If violence, torture, murder and genocide are wrong; then smart machines will not carry them out. So far these things have been the pursuit of humans and not (smart) machines.
That is nice, I suppose. Except that rational and unemotional doesn't tend to see things from any particular sides. I know it sounds a bit far fetched, but what if someone came to your house to murder you and you murdered them? Most people would tend to view you as a hero -- you defended your house, you defended yourself, and you put a murderer off the streets.
A machine would see that you murdered someone in the case where it's if statement didn't contain an "and not" statement about you defending yourself.
An extreme example, but when you try to define everything that can happen in our illogical society into logic statements, something is going to break somewhere. I'm sure there's a solution to the problem, but then again, in every science fiction where something goes wrong with AI, it's generally because somewhere along the line, someone didn't think of the solution before the problem was far too out of control to be fixed.
tl;dr: Too much gray area for ones and zeros.
Slashdot Mirror
If violence, torture, murder and genocide are wrong; then smart machines will not carry them out. So far these things have been the pursuit of humans and not (smart) machines.
Logically define right and wrong.
That is nice, I suppose. Except that rational and unemotional doesn't tend to see things from any particular sides. I know it sounds a bit far fetched, but what if someone came to your house to murder you and you murdered them? Most people would tend to view you as a hero -- you defended your house, you defended yourself, and you put a murderer off the streets. A machine would see that you murdered someone in the case where it's if statement didn't contain an "and not" statement about you defending yourself. An extreme example, but when you try to define everything that can happen in our illogical society into logic statements, something is going to break somewhere. I'm sure there's a solution to the problem, but then again, in every science fiction where something goes wrong with AI, it's generally because somewhere along the line, someone didn't think of the solution before the problem was far too out of control to be fixed. tl;dr: Too much gray area for ones and zeros.