Author: jefkaan
Date: 01:29:36 11/20/02
Go up one level in this thread
On November 19, 2002 at 19:18:56, Bob Durrett wrote: >There are "tons" of complicated items of human knowledge. Much is captured in >large databases of master games. > now you are going back to chess; ok, so the question is, how complicated is the game of chess; obviously its a lot more complicated than tictactoe, or even checkers or so. On the other hand, it's not as complicated as for example the works of Shakespeare; you can store Hamlet in bits and bytes on whatever computer, but there is no computer program which can understand the play. The same still holds for storing GM games up to a certain extent; certain patterns, eg. a fortress are easy to see for a GM, but not at all by a computer program. So what i meant was that we need software 'tools' at a higher level of abstraction. Maybe in some hierarchical way, informationwise, i mean. There are a lot of simple tools in computer chess, see eg. the tools of Paul Onstadt (seems to have deceased). More complicated tools are Chessbase and Chess Assistant. You can search for certain patterns in such programs (probably Chessbase is better in that, but for the opening book theory i prefer CA); now the question is, how the program an engine which can look forward into certain patterns, simple example, recognizing the difference and importance between equal and unequal bishop pairs in the middlegame, depending on other game characteristics. Vincent could tell you a lot about that, but he doesn't :) However you can also read about chess theory. Eg. modern chess strategy by J.Watson. Now such ideas are difficult to absorb in the first place (it takes learning, practising to see what's most important, etc.); a team of GM consultants in theory could help maybe, although digesting the best ideas from their different styles might be difficult, but then still such ideas are even harder to program in assembler, C_sharp, bitboards, or whatever. In fact, we would need higher level programming languages for such purposes; that's what i was meaning. For a relatively simple game as chess building an expert system might turn out 'easy', after 30 yrs or so, when 99.9999% of the games will end up as a draw; for Go, it's already a lot more difficult as a result of a/o pattern recognition and strategy; i suspect such issues are the underlying reason for the different views of Vincent D. and mr Hyatt. Only when programmers will be able to make a program which can translate Ulysses by James Joyce from English to Chinese they also will be able to beat the best human Go players. This imho will take a lot of time, decades to beat the best Go players, and hundreds of years to translate literature. Making a selfconcious computer, which can learn by itself, like some sort of SF 'android' will take even longer; at least thousands of years or so i presume (but i'm always willing to bet about this for a couple of beers, you know :) best regards, jef
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.