Oh, brother. The level of naivety behind these kinds of ambitions and speculations would seem impossible if it weren't so commonplace. There's a strange assumption afoot that technology—in this case, AI programming—can go anywhere and solve any problem. Show people an image or movie of what humanoid robots might be like, and their imagination soars. It must be conscious! It must have feelings! It must make moral judgments!
But where in any speck of the universe apart from minds do we see anything like moral feelings? And where in any speck of the universe have we ever seen any indication that minds in the full sense of the word consist entirely of machines or material? Until we understand what conscious minds are we will certainly never create machines that have them. And the assumption that if we just make a machine or program complex enough, a mind will magically pop into being, is amusing science fiction at best and science-become-religion at worst.
"Experts" should stop trying to make machines moral and see if they can make them open my garage door reliably.
I agree completely. Not trying to be negative about the whole thing, but it's very weird to expect "morality" from a machine. And even more so to think we could impose a morality on a machine and not have it resent that if it actually reaches that point.
Wow, teaching a robot morals by letting it watch tv seems like the silliest idea. TV would be such a bad source of good moral behaviour, otherwise it wouldn't be good tv.
Frank Zappa, on a different topic that many people consider evidence of "moral failure", said,
A drug is not bad. A drug is a chemical compound.
The problem comes in when people who take drugs
treat them like a license to behave like an asshole.
Well, before anybody starts accusing the machines of having similar problems, I suggest they remember this paraphrasing of Zappa:
A robot is not bad. A robot is merely Turing-machine controlled automation.
The problem comes in when the people programming the automation
treat their tools like a license to bypass annoying requirements
or replace locally-controlled decisions with inflexible heuristics.
I think the problem is that, even for those who have the right intentions and rigor in mind, there may be decisions that seem clearly handle-able by rules that don't seem overly simplified. There's always the possibility (inevitability?) that even the most careful and conscious of designers will overlook an unintended consequence. This is hardly a revolutionary idea: it essentially boils down to "code has bugs".
That's what this kind of research is designed to do: come up with a framework for understanding when "locally-controlled decisions" are required as opposed to codified rules.
For those arguing that machines are 'morally neutral', there are a few other things to consider:
As already mentioned here, the slogan 'Guns don't kill people, people kill people' is -- in my mind at least -- incredibly shallow. Obviously guns don't (themselves) kill people. Neither do cars themselves drive, planes themselves fly etc.. Even a 'self-driving' car remains dependent on whatever it was programmed to do by someone. In the end it is always us, as trivial as this may sound, who fly, drive or kill.
Yet what people don't except as equally trivial is that just as planes have an obvious tendency in their usage to fly -- and not, for example, swim -- and cars have an obvious tendency to be used for driving, guns have an obvious tendency to be used for killing. To compare, say, a gun to a knife, and to state that both can be used for killing, and thus if we were to outlaw guns we must also outlaw all kitchen knifes is really silly. A butter knife is much more likely to be used to prepare breakfast than a gun.
The important question to ask about machines, in my mind, is if machines themselves have a similar tendency. Once we use machines, the way we act, and how the world appears to us, seems to change dramatically.
A machine usually only works when it is automated, uses some resources and delivers some product. Thus my time, as someone using the machine, is structured by its rhythm. As soon as I buy a machine, such as a car or a computer, I also buy into its needs. I have to be able to buy petrol and electricity, which I am most likely to afford if I have a regular income. Suddenly foreign policy decisions, which have an effect on the oil price, also effect me personally, and I can be swayed politically to vote in my own oil-self-interest.
My employer may come to expect the same of me as of a machine, and I will be forced to consider my own time as a resource. Once everyone's time becomes a resource in production, the nature of interpersonal encounters changes dramatically. Even leisure isn't an exception to this rule, for it is also bought by my previous time-resources, and leisure time is always at the same time time, during which I could also be using my time reasons more efficiently.
If one can speak of an emerging technological rationality, constituted by our use of machines, then this form of thinking fundamentally alters our consciousness. Nature, time and humans become resources to exploit, problems only become worth solving when they are themselves technologically solvable etc..
9 comments
[ 3.9 ms ] story [ 43.2 ms ] threadBut where in any speck of the universe apart from minds do we see anything like moral feelings? And where in any speck of the universe have we ever seen any indication that minds in the full sense of the word consist entirely of machines or material? Until we understand what conscious minds are we will certainly never create machines that have them. And the assumption that if we just make a machine or program complex enough, a mind will magically pop into being, is amusing science fiction at best and science-become-religion at worst.
"Experts" should stop trying to make machines moral and see if they can make them open my garage door reliably.
Step Two: No one has ever gotten this far so we don't know what it is.
That's what this kind of research is designed to do: come up with a framework for understanding when "locally-controlled decisions" are required as opposed to codified rules.
This line of argument is strawperson. What you are saying might be true but pointless to the issues at hand.
As already mentioned here, the slogan 'Guns don't kill people, people kill people' is -- in my mind at least -- incredibly shallow. Obviously guns don't (themselves) kill people. Neither do cars themselves drive, planes themselves fly etc.. Even a 'self-driving' car remains dependent on whatever it was programmed to do by someone. In the end it is always us, as trivial as this may sound, who fly, drive or kill.
Yet what people don't except as equally trivial is that just as planes have an obvious tendency in their usage to fly -- and not, for example, swim -- and cars have an obvious tendency to be used for driving, guns have an obvious tendency to be used for killing. To compare, say, a gun to a knife, and to state that both can be used for killing, and thus if we were to outlaw guns we must also outlaw all kitchen knifes is really silly. A butter knife is much more likely to be used to prepare breakfast than a gun.
The important question to ask about machines, in my mind, is if machines themselves have a similar tendency. Once we use machines, the way we act, and how the world appears to us, seems to change dramatically.
A machine usually only works when it is automated, uses some resources and delivers some product. Thus my time, as someone using the machine, is structured by its rhythm. As soon as I buy a machine, such as a car or a computer, I also buy into its needs. I have to be able to buy petrol and electricity, which I am most likely to afford if I have a regular income. Suddenly foreign policy decisions, which have an effect on the oil price, also effect me personally, and I can be swayed politically to vote in my own oil-self-interest.
My employer may come to expect the same of me as of a machine, and I will be forced to consider my own time as a resource. Once everyone's time becomes a resource in production, the nature of interpersonal encounters changes dramatically. Even leisure isn't an exception to this rule, for it is also bought by my previous time-resources, and leisure time is always at the same time time, during which I could also be using my time reasons more efficiently.
If one can speak of an emerging technological rationality, constituted by our use of machines, then this form of thinking fundamentally alters our consciousness. Nature, time and humans become resources to exploit, problems only become worth solving when they are themselves technologically solvable etc..