system painting a target for another automated system that does the killing is much the same thing as an automated system killing people? Not entirely sure how these systems work, so I could be wrong as to how much human input there is in the process, but I think that painting the target is usually the bit that's done by humans, hitting it is largely automated (guided weapons and the like).

Sending a drone into a hot environment and asking it to defend itself without using pre-emptive action is probably the most ethical way of employing an automatic killing machine, because it has a guarantee that anyone it kills was hostile - they fired first.



In air-to-air warfare, how is this done now? A drone would have this much easier, because one could standardise a "ceasefire" frequency or signal, which could then be used to surrender. Probably far more reliable than a crackly radio in a different language against someone you just pissed off by shooting at them.

In a land-based scenario, this is one of many things where drones will fall short when trying to identify targets. I'm repeating myself I think.



I think you're putting a little too much significance on it. Well-programmed cold and calculating is no better or worse than emotional. Machines inherently cannot hate or be upset (as you stated above); humans can be. One can program a computer to account for casualties inflicted so far, but at the end of the day, your machine is mirroring the morality of the man who programs it. Why is that machine so much scarier than the man?



You make it sound like men have a right to make life and death decisions. The only thing that sets them apart is that man has a conscience. The only problem with that reasoning is... Some don't, and some listen to other emotions first.


And why's that?
More importantly why is it moral and ethical to clone animals as opposed to humans? Is it? I think (from the way you say it) that you say that because it's something you've always heard or thought or been told, but never really thought about.


Sorry if I'm repeating myself a lot, I wrote this over the course of hours so I don't know how it's come out...