Slashdot Mirror
UN to Debate Use of Fully Autonomous Weapons, New Report Released
concertina226 (2447056) writes "The United Nations will debate the use of killer robots for the first time at the UN Convention on Certain Conventional Weapons (CCW) this week, but human rights activists are calling for the robots to be banned. Human Rights Watch and Harvard Law School's International Human Rights Clinic have published a new report entitled 'Shaking the Foundations: The Human Rights Implications of Killer Robots', which calls for killer robots to be banned to prevent a potential arms race between countries. Killer robots, or fully autonomous weapons, do not yet exist but would be the next step after remote-controlled armed drones used by the US military today. Fully autonomous weapons would have the ability to identify and fire on targets without human intervention, putting compliance with international humanitarian laws in doubt. Among the problems with killer robots highlighted in the report is the risk of criminal liability for a military officer, programmer or weapons manufacturer who created or used an autonomous weapon with intent to kill. If a robot killed arbitrarily, it would be difficult to hold anyone accountable."
Don't mines qualify as "autonomous weapons"?
Set your phasers on "funky"!
Bans will not only not prevent them being developed, probably even by a technologically advanced State that is a signatory to the treaty, but it will also not prevent them being used by rogue or puppet states who don't care about bans, or who use them at the behest of a signatory state that is just using them to do their dirty work.
c development!"
When I read things like this I wonder how these people even function in daily life without eating pebbles and glue sandwiches. The fact that the law is not currently equipped to assign guilt in the case of the malfunction of an autonomous robot is not a good enough incentive to stop scientific progress. First of all, robots can't kill arbitrarily, they can only kill who their programming specifies they should kill, even if that programming encounters a bug or operates in a manner unforeseen by its programmers. Arbitrarily would be without reference to a standard, randomly, like an earthquake or lightning. Second, banning killer robots will not prevent an arms race. It will simply hamper the combat effectiveness of the side who holds itself to the treaty. Third, it would be much more effective if the money spent on ethicists worrying about how scary science is to them went to the scientists instead, so that it could go into development and research of the very thing the ethicists are so afraid of, to make it better understood and less scary.
Auto-targeting weapons are only a matter of time. If a college student can make a gun that spits out paintballs with high accuracy, then the best and brightest likely have items far superior.
Yes, the UN will debate it, but it will be like the debate on land mines. A lot of hand wringing, but nothing really getting done, and the belligerent parties will still make them.
Right now, it is only a matter of perfecting manufacturing. I wouldn't be surprised to see in 5-10 years that sentry robots, which shoot at anything that doesn't have some form of friendly transponder, will become the norm on not any military post, be it Russian, Chinese, Saudi Arabian, or any other place that needs area denial.
Lets be real here... a couple independently active robots with high RPM machine guns are a lot more reliable than soldiers/guards, have no moral issues, have no morale issues, and will "just work". Someone takes one out with a rocket, another can easily return fire.
Add sentry UACVs to the mix, and a rocket attack would be responded in kind.
I wouldn't be surprised to see even civilian warehouses (a data center in a rural area) protected by autonomous firing machines soon. Might makes right, and SCOTUS has shown that money is speech, so any casualties from these would have no criminal/civil consequences ("there was a warning sign".) I would also not be surprised to see this on train tracks and other places, where there isn't a need for it, but the fear of being gunned down by a robot will keep kids from putting pennies on tracks.
Look how tasers are overused. Expect the same thing with these.
Stop with the "3 laws" nonsense. Asimov's "laws" were never intended as actual laws, they were a plot device, and they're certainly not something you "delete" because they were never there in the first place. We already have regulations about machine safety (I work with them every day). The laws govern the control of hazardous energy in a system, with various guarding and interlocks being required to protect humans from injury when they interact with the system, and design constraints determined by how likely certain safety critical component failure is, and redundancy, etc.
Nobody building a killer robot is going to be worrying about any laws, pretend or otherwise. They're worried about how many units they can sell.
