Author: Dave Gomboc
Date: 11:42:28 08/23/99
Go up one level in this thread
On August 23, 1999 at 11:11:35, KarinsDad wrote: >Thorsten, > >My example yesterday was poor. Let me give you some other examples of a computer >emulating mistaken human behavior where it would be considered a bug UNLESS it >is a special mode of play which is not the default: > >1) The program does not accurately keep track of the moves. Humans do this ALL >of the time (I forget to write down my opponent's move and my move about once >per tournament and get my sheet off by a move). If the program does this, then >it might "rush" it's moves in order to get 41 moves (40 moves to the program). >Or, the program might claim a win thinking the opponent only made 39 moves by >time control (although the opponent had actually made all 40 moves). > >2) The program could claim a draw by repetition when the position only repeated >twice. > >As you can see, there are a LOT of human mistakes that could be emulated by a >computer program. And, there is nothing wrong with emulating those mistakes, as >long as this is not the default mode of play and the customer can turn these >features on and off. > >However, when you have a human-like mistake which is not expected and is the >default behavior with no way for the customer to disable the mistake, then it is >a bug. > >Your claim is that humans go over time control all of the time, so programs can >do it to. Nobody is arguing that the program is not ALLOWED to do that. We are >arguing that it is STUPID for a program to do that since it does not need to >have the human frailties of poor time management. And when the customer does not >have the ability to turn this feature off, then it is a bug and not a feature. > >There is no difference between this and my two examples above. If you would >consider the two examples above to be bugs, then you should consider spending 4 >minutes on the 40th move in a won position and overstepping the time also as a >bug. > >You gave the example of a program which took time to make it's opening moves. >That is a fine feature as long as the customer can turn it off and on. > >The position you are taking is an arrogant one of: > >We are the programmers. We are god. We know what is right. KarinsDad has not >even played the program, so he is "prejudiced and talking nonsense". He "does >not know what a bug is". How much more arrogant can you get Thorsten? You think >that since you are an intelligent programmer that nobody else has a mind and is >capable of rational thought? You think that one must PLAY CSTal in order to be >able to distinguish between a bug and a feature? You think that people such as >myself that have been in the computer industry for 20 years and have been >playing chess for 30 years are incapable of knowing a bug in a chess program >when they read about it? > >Sorry to disappoint you pal, but I am not the only one with this position on >this forum. And, some of the other people who agree with me HAVE purchased >CSTal. And, some of the other people who agree with me HAVE been writing chess >programs for decades. > >What you are describing COULD be a feature IF a person playing the program could >turn it off and on. But without that capability, it's just another bug. We ALL >understand your position of human-like play, we just do not understand you >position of human-like mistakes having to be forced down the throat of your >customers. It's unfortunate that you are incapable of understanding our >position. > >KarinsDad :) It's not human-like at all. Humans that chronically get into time-trouble often take way too long early on, and end up playing their last fifteen moves in ten minutes or something. Humans tend to lose on time when they are in complicated positions (or simply busted.) If they are completely winning, they'll just play a series of moves that seem to hold the fort together until they make time control -- maybe not very good moves, but that's what time pressure is all about. CSTal isn't modeling these behaviors at all. If CSTal was using time in such a fashion, I could believe that it's a "feature". However, we know this isn't the case, and furthermore, there is a well-known programming flaw that explains the behavior that is seen. Thorsten's claim is completely bogus. Dave
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.