Isn't Deep Blue (the chess program) an excellent example of an application in AI? From what I know of the program, it can make novel decisions based on past "experience", which means that there's some "learning" taking place.
Re: Re: Re: Artificial Intelligence Programming in Perl
Replies are listed 'Best First'.
No, I would not consider any chess program to be an example
of an application in AI. Most, if not all, chess programs
are just brute force algorithms, combined with extensive
libraries for the opening game. After the opening, they just
explore all possible plies (a ply is a half-move) to a certain
depth, evaluate the postions, and prune to select the best moves
so far. Then they repeat the process of "try all, evaluate, prune
to select best so far" zero or more times.
This doesn't even come close to how a human chess player works.
Not much is know how exactly a human chess mind works, but
(grand)masters only explore a few moves in depth when its their
turn to move.
As for the learning process of Deep Blue, the only ones that
are "learning" are the programmers.
After the opening, they just explore all possible plies (a ply is a half-move) to a certain depth, evaluate the postions, and prune to select the best moves so far.
This process of examining and evaluating usually uses some set of heuristics to value each move.This process is an example of applied AI. I have however heard it said, that once AI become applied it ceases to be AI.
Cool! Many of the problems I've worked with have been (provably) NP-hard. Of course, the first thing you learn as an algorithms guy in the industry is that NP-hard problems are not insurmountable: typically the instances of the problem which actually show up are easy (in some appropriate sense), and a good heuristic handles most of them quite well (although not optimally).
I designed and coded some heuristics to speed up such searches. Does that make my work "an example of applied AI"? Or is it just a solution to a practical problem?
Perl's regexp pattern matching is NP-hard (I believe Abigail-II wrote about this very topic). But ingenious heuristics (coupled with occasional "better" choices of regexps) make Perl's regexp matcher indecently fast for most practical applications. Is the Perl regexp matcher "an example of applied AI"? Or is it just a solution to a practical problem?
Ask yourself this: Are these programs displaying anything like intelligent behaviour? Or are they just glorified adding machines?
Once AI becomes applied, I'll start believing in it. For the time being, the burden of proof is upon its proponents.
Actually, it values each position, not each move. But if you call this
technique "applied AI", then that term because shallow
and not very meaningful anymore. From there on, it would
be easy to call any algorithm "AI".