Author: Peter Fendrich
Date: 13:39:36 04/16/99
Go up one level in this thread
On April 16, 1999 at 14:50:02, KarinsDad wrote: >On April 15, 1999 at 18:33:11, Peter Fendrich wrote: > >>At the time I worked with Knowledge based systems (or programs) we used >>something like this as a definition: >> >>A computer system that imitates human problem-solving by using a "knowledge >>database". >>The knowledge database is a set of "knowledge chunks" expressed using some >>formal knowledge representation language. >> >>This or some similar defintition is commonly used within the AI community. At >>least it was some years ago. >> >>The idea is to capture the knowledge from experts, express it in the knowledge >>represenatation language and put in the knowledge database. The knowledge >>database was later searched in runtime by the program to find out what to do. >> >>I know of no chess program that uses this approach and that plays descent chess. >> >>IMHO, assuming that the definition still holds, it is not a good idea to use a >>different definition for knowledge based chess programs visavi knowledge based >>programs in general. >> >>//Peter > >That's hard to say. If you use a pre-processor on the current position to >determine if there are certain squares where you may want to place some of your >pieces and certain squares where you do not want your opponent to place certain >of his pieces, and you change your sort order based on that, it could be >considered that this would be a program with a knowledge base (it ignores what >it considers bad moves and searches what it considers good moves more thoroughly >based on pattern recognition and "expert" data). However, it may not be touching >a knowledge database at all. > >If you use an opening book (a knowledge database created based on expert >playing) and use it to find out what to do, I would not consider that as >creating a knowledge based system (with the possible caveat that it is an >opening knowledge based system), but from your definition, it would have to be >considered so. Well, knowledge based programs doesn't have to be smart by themself - the knowledge is outside the program. There is a great deal of knowledge collected during centuries, concentrated in the opening theory behind the opening book moves. Opening books and EGTB's could be regarded as a knowledge databases but they are not very typical and I agree that they do not justify the program to be called a knowledge based system/program. Just because of the small increase of playing strength yielded by these things. A bad engine isn't much helped by books and EGTB's. >The problem lies in whether you consider only some of the moves being acquired >from a knowledge database (such as opening books and tablebases) as indicating >that the program is knowledge based, or if all moves must use knowledge (which >again they obviously do to some level due to the evaluation routines). Hence, it >would seem that ALL chess playing programs are knowledge based. There is just a >difference on how much or how well they use that knowledge. And the rest is >probably just semantics. Yes, of course all programs (chess or not) uses knowledge and are based on knowledge. But then the term loses its meaning. The problem seems to be the different meaning of "knowledge based" in general and when it is used in "knowledge based systems/programs". My point was that we shouldn't redefine a commonly used term. That just causes confusion - more confusion than there already is... I will call it a knowledge based program when Kasparov, Anand or some other chess guru sits down and really tries to specify their knowledge and some other guy is trying to represent that in a database used by the program. This is a project that would take a very long time and it would require full attention from the participants. ...and it would probably fail to beat the top programs anyway... :) > >KarinsDad :)
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.