Author: Fernando Villegas
Date: 19:30:01 02/25/98
Go up one level in this thread
On February 25, 1998 at 00:21:52, Robert Hyatt wrote: >On February 24, 1998 at 22:29:46, Fernando Villegas wrote: > >>Hi all after my long vacations: >>This F5 rating has arisen once again the old, somewhat absurd discusion >>about knowledgeable Vs fast searchers programs. I say absurd because, >>as ever, discusion is going on as if "knowledge" as such was already a >>defined, decided and unquestionable thing. But is not. If for knowledge >>we call the somewhat unorganized sum of practical chess signals human >>experience has got trought centuries of playing the game, OK, but then >>remember that after centuries of common sense and experience people was >>certain the Earth was a flat surface. That seemed undisputable >>knowledege to date. Before and elsewhere, other thought the world was >>supported by a number of elephants, the elephants by turtles and the >>turtles...well, as they said, don't bother with silly question. If you >>take a look at almost any specifi bit of chess "knowledge", you will see >>that does not amount to much more than to a general, umprecise statement >>based in a determinate amount of specific, local, narrow experience, and >>that's the reason real games offer at each moment - more often that >>presumed- clear signs of exceptions. What kind of "knowldege" is that >>that can be refuted once in two or three tries? Take a llok at any >>theroic book and you will see that at each step. "Yes, knights are good >>here, but then"... At leat defenders of knowledge approach should do >>something better than to trash "fast stupid searchers" and give us a >>more detailed and deep analysis of what is that knowledge they defend so >>much. Have they a more specific slice of principles we don't know yet >>and that are the "truly knowledge"?. My impression is: a lot of it is >>trash itself and that`s the reason a fast progran not too much loaded >>with knowledge can give a very good perfomance. Another impression: for >>the kind of searching abilities computer have, knowledge probably means >>or should mean something very different to our clumsy road signals. >>Fernando > > >I think "knowledge" here means evaluation code. IE a program that >recognizes >more "features" in a game has more "knowledge". Whether this is >analogous >to human knowledge has nothing to do with anything. Well Bob, with due respect to your genius as a chess programer and a great personality in this field, here I differ from you quite strongly. Until now, before aliens decide to show his faces from his flying saucers, Knowledge means something only if means what you call human knowledge. And certainly so it is in chess programing. Code lines are done for mimicking human knowledge. Surely to recognize features, as you call it, has to do with using the known chess human techniques and modes to play the game, by example, what to do with the king againts a pawn in the h colum and so on. So all this has to do strongly with human perception of what it is chess knowledge. My point is if what we NOW call chess knowledge if enough a real knowledege to try so intensely to put into engines. I think is not. I think is a heap of recipes that sometimes do the job and sometimes do not. That's why most of chess programming has something very similar to alchemist practice, an issue of mixing things and then to see what happens. A pleasure to talk with you, Bob, and awaiting a full, complete Crafty version in CD. fernando The question is, >does >fritz rely more on search, or more on evaluation. Ditto for a program >like >Rebel... > >that was the point...
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.