"I have never let my schooling interfere with my education." - Mark Twain
I don't know how robot soldiers identify targets, but presuming they have some mechanism whereby they only kill armed combatants it's not hard to see some advantages over human soldiers at least with respect to civilian noncombatants.
More accurate fire -- ability to use the minimal firepower to engage a target due to superior capabilities. Fire back only when fired upon -- presumably robots would be able to withstand some small arms fire and thus wouldn't necessarily need to shoot first and wouldn't shoot civilians.
Emotionally detached -- they wouldn't get upset when Unit #266478 is disabled by sniper fire from a village and decide to kill the villagers and burn the village. You don't see robots engaging in a My Lai-type massacre.
They also wouldn't commit atrocities against civilians, wonton destruction, killing livestock, rape, beatings, etc. Robots won't rape and pillage.
I expect this will be as successful as the UN's 1990-era anti-mine treaty (the Ottawa Treaty - http://en.wikipedia.org/wiki/O...), with over a hundred signatories, but not Russia, China or the United States. http://en.wikipedia.org/wiki/L...
Taking such action really is a bad idea. An autonomous killing machine could be as complicated as as a military drone with hellfire missiles or as simple as a car loaded with autonomous weapons designed to engage any anything that move, with a GPS pre-determined route and self-driving capability, sitting like a mobile minefield in an abandoned house long after the occupants have left, waiting to be activated.
I think the appropriate course of action would be to feed international condemnation of such tactics until they are treated with ruthlessness by the international community against any involved in use of such weapons, for any infraction. Just like the use of chemical weapons should have been...
Autonomous weapons are far more frightening that WMDs... And nowhere is safe.
Then again, I wrote a book on the creation of a universal standard for determining if an autonomous weapon could be trusted with the decision to kill, so perhaps I am somewhat hypocritical there.
GrpA
Enjoy science fiction? "Turing Evolved" - AI, Mecha, Androids and rail-gun battles. What more could you want?
Well, you make his point eloquently.
The holocaust was conducted clearly by an advanced state, signatory to many treaties and international obligations and "laws", none of which served to make any difference whatsoever, when that state decided they didn't care what the rest of the world thought.
But why stop there? Rwanda, Stalin's purges, China's Cultural Revolution, Kashmir, Iraq-Iran, until the U.S. got actively involved, all the U.S. wars against brown people, etc., etc., etc. When has international law, regulations, or even opinion, ever changed the conduct of an aggressor nation when they decided to go to war? The reason nukes haven't been used since Nagasaki is only because everyone who has them is afraid if they used them in aggression, it would trigger a much higher escalation, and has nothing to do with any treaties, laws, or world opinions.
"Unheard of means only it's undreamed of yet,
Impossible means not yet done." ~~ Julia Ecklar
...as with most technological weapon issues, those with them, or with a reasonable chance of developing them will defend the idea.
Those without will roundly condemn it using a great deal of moral and ethical language, but their base issue is that they cheerfully condemn the use of any weapons that they cannot yet field.
The UN as a clearinghouse organization for multinational efforts does a massive amount of good that would otherwise be difficult to enable. ... for reasons based entirely on their OWN narrow self-interests. (Not to mention its main actual value: a way for the favored scions of grubby tinpot regimes to be prostitute-frequenting scofflaws in a place far nicer than their own pestilential capitals.)
The UN's general chambers are worthless talking shops where inconsequential states get to criticize significant, powerful states for acting in their own narrow self-interest
-Styopa
The Super Aegis II has a 12.7 mm machine gun and a grenade launcher. laser and infrared sensors that see 3 km in the day, 2 at night. But the gun probably can't shoot that far - it just sees that far.
excitingthingstodo.blogspot.com
Beyond that, I see another problem with the idea of banning development of autonomous weapons: most of the technology involved would probably be developed anyway because it would be widely applicable.
Think about it. If you were going to make a killer robotic soldier, what technology would be hard to develop? It's difficult to make a robot that can easily traverse diverse terrain, but we'll work on that for other reasons. Making an AI that can accurately identify people by facial features, clothing, and speech patterns would be hard. We'd also do that for other reasons. There are a million reasons to develop and intelligent free-roaming robot who can identify people and interact with them.
