Slashdot Mirror
US Navy Wants Smart Robots With Morals, Ethics
coondoggie writes: "The U.S. Office of Naval Research this week offered a $7.5m grant to university researchers to develop robots with autonomous moral reasoning ability. While the idea of robots making their own ethical decisions smacks of SkyNet — the science-fiction artificial intelligence system featured prominently in the Terminator films — the Navy says that it envisions such systems having extensive use in first-response, search-and-rescue missions, or medical applications. One possible scenario: 'A robot medic responsible for helping wounded soldiers is ordered to transport urgently needed medication to a nearby field hospital. En route, it encounters a Marine with a fractured leg. Should the robot abort the mission to assist the injured? Will it? If the machine stops, a new set of questions arises. The robot assesses the soldier’s physical state and determines that unless it applies traction, internal bleeding in the soldier's thigh could prove fatal. However, applying traction will cause intense pain. Is the robot morally permitted to cause the soldier pain, even if it’s for the soldier’s well-being?'"
Imagine us trying to teach a robot morality when humans have little agreement on what is moral. For example would a moral robot have refused to function in the Vietnam War? Would a drone take out an enemy in Somalia knowing that that terrorist was a US citizen? How many innocent deaths are permissible if a valuable target can be destroyed? If a robot acts as a fair player could it use high tech weapons against an enemy that had only rifles that were made prior to WWII? If many troops are injured should a medical robot save two enemy or one US soldier who will take all of the robot's attention and time? When it comes to moral issues and behaviors there are often no points of agreement by humans so just how does one program a robot to deal with moral conflicts?
US armed forces should want leaders with morals and ethics, instead of the usual bunch that send them to die based on lies (I'm looking at you Chenney, you bastard).
`echo $[0x853204FA81]|tr 0-9 ionbsdeaml`@gmail.com
Every single one comes down to "do I value rule X or rule Y more highly?" Who gives a shit. Morals are things we've created ourselves, you can't dig them up or pluck them off trees, so it all comes down to opinion, and opinions are like assholes: everyone's asshole is a product of the culture it grew up in.
This is going to come down to a committee deciding how a robot should respond in which situation, and depending on who on the committee has the most clout it's going to implement a system of ethics that already exists, whether it's utilitarianism, virtue ethics, Christianity, Taoism, whatever.
If they are talking about the moral of the US government, I rather have the robots from Terminator.
And they are talking about helping wounded soldiers. Why talk about the (US) marine with the broken leg? What about the injured Al-Quaida fighter?
The question of causing pain for the better wellbeing of the patient is obvious for most people. What if it means killing 1 person to save 10? What if that one person is not an enemy?
What if it realizes that killing 5% of the US population would save the rest of the world? What if that 5% is mostly children? Even if you can answer that as a human being, would you want it enforced by robots?
Don't fight for your country, if your country does not fight for you.
It would be great if they could develop a politician with morals and ethics........but I doubt even the Pentagon's budget would be big enough...........
"The average reporter we talk to is 27 years old......They literally know nothing." - Ben Rhodes
If they calculate that you can't be helped and must be left to die, just say, "Sorry, I've been given specific orders to do X, so I can't help you."
All of this 'ethical debate' surrounding robots that can make life-or-death decisions has absolutely nothing to do with technology, or AI, or any issue that can be resolved technically at all. All it boils down to, is that people are mad that they can't hurt a robot that has hurt them. See, before machine intelligence we had a pretty sweet system. When a human being commits a crime, we stick them in prison. It doesn't feel good to be in prison, therefore this is "justice." But until robots can feel pain or fear or have a self-preservation instinct, prison (or, hell, even the death sentence) wouldn't affect them at all. And that's what drives people nuts. That technology has shown us that beings can exist that are smart enough to make life-or-death decisions, but lack the concept of pain or suffering and if they do something bad there's no way we can PUNISH them.
A fool and his hard drive are soon parted.
The chassis is the hard part, not the ethics. The ethics are dead simple. This doesn't even require a neural net. Weighted decision trees are so stupidly easy to program AIs that we are already using them in video games.
To build the AI I'll just train OpenCV to pattern match wounded soldiers in a pixel field. Weight "help wounded" above "navigate to next waypoint", aaaaand, Done. You can even have an "top priority" version of each command in case you need it to ignore the wounded to deliver evacuation orders, or whatever: "navigate to next waypoint, at any cost". Protip: This is why you should be against unmanned robotics (drones): We already have the tech to replace the human pilots and machine ethics circuits can be overridden. Soldiers will not typically massacre their own people, but automated drone AI will. Even if you could impart human level sentience to these machines, there's no way to prevent your overlords from inserting a dumb fall-back mode with instructions like: Kill all Humans. I call it "Red Dress Syndrome" after the girl in the red dress in The Matrix.
We've been doing "ethics" like this for decades. Ethics are just a special case of weighted priority systems. That's not even remotely difficult. What's difficult is getting the AI to identify entity patterns on its own, learn what actions are appropriate, and come up with its own prioritized plan of action. Following orders is a solved problem, even with contingency logic. I hate to say it, but folks sound like idiots when they discuss machine intelligence nowadays. Actually, that's a lie. I love pointing out when humans are blithering idiots.
