Author: Bruce Moreland
Date: 13:54:08 06/27/01
Go up one level in this thread
On June 27, 2001 at 00:33:26, derrick gatewood wrote: >ok, I made some general statements about AI and chess and how they have been >closely related... this is the response I get back. I was wondering what you >guys think about this. If you can rip its reasoning apart, please do. I would >like a little support when I do it =) > > >"Umm.. Lets see. Chess programs and AI. > >Of course what follows is an Opinion, but here goes anway.. > >What exactly *is* intellegence? > >According to Websters: > >Main Entry: in·tel·li·gence >Pronunciation: in-'te-l&-j&n(t)s >Function: noun >Etymology: Middle English, from Middle French, from Latin intelligentia, from >intelligent-, intelligens intelligent >Date: 14th century >1 a (1) : the ability to learn or understand or to deal with new or trying >situations. > >To learn or understand.... Intelligent behavior is adaptable behavior. It's the ability to generalize over a complex problem domain. >Chess programs do neither. What chess programs do is heuristic search of various On the contrary. If you give a chess program a game situation it has not been specifically prepared for, it will perform well. The program exhibits intelligent behavior -- it can handle a large variety of positions with a degree of competence. >board configurations using a Minumum and Maximum approach. Each board >configuration had a heuristic value associtated with it saying how good or bad >it is for a certain side. All a chess program does is search this database for >the next move that has the best score for it and the worst score for the >opponet. Its only slightly more complex then a sequal query searching for your >friend Joe's phone number. This is not true at all. The database searching domain is not complicated. It's true that designing and building a database is a complex issue, but once you have it built, it's not hard to go look something up. >It does not learn or understrand, hence its not intellegent although its >commonly called AI. Real AI doesnt exist and were no where near making it. When >you can tell a computer an Aesops fable and it can give you the moral of the >story, then you will have an AI, not any sooner. "Learning" is not necessarily part of AI. A program that can decipher handwriting is exhibiting intelligent behavior, since everyone's handwriting is different, and there are cases that aren't specifically predicted. Such a program does not have to adapt in a long-term sort of way. The adaptatation is a shorter-term thing. A human is still intelligent if you wipe his ability to remember things for more than a few minutes. This happened to me when I smashed my head on the ground. I could still react to situations and solve problems, I would just react exactly the same way if I encountered the same circumstances in five minutes. To insist that human-style learning adaptability *must* be involved in intelligence is like insisting that a car is not a car unless it has an ash tray. >Big blue didnt beat that russian chess player (I forget his name), a team of >people beat him, the IBM chess programming team. Chess has been distiled to a >mathimatcal abstraction and it still takes super computers forever to find the >best solution. This is a little like saying Kasparov's parents beat DB in the first match. The entity is created. It has capabilities. Maybe it adds to them and maybe it doesn't. But it deserves credit for its accomplishments. The guy can't remember the name of either opponent, which I think might class him as non-intelligent according to his argument, but I'll overlook that. >Something as abstract as Starcraft, for instance, is VASTLY more complex then >chess. To compare the size of the problem would be like comparing a grain of >sand to the sun (chess being the grain of sand). Even that might not be big >enough. In chess there are only a few units, and they all interact with other >units in the same fashion, if they can move into a spot cointaining another >unit, the other unit is destroyed. Not only are there far more units in >starcraft then in chess, but there are far, far more board positions and each >unit can interact with other units in differnt manners. IE: Hydrolisk vs a ling >is differnt then a Hydrolisk vs a ling in a dark swarm. Games like that can't be handled with a tree search or any other sort of perfect knowldge thing. It is possible to create AI strategies for those games that allow the AI to handle typical cases well, and it may be possible to create an AI that handles oddball cases and generally plays well. The problem with AI in those games is not so much that it is a very hard problem, which it is, but rather that people who write those games seem more concerned with the graphics aspect of the games, and the games are designed with human play in mind first, without any thought put into how the computer is going to play succesfully. I think that with more thought and effort, you get an AI that can handle more situations better. If Starcraft is infinitely more complex than chess, then so is football. I think that in some senses, apples and oranges are being compared here. >Now games like EQ and AO make Starcraft look like a tinker toy. Again the world >is MUCH bigger then in Starcraft and the units are much more complex and the >enviroment is much much more complex. I don't know what EQ or AO are. >Whats the point? Its not gonna happen until we get something like HAL, and HAL >isnt even on the drawing board. Give it up. " The author demands too much. He's demanding much more generalized human-style intelligence. It's possible to be intelligent in a smaller domain. A program doesn't have to be able to write a sonnet before you can apply the term "intelligent" to it other than as marketing hype. bruce
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.