Author: Fernando Villegas
Date: 09:36:35 05/22/98
Go up one level in this thread
On May 22, 1998 at 12:24:49, blass uri wrote: > >On May 22, 1998 at 10:22:26, Mark Young wrote: > > > >>I don't know what other people have said in the past. I just think the >>skill part is making a top program with less knowledege. The programmers >>need to know what kind of Knowledege >>they need to give the program. Then letting the program get much of what >>it needs positionally from its deeper search. If this is done >>incorrectly the program will play awful chess. Yet these three >>programmers have the best programs out. To me, it would seems much >>easyer to just add Knowledege. Not take it away. > >I believe the important thing is to add knowledge. >there are positions all the programs I know do not understand >in my nunn match I saw that fritz5, Junior4.6 and all the programs I >know >do not know to evaluate danger to the king and sometimes the programs >evaluate one side win for many moves when the result is the opposite >or draw. >in the 5th game both programs evaluated the position as more than >3 pawns for Junior(and a big part of the game as more than 2 pawns for >Junior) > and the game was drawn! > >other kind of important knowledge is what to search >and I see that programs are very weak in this knowledge. >for example no program I know discover that the position kasparov >resigned >against deeper blue is a draw. >an interesting question is what program is less weak than others >in the knowledge of what to search. > >Uri Dear Uri: This discussion is becoming a kind of example of logic falacy. The issue is not to state or to deny if programs have or have not a full positional understanding , but to discuss if positional understanding is or can be solve with other ways that the usually asked "knowledge". In other words, the holes you ha spotted so well are one thing and to think that those holes coul be filled with "konowledge" is another matter. Not necesarily one is the product of the lack of the other. In my view the esential factor here is the quality of the search algorythm and this one is not, as you know, just a question of filling the source code with tabels, functions, etc. In this, like in any other things, the point is the mix and the best mix not ever is the greatest mix. Let me give you an example: all the history of technology show that a the greatest steps ahead were produce when some very comlicated and improved technology gave way to another far simpler but superior in his use of energy. Never steam locomotives were more stupnednour work of engineering that the day previous to the arrival of the very simple electric motor. Maybe I am wrong, but some lesson could be drawn from that in the field of chess computers. Shalom Fernando
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.