Author: blass uri
Date: 15:58:42 09/11/98
Go up one level in this thread
On September 11, 1998 at 17:58:02, Serge Desmarais wrote: >On September 11, 1998 at 03:02:18, blass uri wrote: > >> >>On September 11, 1998 at 01:23:39, Serge Desmarais wrote: >> >>>On September 11, 1998 at 00:01:10, Roy Brunjes wrote: >>> >>>> >>>>I think the subject says it all. Of course this is a huge matter of opinion, >>>>but if a large group like this comes close to a consensus, then I'll consider >>>>the input valuable. To me, human play contains more subtle, >>>>positional/strategic stuff as well as speculative sacrifices (fairly rare for >>>>programs I gather - though some claim Hiarcs 6 does spec sacs). >>>> >>>>Thanks! >>>> >>>>Roy >>> >>> >>> What does it mean "playing like a human"? If you look at the games of a >>>Petrosian and compare them with those of Tal, Fischer, Kasparov etc. you will >>>find that ONE SINGLE human style doesn't exist. Your question seems to imply >>>that humans play chess in a way, while computers/programs play another way. In >>>any given chess position, you do not have an infinite number of good/playable >>>moves, but still more than one (usually or at least often). Now, today's >>>programs do play moves that any human could/would play >> >>Sometimes computers play stupid moves that no human in the level of at least >>2000 elo could play. >>I cannot say it about grandmasters >> >>Uri > > Players rated above 2000 do make mistakes and blunders. Now, what is a >mistake, error or blunder, if not a "stupid move"? But I know what you mean >:`stupid moves made after an error in calculation (tactical error) aren't the >same as mistakes caused by a misunderstanding/misjudgement of a position. I meant mainly to endgames There are positions that I can find the right moves and every human with at least 2000 elo can find and some good programs or all the programs fail to find the right move one case is when they do not do a right evaluation of a pawn ending. another case is fortress positions. There are also some positions in the middle game when the computer is too greedy. But I >am persuaded that for an important percentage of games, showing these without >naming the players, it would be tough (even for "specialists") to tell what >players were computers and in what games. That would be a nice test to do! I >think you would/could be surprised by the results. I think that the program that can be most succesful in "cheating" humans to think it is the human has more human style. Uri
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.