Once you have a robot that can do those kinds of things, turning the "interaction" into "fire a weapon at that person" is easy. It's not hard for a computer to aim once it has its target. It's not hard for a computer to trigger the weapon itself.
Actually, part of the purpose of the 3 laws stories was to show that even if you built robots from the ground up to not harm humans, you can still end up in situations where robots are dangerous to humans. Almost every 3 laws story revolves around trying to determine why the three laws failed. This becomes more and more true as the robots become more and more sophisticated; primitive robots cause minor hassles, more advanced robots risk death and serious injury, more advanced yet take over the planet to reduce the total harm to humanity in general (yes, the stupid movie plot is, in fact, based (loosely) on one Asimov's stories, though in Asimov's story the takeover was completely non-violent).
You don't see robots engaging in a My Lai-type massacre.
They also wouldn't commit atrocities against civilians, wonton destruction, killing livestock, rape, beatings, etc. Robots won't rape and pillage.
Well... You won't see them independently decide to do something like that. But orders are literally orders to a robot. You tell them to burn a city to the ground, shoot anyone who tries to flee, and they will burn that city to the ground and shoot everyone who flees. Without remorse, without second guessing orders, without a moment of any hesitation.
Which frankly, worries me a bit more. Because the upper levels of command in just about every model of human hierarchy always seems to have worrying numbers of psychopaths/sociopaths beyond what you'd expect in a normal pool of the population. On top of that - they're physically removed from the carnage. It's a lot easier to order the leveling of a rebel-occupied village when you will never personally see the slaughter of innocents that result.
That's not to say humans never do these things. Just that, humans are capable of refusing to do these things. Robots aren't.
I like killer robots. Know why? Cause I know how to make them, muthafuckers! Mwhahahahahahha!!!
-1 Uncomfortable Truth
Automated weapons are already deployed - the Korean DMZ is guarded, in part, by autonomous sentry drones. If it moves, they shoot it - and they're armed with machine guns or automatic grenade launchers.
That's a good model. Don't try to make a drone that can distinguish targets from non-targets - make something that treats everything as a target, and deploy it only when you don't have non-targets to worry about. Or, to allow your own forces to operate in the area, provide an IFF transmitter to designate them as non-targets (civilians are still fucked though - so don't use it anywhere near civilians). Works fine for air, land and sea - we already have an established concept of "shoot to kill zones", this just replaces the soldiers under orders to shoot anything that moves with robots under programming to shoot anything that moves.
For automated weapons deployed outside such areas (or even ones within), I would say that a human still has to give the fire order. The automated system can identify targets, track them, pursue them, prioritize targets, do basically everything but pull the trigger, but it has to request permission to fire from a human operator. And for all legal and ethical purposes, that human operator can be considered the one who pulled the trigger. It's still some massive force multiplication even compared to modern drones, so I don't see why the military would have much problem with it.
Let's wait until after we get true AI before we try to give machines the responsibility to decide whether or not to kill someone.
Will these killbot have pre-set kill limits?
No, but if you stay off my lawn, you should be safe
-1 Uncomfortable Truth
Anyone citing them also has a serious lack of understanding in how programming works.
Hey I know we should create 3 laws of software to be implemented in all code thats written:
I think I just solved the problem of software bugs!
The point is that massacres like Mai Lai are caused when soldiers go a bit nutty due to the emotional stress of seeing their friends cut down by an insurgent resistance. That stress isnt going to be there if you're a remote operator, and youre generally going to be much better supervised by pencil-pushers as a drone operator than as an infantryman in a hostile country.
Just that, humans are capable of refusing to do these things. Robots aren't.
And as history shows us, humans dont. Forget about the holocaust, the cultural revolution, the soviet purges? Humans go with the crowd, especially in emotionally charged situations. Anything that brings down the emotion and brings some sanity to combat situations is a good thing.