Author: Evgenii Manev
Date: 03:45:45 04/30/05
Go up one level in this thread
On April 30, 2005 at 04:43:48, Walter Faxon wrote: >(Since the Kasparov-DB threads are heading off the right end of the screen, I >thought I'd start anew.) > > >Dear Dr. Tueschen, > >Since the areas of science and justice are clearly important to you, please >enlighten this forum as to exactly what changes you would like to see made in >the rules for future human-computer public chess matches. > >The rules used for matches to date are obviously flawed since they make all >sorts of false assumptions about computer chess-players and the human reactions >to them. > >Some suggestions for your consideration: > >- The game scores of at least 40 games played by the current version of the >program and hardware, against strong international opposition, must be made >available to the human player at least six weeks prior to the start of the >match. > >- If the computer plays any private training games within six weeks of the start >of the match, the game scores of these games must be made available to the human >player in a timely manner after each such game is played. > >- The human player must be allowed to practice with the current version of the >program and hardware, starting at least four weeks prior to the start of the >match. The computer will not be allowed to "learn" from this practice. > >- There will be no changes to the computer's opening book allowed during the >match, except changes made by the computer itself. > >- There will be no program changes allowed between games. Hardware changes can >be made only to clearly inferior versions of the required hardware. > >- The human player will be allowed access during each game, to a separate >computer database of chess openings, as chosen by him or her. > >- The human player will be allowed access during each game, to a separate >computer database of endgame positions, as chosen by him or her. > >- It must be made easy for all inputs to and outputs from the program to be >monitored by agents of the human player during each game. This may require that >an agent of the human player be present at every computer control point, e.g., >terminals, access panels, etc. The sponsoring organization will pay the >salaries for all such agents. > >- To reduce the possibility of outside "interference" with the computer's play, >all the relevant computer hardware may be examined in detail by competent agents >of the human player, at any time during the course of the match. The sponsoring >organization will pay the salaries for all such agents. > >- A complete print-out of the computer's search will be made available to the >human player after each game. (I.e., the principal variation and score after >each search iteration, etc.) > >- The computer's estimate as to the value of the current position will be >provided as running commentary to the human player. > >- The computer's estimate as to the human's best move in the current position >will be provided as running commentary to the human player. > >- The computer's estimate as to the principal variation for the current position >will be provided as running commentary to the human player. > >- Computer memory and mass storage is not limited; however, long-term storage of >concrete chess positions (openings and endgames combined) will be strictly >limited to a maximum of 10,000 positions. The complete source and object code >for the chess-playing program and all its auxiliary maintenance programs and >files will be provided to the human player to allow verification of this. > >- Computer processing power is not limited; however, computer examination of >concrete chess positions during its search will be strictly limited to a maximum >of 1,000 positions per second (200-500 times the maximum for human >chess-players). The complete source and object code for the chess-playing >program and all its auxiliary maintenance programs and files will be provided to >the human player to allow verification of this. > >- If, during the course of the match, the human player comes to believe that the >current rules for the match are unjust or do not promote the advancement of >science, and that a particular change will improve the situation, he or she can >unilaterally amend the match rules accordingly. > >- It is a responsibility during the match, of both the team supporting the >chess-playing computer and the match sponsors, to anticipate possible problems >that may be encountered by the human player, and to suggest to the human player >amendments to the match rules that might alleviate or prevent said problems. > >- The human player will be allowed extra "off-clock" time in difficult >positions. > >- The human will be allowed to adjourn any game at any point, to be continued on >the next playing day. Each game may be adjourned multiple times. > >- The human player will be allowed to take back one move in each game, before >the computer has made a move in reply. Clocks will be reset to the times >elapsed at the start of the human player's turn to move. > >- The human player will be allowed to restore the game to any prior occurring >position, once in each game. Clocks will be reset. > >- If the human player resigns in a position that is subsequently discovered to >be drawn, the game will be re-scored as a draw. > >- Any games where the human player believes that he or she played "below normal >strength" will be annulled and replayed. > >- Any public insinuation by the human player to the effect that the computer's >play was the result of cheating, will be accepted without question and with >immediate apologies. The human player will be declared the winner of the match >and any purse distributed accordingly. All so-called computer scientists >associated with the chess-playing computer will leave the field in disgrace. > > >I invite CCC readers to suggest additional match rules. > > >Dr. Tueschen, lest you think that some of these suggestions are serious and some >are jokes, let me assure you: All are jokes. > >-- Walter hi Walter all is needed in human-computer chess matches is as follow: 1.computers do not have access to any opening library accept one made from its previous games 2.computers do not have access to endgame tablebases the idea is that computers have to learn how to play some parts of the game, not just repeat already found best moves by memory; progammers tend to disregard ability of their creations to understand some aspects of chess game because there is an easy way - opening libraries and endgame tablebases - may be its too hard to 'teach' chess engine how to play openings and endgames, or how to play 'positionally' but IMO thats the way programmers should go forward Position and Book learning is all that is needed for chess program, i think. Everything else lays on the computer speed power. of course, for analysis puproses its very good to have opening library and access to huge endgame database but when the matter is to see who is stronger - a human or computer - such things that gave advantages to the one side, should be switched off regards, Geno
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.