Is funny because since WWII the army has worked to get the kill rates up. In WWII only 15% of soldiers shot to kill, but they the army brainwashes them so that 90% kill. Moral. Killers. Can't have both.
And Moral and Ethical for the NSA? LMAO.
Navy says that it envisions such systems having extensive use in first-response, search-and-rescue missions, or medical applications.
Just like drones were first used for intelligence gathering, search and rescue and communications relays.
That's our life, the big wheel of shit. - The Fat Man, Blue Tango Salvage
What they want is a robot that will not embarrass them, but that will do their killing for them. I want a pony, but I can't have one. The situation here is similar. Coding up a robot that makes ethical choices is so far beyond the state of the art that it's laughable. Sort of like the story the other day of the self-driving car that decides who to kill and who to save in an accident.
When will they figure out that what you really need is a robot that will walk into the insurgent's house, wrestle the gun from his grasp, and cuff him? There's no need to shoot anyone—the robot is not in danger.
He devised a system he called "Utilitarian Dynamics"
He had a formula, V=DNT
V is value
D is degree; i.e how happy or unhappy somebody is
N is number; the number of people
T is time, how long they were affected
Morality is very tricky, but objective attempts to quantify and make optimal decisions cannot be a step in the wrong direction. Maybe well programmed machines will help improve human behavior.
Greed is the root of all evil.
>There are lots of ways of doing this, but the best way is by killing them
Correction: The most effective way is killing them. There's a difference. In a real war it should always be remembered that the folks shooting back at you are just a bunch of schmucks following orders, just like you. The actual enemy is a bunch of politicians vying for power who have never set foot anywhere near an active battlefield. And not necessarily the ones giving orders to the *other* side.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
Actually, with the invention of the NATO round, bullets are designed to maim instead of kill. This way, one bullet can take out 2 or 3 people from the immediate "action". One to get shot, and up to two to carry the wounded solider to safety.
"Freedom in the USA is not the ability to do what you want. It is the ability to stop others from doing what THEY want"
I think what the parent means to say is that, in a war created by politicians, it should be fought by politicians. My Prime minister doesn't like your president. Ok. Grudge match! Stick em both in a ring and let them fight it out. First blood, till death, whatever. Doesn't matter. Or perhaps a forfeiture of that leader's assets should be on the line. Hit em where it hurts. You lose, you retire and lose the entirety of your assets to the victor.
Point being... leave the rest of us out of it.
"Freedom in the USA is not the ability to do what you want. It is the ability to stop others from doing what THEY want"
The problem is not putting morality into machines. The problem is letting these machines execute this "morality" in a complex environment with life-or-death stakes.
I program protective systems in factories. A guy opens a monitored gate and walks into a conveyor area. If the conveyor runs while he's in there, he will have a messy and painful death. The conveyor "knows" it's "wrong" to move under those conditions.
We don't use the word "morality", we say "safety". When auditing the software that lets the conveyor run I find a potential for exposure. Hang a lock and shut it down until it's fixed. The first law. In all except one case over 20 years I have received full support from production for similar decisions. That's precedence of the first law over the second law.
We know what's "right", and we try to teach that to the machines. Industrial safety operates with a very limited set of variables, and exception handling is simple--just STOP. Immediately stop moving and disconnect all power.
This idea that we can "program" morality and justice is not problematic. Of course we can; we write code for those things all the time. Heck, it's even called "code". Here's some. Here's some more. The execution engine for this code is a group of complex elements (people) from which emerges an even more complex "society". This "morality" execution engine constantly goes hideously and indefensibly wrong.
Now we want to create far simpler code and execute it with a far simpler machine. But with the same stakes and in the same environment. And the source of this simplified code and execution engine is the existing society, particularly the part of it most directly involved in many of the atrocities.
It's wise they're starting with autonomous "good" machine--search and rescue, first responder, etc. Maybe they'll learn something before humanity starts building autonomous "bad" robots. Thankfully, those exist only far into the future...
"Reality is that which, when you stop believing in it, doesn't go away." - Philip K. Dick
Not at all. Every institution needs some kind of morality to guide its actions. One can debate whether a particular institution should exist, altough in our current level of development it seems unlikely we could do without military, and of course even in a completely peaceful one we'd still need personnel and equipment for difficult or dangerous missions, such as search and rescue; but once one does, it needs rules about what's desirable or acceptable and what's not.
But then again, I guess posting edgy slogans is less demanding than identifying specific problems, evaluating possible fixes to them, or even discussing the ethics of national defence in general.
Forget magic. Any technology distinguishable from divine power is insufficiently advanced.
”So many vows. They make you swear and swear. Defend the king. Obey the king. Obey your father. Protect the innocent. Defend the weak. What if your father despises the king? What if the king massacres the innocent?” - Jaime Lannister
APK, you are a fucking psycho. Enough with the stalking, already. And no, I'm not K.S. Kyosuke, nor do I have anything to do with him.