Author: Rob
Date: 02:31:11 11/25/03
Hi, in the past the competition between man and machine is really mainly a competition between man and a team of man + machine. So as a true test of ai those matches have shortcomings. The main influence of the human team is in opening preparation for the opponent and adjusting the book during the match. The latter was already adressed in the Kramnik match against Fritz where human intervention during the match (adjusting opening book) was not allowed. While a step in the right (ai wise) direction it is not sufficient. My proposal is that PRE-MATCH preparation should also be done by the machine alone. Rules could be as follows: 1) The machine has no opening book per se 2) the machine has a database of all games ever played which is unqualified by human interpretation (In Chessbase lingo: NO MEGABASE, which has games selected by humans but a GENERIC BASE like big base which contains everything from club level to world championship matches) 3) The machine is given the name of the match opponent which is also used in the database according to 2 4) The machine can then use the time up to the match to compute its opening book on its own - no human intervention (this could be lengthy process if the algorithm analyses a subset of games with the engine - so time management is important) 5) no human intervention during the match 7) Verifikation: source code for the match is given to a neutral party to verify and this version is used for the match. if pseudo random numbers are used the seed is to be given to this party to reproduce the decisions. I think Kramnik is to be thanked for thinking about this with his rule of no human intervention. This is just taking it one step further to make the man vs machine matches more interesting in the future (5 to 10 years) without having to handicap the machine. regards, Rob
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.