> What does everyone think of artificial intelligence, or AI? Is it possible to create a truly intelligent machine? For that matter, should we try?
>
> Saku"had to do a report on this last year and was troubled by a few of the things she found"ra

The purpose of AI is not simply to create intelligence. Unfortunately there is a heck of a lot of propaganda out there by people who don't understand this -- even by people who work in the field. Most of what you read about AI is bunk. The AI subfield of computer science is designed to produce workable engineering solutions to problems that take too many computations to solve dumbly. "AI is search" was the mantra of my AI professor in graduate school.

For example, if you have an online phone book, and you're looking up someone's name in it, it's no problem for a computer to check each name, in order, comparing it to see if it's the same as the one you're searching for. If you have 100,000 entries in the phone book, the computer algorithm takes 100,000 comparisons to find the name in the worst case and an average of 50,000 comparisons. This is not hard for a computer these days. However, it is a little more intelligent for a computer to rely on the fact that phone books are sorted to find the name it is looking up. It could check the middle name first, then, if it's not it, figure out whether the target name comes before it or after it. Then repeat on the name in the middle of that half. The name is found MUCH quicker this way; both the average and worst cases are much smaller.

This method of "intelligent" searching is actually technically artificial intelligence. No, the algorithm does not incorporate the ability to improve itself and accrue knowledge upon which it can change its method of operation, but that's not all there is or even the most important of what there is to AI. AI is searching a large search space intelligently. But more interestingly:

A game of chess or checkers or bridge. Artificial intelligence is great for gameplaying. Take chess. On the first move, there are 20 possible moves. In the middle game, usually there are more. To make a good move in chess, some "lookahead" is required, which means you need to take into account what your opponent can do after your move, then what you can do after that, then what you can do after that. With 20+ possible moves every turn, looking ahead just two moves for each side comprises 160,000 possibilities. Four moves each is 25,600,000,000 possibilities. A good chess program needs to look ahead maybe eight; a brute-force evaluation of all possible positions after any eight moves would take from now until -- wait for it -- THE HEAT DEATH OF THE UNIVERSE. However, the vast majority of these paths can be discarded. If the move you're considering kills your queen without getting anything back in return, it's probably safe to assume you don't have to lookahead very many more moves AFTER that to discard the possibility, so there's a whole branch of possibilities consisting of millions or billions of game positions that you're able to assume don't lead anywhere worthwhile. By intelligently evaluating each move and making an educated guess on whether it's worth it to look ahead further, you can find a good move after FAR fewer calculations.

That is artificial intelligence. There is very little to artificial intelligence beyond that kind of thing: intelligent searching. The thing about AI is, it truly does make it look like computers are thinking and learning. Once you know how it all works, it's shockingly underwhelming. The potential for computers to become self-aware and ultimately consider humans unnecessary and candidates for extermination isn't there.