Author: Robert Hyatt
Date: 07:37:54 05/08/00
Go up one level in this thread
On May 08, 2000 at 08:35:08, Hans Gerber wrote: >On May 06, 2000 at 23:25:47, Robert Hyatt wrote: > >>But _none_ of this has anything to do with being able to detect or prove the >>absence or presence of cheating. It simply can not be done. You give me an >>experimental set-up that you like. I'll explain how I will still cheat without >>your being able to detect it. >> > >Just to remind the group: we were talking about the DB vs Kasparov match in >1997. > >Isn't it obvious that the scientists should have granted Kasparov a look into >the prints of the logfile if the proof that nothing had happened that could be >called cheating? Absolutely not. First, it would give Kasparov a chance to see how deep the thing searches, how it extends. How it evaluates some positional considerations. It would be a decided advantage for Kasparov had he had this kind of information. Injecting it into the middle of the match would have definitely tainted the results. Second, the logs could have been 'doctored' as the game was played. A human overrode a move. He could edit the log immediately and make it appear that the machine liked the move. > >My "set-up"-hint? > >To find right from the beginning of your research how to control and thus assure >objectivity of your results. > >Is it so difficult to understand? > >IMO the long tradition of computerchess events has shown that the actors, most >of them scientists, had a mutual agreement about the procedere. They were >scientists and behaved like that. It was more or less fun. > >Therefore they neglected the genuine question of control. Although the >possibilities for cheating were there. > >As we know there is a long tradition of so-called experimental matches between >strong computers and humans. The stronger the hardware became, the better the >machine could play. Even GM players had difficulties to get good results. Still >the whole events were more or less fun. > >Then came the close cooperation between the DB team and Kasparov. If we neglect >the question of money, it's still fun. Look at Kasparov's own comments on the >strength of DEEP BLUE after the first match. If DB had all the weaknesses >Kasparov detected it's completely nonsense to talk about the ultimate match to >decide the fight between The Machine and The Human Race ... > >You can't have "fun" and at the same time an event that should _prove_ who >really should be regarded as the strongest chessplayer of the world. > >The participation of scientists however seemed to guarantee that this proof >could be presented. > >Fact is that G. Kasparov had the impression that something wasn't kosher with >the second game. He thought that he could win some insight if he could see the >logfiles. They were promised then denied. So for Kasparov this was a support for >his suspicion. He could no longer concentrate on the match, his thoughts went >back to this game number two. The outcome of the match did no longer interest >him which can be seen especially in the 6th game. > >The proof that the machine could beat the human was not there. > >I have no idea how we could _control_ the output of the machine, but many might >be here who have ideas for a solution. I have been doing this almost forever. I have participated in numerous discussions on how to prevent cheating in comp vs comp matches. The bottom line is that it is impossible to do. > >Kasparov should not be regarded as an especially difficult human being. All >chessplayers hate being cheated. So, if we want proof if a machine could play >better chess than the best humans we should at _least_ guarantee that the >machine won the match on its own. Including match psychology and so on. > >Any ideas? Note that several people _did_ look over the output. Ken Thompson was one such person that was a friend to Kasparov as well. He said that everything looked perfectly normal. Kasparov would have no idea of whether the machine's output indicated cheating, or anything else. > > >Hans
This page took 0.04 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.