South Korea Drafting Ethical Code for Robotic Age

goldaryn writes "The BBC is reporting that the South Korean government is working on an ethical code for human/robot relations, 'to prevent humans abusing robots, and vice versa'. The article describes the creation of the Robot Ethics Charter, which 'will cover standards for users and manufacturers and will be released later in 2007. [...] It is being put together by a five member team of experts that includes futurists and a science fiction writer.'"

Comments

Will the next step be "robot rights"?

[Opportunist]

Because one thing's quite blatantly clear, robots are by their very definition slaves. They are owned, they exist to do work we don't want to do (or which is hazardous), they don't get paid and they are only given what's needed for their sustainance, they can't own property etc.

I fear the day when we create the first truely sentient robot. Because then we will have to deal with that very question: Does a robot have rights? Can he make a decision?

And I'd be very careful how to word the charta. We have seen that the "three laws" ain't safe.

Re:Will the next step be "robot rights"?

[ubergenius]

If we ever created a truly sentient robot, it would have to be given rights. That's not debatable.

What is debatable is, when do we know a robot is sentient? We barely have a definition for sentience, much less a method for identifying it's existence in a machine. Until we figure that out, it will be near impossible to tell if a robot is sentient or just really well programmed.

Re:Will the next step be "robot rights"?

[oddaddresstrap]

I fear the day when we create the first truely sentient robot.

And we all should. If (some would say "when") that day comes, the robot will likely have more or less unlimited knowledge at its disposal (fingertips?) and the ability to process it much faster than people. The first thing it will figure out is how to eliminate or at least control people, since they will be the greatest danger to its survival. After all, that's what we do to species that endanger us.

Re:Will the next step be "robot rights"?

[gsn]

You raise a great point but its even harder than that.

Until we figure that out, it will be near impossible to tell if a robot is sentient or just really well programmed. Is there a difference? For humans even? What if in the process of creating sentient robots we find that we aren't really all that free thinking (I'm not implying any kind of design here but someone is going to raise that issue as well).

I argued this for a hypothetical cleverly programmed machine that could pass a Turing test. Strictly, it would simulate human conversation based on some clever programming, which my professors claimed did not amount to machine intelligence. The counter being how do you prove that human conversation is not based on some clever rules.

It might be possible to define a set of rules for conversation between humans in restricted circumstances - I wonder if anyone has actually tried doing this. I'm fairly certain a lot of /. would like the rule set for conversation with pretty girls in bars.

Re:Will the next step be "robot rights"?

[kalirion]

Pascal's Wager basically says that if there's no god then it doesn't matter what you believe, but if there is a god then you had better believe in him. Even though Pascal's wager may be invalid when it comes to belief in a particular god, it may be reasonable when applied to a meaning for the universe as a whole. In other words, if the universe is pointless then it doesn't matter what you do, but if there is a purpose to the universe then it does matter what you do. If you don't know what the purpose is then I guess the first step is to figure out what the purpose is.

To me, that's a pretty invalid argument. Why should I give a rat's ass what the universe's "purpose", if any, is for me. I only care what purpose I give myself. Even should the universe have a purpose for me and I learn what that purpose is, if it turns out to be something that I don't agree with, too bad for the universe. And yes, if God exists this applies to Him/Her/It as well.

I remember a short scifi story, I think it was in Clifford Simak's "Strangers in the Universe" collection. There humans on some planet spent millenia on building a computer that could answer any question. The questions were "What's the purpose of the Universe" and "What's the meaning of life." The answers were something like "The Universe has no purpose, the Universe just happened" and "Life has no meaning, life is an accident." After learning this, the humans abandoned all their technology and settled to an Amish-like lifestyle. Which as far as I'm concerned is fine if that's what they really wanted to do, but I see no reason why people should give up just because something higher than them doesn't assign them a purpose.

In other words, if God exists and he created humanity for no purpose other than to have someone to worship Him, would you accept that purpose, or would you attempt make your life have a meaning beyond that?

Who represents the robots?

[VWJedi]

It is being put together by a five member team of experts that includes futurists and a science fiction writer.

If we're creating laws about how humans and robots should treat each other, shouldn't the robots be part of the decision-making process? This sounds a little too much like "the founding fathers" determining what rights slaves had (not many at the time).

seriously, why does anyone care?

[HelloKitty]

make robots without emotions - essentially machines, pistons, actuators, CPUs, etc... and WTF, who cares how much you use it, replace the parts as they wear out like any machine...

why would anyone install emotion into a worker robot anyway?
and even if it had emotion, the only reason to "treat it right" is so they don't start the robot uprising against humanity. which is a good reason... but that begs the question, why give real human emotion to something you want to abuse? for menial labor, keep the emotions out, let it be purely a machine.

this is a waste.

Re:seriously, why does anyone care?

[Rei]

Emotion could be seen as inherent in sentience. If you try to create a sentient robot for any of a wide variety of reasons (say, a military robot that can't be easily outsmarted by insurgents, or a household robot that needs to be able to interact with people and understand more than basic commands), its neural net will need to be trained: rewarded positively when it gets things right, negatively when it gets things wrong. Emotion could potentially be an emergent phenominon from this kind of reward/punishment.

Re:seriously, why does anyone care?

[KDR_11k]

Animal abuse is illegal in many jurisdictions, it's quite thinkable that we'll have robots that are as intelligent as most pets.

Re:seriously, why does anyone care?

[Rei]

Which is why I didn't try to use my ANN as an appeal to authority, unlike the person I was replying to :) That's quite true: we don't know how far we are from sentience. However, it can be said that what we have is, from a purely subjective, observational standpoint, not even close. It's very good at pattern detection. Beyond that, well, we don't have much. Whether there is some small change that will lead to a big subjective leap, or whether it will take a long, tedious process of incremental improvements, who can say? I read an interesting paper a few months ago looking at why we haven't achieved more, postulating several theories. There's the "we just haven't thrown enough processing power at it" theory. There's the "There's something biological that we don't know about yet" theory. There's the "There are a combination of known biological factors that, while we know about and have tried them individually, it is their net action that is problematic" theory. And on, and on.

The question is where to invest your resources. Do you simplify your model of a certain feature so that it can be simplified mathematically for more effective computing power, but risk losing the effects caused by what you simplified? Do you do learning, genetic algorithm selection of fixed nets, or take the major computing power hit and try to do both? What biological features, exactly, do you choose to include? Or do you go for an abstract system not based on biology at all, but something that should lend itself to computing better? There are so many possible tradeoffs one can make.

A good example of how much a little effect can make a big differences comes from a (Navy?) audio research project several years back that I read about at the time. They had a net for audio processing, designed to detect submarines. Their earlier models had performed very poorly, but their latest had worked incredibly well. What did they change? Just one thing: they modelled the delay for signal propagation between neurons. That one little thing made their net go from performing a fraction as well as a human to performing many times better than a human.