Author: Dann Corbit
Date: 11:55:09 01/20/05
Go up one level in this thread
On January 20, 2005 at 14:09:32, Steve B wrote: >In his 1997 book"Beautiful Mates:Applying Principles of Beauty to Computer >Chess Heuristics",Ben Wallis attempts to program a computer to play chess more >like a "Human" > > >clearly the way current computer programs play is nothing like the way humans >play > >current day programs rely chiefly on brute speed and the ability to analyze and >evaluate millions of nodes per second > >while the algorithms employed to evaluate the positions are of course important >,the sheer computing power and speed of current day hardware is chiefly the >reason for the very high ratings achieved within the last decade > >an example of this are the last two Kasparov matches against Deep Blue >in the first match Deep Blue using 40 processors lost to Kasparov by a margin >of 2 points(although it did defeat him in the first game) >in 1997 Deep Blue Employed 512 processors and as we all know defeated Kasparov > >Defining "Human" like play is not so easy > >based upon a previous study by Margulies who assembled a panel of 30 players who >were rated over 2000(Elo), >several concepts of "human like" play were created >one such concept was "Beauty" > >shown this position : >[D] k7/P1r5/K7/3N4/8/8/8/8 w - - 0 1 > >90% of the rated experts preferred to deliver the mate by checking with the >Knight at B6 rather then taking the Rook on C7 > >the idea of "Using the least amount of Force "was then included in the concept >of "Beauty" Beauty is in the eye of the beholder. I would take the rook. Besides, all checkmates are beautiful, unless you are the one checkmated. In that case, they are very, very ugly. >Reuben Fine seemed to agree with this idea when he critiqued Margulies study > >the book goes on to establish several different ideas like this all in an >effort to describe or define human like play >clearly no computer today is programmed to evaluate a position with "Beauty" in >mind > >in the end Wallis programmed a computer to solve mating positions using several >"human like" algorithms such as ..deliver mate with least amount of force" >the program was then subjected to a series of mating positions and the results >were compared to see how closely it compared to the panel of experts >an argument can be made that Wallis program exhibits more human like play then >that of todays programs (or at least solved mates more like a human player) > >i do not know what became of his program or if even was ever released > >the funny thing about all this is.. >if i were White in this position and i was playing a rated tournament game..i >would snatch that rook off with my knight and slam it down on c7 with as much >fanfare as possible..not forgetting of course to bark out...MATE! > >now i am certainly no computer..but then again..i am no rated expert either >:)) > >Best >Steve
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.