Author: Uri Blass
Date: 04:13:38 02/19/03
Go up one level in this thread
On February 19, 2003 at 06:31:34, Rolf Tueschen wrote: >On February 18, 2003 at 21:34:44, Robert Hyatt wrote: > >>On February 18, 2003 at 18:16:10, Rolf Tueschen wrote: >> >>>I have made a white paper for a simple question: what is your opinion about >>>Crafty against humans in tournatment mode? I had a littler dispute with >>>Matthew... I said you never claimed GM status for either Crafty or other >>>commercials. Then Matthew said Crafty drew to some unknown (to me!) GM. Does >>>that mean you now say that also Crafty "is" GM in long tournament mode? >>> >>> >> >>First, a clarification. I have _never_ said that Crafty is a GM-level >>player at tournament time controls, namely 40 moves in 2 hours. I believe >>it is clearly a super-GM at blitz (game in 5 minutes). In fact, it is >>most likely a "ultimate-GM" if the time control is really game in 5 minutes, >>period. > >First of all thanks. Amir might respond for the superiority of DJ. > >Then I want to make a little experiment. > >Challenge and Refutation: > >(1) Take GM Roman in a 6 game Blitz with 5' for the game and give Roman 1 >million US$$ in advance cash. If he wins it's all his money. A draw or loss and >he will have to give back the whole money. I bet that Roman will win that >against any Crafty available. Period. :) I think that it will be but unfortunately we need some sponsors for the challange and I do not think that you will get a sponsor that will give 1.000,000 dollars If there is a sponsor who can give 1000$ then I suggest the following idea Every player can play Crafty 6 game blitz match but need to pay 10$ for the match. In case of winning the player win all the money 1000+10x$ when x is the number of matches that crafty played. In case of losing the player lose 10$. Note that if there are players who believe they have 30% chance to win the match then it means that they can probably win the 1000$ by playing some matches when in the first matches they will lose 10$ but in the last match they will earn them back together with another 1000$. > >(2) We are making too many wild guessings without exact analyses. In general >people have little experience in psychological experimental design. Again, that >was very clear - for me at least - in 1997 already. The factor 'human' is out of >sight when you chess programmers do your work. That is why Amir could make his >incredible statements. If already a machine is inpredictable in its behavior >[parallelism etc.] then human beings could never be out-calculated. In the final >moment they still could decide for the other option. - I say that you have lots >of data about humans playing comps but in general you know zero about the human >players' motivation. We simply have too little data. So I prefer the theory that >super GM still are stronger than machines also at Blitz. > > > > >> >>Second, I've come to modify my "GM level" a bit. I now suspect that in >>tournaments/matches that span multiple days, that most computer programs can >>play at a modest GM level. Not because they are positional GM players most >>of the time, and not because they are tactically superior to most GM players, >>but because of their consistency and stamina. They don't get tired and will >>play the last game of a consecutive 100 game match just as well as they play >>the first, perhaps even better when you factor in "book learning." Humans >>can't do that, as they get tired. >> >>So we have yet another "superiority" that the machines hold over us. Better >>memory than most (most but maybe not all). No fatigue. No distractability. >>Not subject to illnesses/bad days/etc. Perhaps those things _do_ make up for >>the positional/tactical shortcomings programs exhibit. > > >Objection! > >For the reasons given above you simply can't say that. We have too little data. >Kasparov and the couple of other GM who drew comps in show events are not data >enough. If you buy this fatigue story then you are a very prominent victim of >the propaganda spread around the event in NY. You know how Kasparov is. Of >course he seeks easy excuses for his blunders. And of course Amir tells us that >he has not seen a single blunder in the games. But this is all propaganda. I also do not believe the fatigue story but I also do not believe that kasparov lost on purpose. Note that Kasparov lost the first game against DB96 but fatigue was no problem. I believe that humans are going to learn from computer to be better players and I am not talking only about tactics so I believe that in the next few years we are going to see again close matches between humans and computers. 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.