Amos, you're assuming that because we don't know how you or I "intend" something, it is and will always be inexplicable. I guess you're also assuming that "intending" requires the same physical process (ie. a few pounds of grey mush using currently-unknown methods) to be used for anything else that can "intend".

With emergent systems, very simple systems with very simple rules can have amazing behaviour. That behaviour comes from the initial rules, but you couldn't predict that behaviour in advance from the rules. And if you make the system and rules more complex, you're increasing the scope of what behaviours are possible. With a computer, you always know you're feeding it a program, so you know in advance that your neural network program would have the capacity to learn behaviours, for example - you've set it up so the rules allow that. But then instinct is nothing more or less than preprogrammed behaviours, hardwired into our brains, and the ability to learn is certainly hardwired in us, so we ain't so fundamentally different.

It's certainly true that no computer today is capable of meeting the Turing test - fooling a human into thinking they were talking to another human. But with the powerful hardware available today, that's more a factor of lack of software than lack of hardware. Keep plugging at the software, and I reckon it'll happen eventually.

Remember that we've had computers (as we know them) for less than 40 years, and it's only in the last 10 that computers with any reasonable power have been available for these kind of experiments. I think you're doing the equivalent of a Galileo-era person saying "OK, maybe Mars is a planet like ours, but it's never going to be possible to do anything with it, because even our best siege engines couldn't shoot an arrow that far..." ;-)

Graham.