| I suspect if someone managed to create a system (probably a distant descendant of current neural nets) capable of human-level intelligence, certain intellectual "emotions" would develop, especially if the system was partially developed in an evolutionary manner. They wouldn't be the same strength as human emotions since the machine would lack the hormonal and other chemical boosters, but things like anger and friendship have real survival advantages. Under the right circumstances, they could probably also be taught to a still-developing network.
Ok, got a bit off track there. Anyway, if whoever is training the system is capable of defining actions as "right" and "wrong", they would exist sufficiently to train the network. If the network is sufficiently complex, it should be possible to start with a basic punishment/reward basis of "right and wrong" like you would with a small child, and then work up through the various more complex stages of right/wrong evalution.
Remember, the human brain is essentially a hypercomplex neural network, and there's no real reason to presume that abstract thought is based in hormonal and chemical influences. In fact, given that the ability of many people to consider such abstractions as right and wrong degrades significantly when being influenced by strong emotions, it seems likely that hormonal effects reduce the ability to deal with abstract concepts. |
