15:45
So yes, one could say similar things about machines (designed by humans in order to do very specific things). The machine may have some features put in there by the human designer but it also may have other unexpected features that are exposed through use.
@Cerberus 'everyone knows'... I don't think that is the case.
@Cerberus It is very very hard to understand 'intelligence'. Just trying to understand our own is hard enough.
@Cerberus The usual dichotomy is extrinsic vs intrinsic: things you can see from the outside (behavior, functionality, what it can do) vs things you can see in the construction (how it was built/designed, how it works). I think the latter is what you mean by 'structural'.
Behaviorism/functionalism/instrumentalism/physicalism (there are slight differences among those) philosophically work mostly on just extrinsic things. They may be inspired by our introspections or imaginations, but what we can know is only by external data.
Behaviorism as a psychological fell out of favor 70 years ago because it was too extreme almost denying thought altogether (which is, despite thought being inaccessible to others, kind of ludicrous.
But behavior as a definition of intelligence is scientifically more supportable (like we can do intelligence experiments with babies even though they can't talk but there are other things we can measure like heart rate, time spent looking at something, or rate of sucking).
I've probably spoken like this, that some new AI technique is claimed as being great and then I'll say it's not so great because 'look how it is built, that's not intelligent'.
eg chess programs, they're 'just' using a tree search, that's not 'intelligence'.
But one could counter that that doesn't matter, people can't beat the chess programs.
Or another example, since about 1890, cars have been faster than horses. Who cares that cars use metal and mini gas explosions (or an electric motor) instead of muscle and bone? Cars are faster, why would you care how it did it?
As to 'faster', totally, cars are better.
But for the term 'intelligence' people tend to use that word badly. "Deep blue is so intelligent it can beat grand masters at chess. How about checkers, a much simpler game? Oh. No, we'd have to reprogram it."
"Hey we have an AI program that predicts recidivism with 95% accuracy. How does it do it? Oh it's actually just checks their race and zip code and if it is over a threshold .."