Author: Robert Hyatt
Date: 20:32:52 01/11/03
Go up one level in this thread
On January 11, 2003 at 10:30:44, Frank Phillips wrote: >On January 11, 2003 at 09:39:41, Uri Blass wrote: > >>On January 11, 2003 at 09:12:20, Vincent Diepeveen wrote: >> >>>On January 11, 2003 at 08:51:18, Frank Phillips wrote: >>> >>>>On January 11, 2003 at 08:38:50, Uri Blass wrote: >>>> >>>>>On January 11, 2003 at 08:23:43, Frank Phillips wrote: >>>>> >>>>>>"The next question is, and many people are asking it, do we know how Deep Junior >>>>>>compares in strength with Deep Blue? The really interesting thing, from the AI >>>>>>point of view in general and for computer chess researchers in particular, is >>>>>>that Deep Junior examines something like one percent of the number of positions >>>>>>per second of Deep Blue. But despite this Deep Junior may well play better chess >>>>>>because its "understanding" of the game is better. It appears to have more chess >>>>>>knowledge and understanding in its evaluation function than Deep Blue did, and >>>>>>this compensates for the difference in positions-per-second.." Extract from >>>>>>Levy on Chessbase.com site >>>>>> >>>>>> >>>>>>From what I read in Behind Deep Blue I find this surprising. But then again, I >>>>>>no nothing about Junior other than it is an awesome program. >>>>> >>>>>I think that comparison between the quantity of evaluation is meaningless. >>>>>The right comparison is comparison of the quality. >>>>> >>>>>It is easy to add a lot of knowledge without testing for bugs but the result >>>>>will be a disaster. >>>>> >>>>>It is possible that the program that has more knowledge and understanding simply >>>>>understands things wrong because of bugs or understand the wrong things. >>>>> >>>>>I think that discussion about Deep blue's evaluation is meaningless unless Deep >>>>>blue team post the source code of their evaluation. >>>>> >>>>>people simply are not going to agree. >>>>> >>>>>If deeper blue post the source code of their evaluation then it will be possible >>>>>to compare it by changing the source code of the free programs to have the same >>>>>evaluation as deeper blue's evaluation and use games with fixed number of nodes >>>>>per move. >>>>> >>>>>Uri >>>> >>>> >>>>I agree. I do not understand therefore how Levy came to his conculsion. >>>> >>>>However, if I had to guess I would say that Deep Blue has more knowledge (it was >>>> free in terms of calculation) and much of it seems to have been tuned by >>>>Grandmasters. Hsu talks about Deep Blue going to Joel Benjamins chess school. >>>>This is not necessarily the same as understanding (strange term), but I know >>>>where I would put my money. >>>> >>>>I find your insistance that only Deep Blue publish their code, biased. It would >>>>be interesting (although commercial sucide) to compare both evaluation >>>>approaches against each other and in the way you describe. Levy offers no data >>>>to backup his claim. Deep Blue not demonstrating it is better by publishing its >>>>code, does not make Junior better. >>>> >>>>Frank >>> >>>Look to the many idiotic moves Deep Blue played and try all those positions on >>>Junior. >>> >>>Then you'll know enough. >>> >>>Easy test nah? >>> >>>Levy as an IM knows more than enough from chess to judge the quality of the >>>moves made by Deep Blue. >> >>I also guess that Deep Junior has better evaluation but >>note that David Levy lost against Deep thought 4-0 in 1989 so I do not think >>that the fact that he is an IM is a reason to support my opinion. >> >>I see nothing productive from another disagreement with Bob Hyatt about the same >>thing and this is the reason that I prefered not to express an opinion in the >>first post. >> >>I do not see what is the commercial interest of Hsu and IBM in deeper blue's >>evaluation so I do not see a logical reason for them not to post their >>evaluation. >> >>In the case of Junior there is a clear reason not to give the evaluation of >>Junior because it is a commercial program. >> >>This is the reason that I suggested that deeper blue post their evaluation code >>including the bugs that they had so we can compare it with today's programs. >> >>Uri > >Accepted. > > >I was not aware that there had been big progress in 'the static evaluation of >chess positions' since Deep Blue, but maybe Vincent will post some examples to >illustrate his point. > >Frank I doubt it. But his hand-waving will no doubt stir up the equivalent of a hurricane... I'm sure there have been _some_ developments since 1997. But Vincent seems to assume that because his evaluation was crap, that everyone's eval was crap. Even though he knows that much of what I do today I did back in 1994. That includes passed pawns, outside passed pawns, pawn majorities, distant majorities, weak pawns, etc. I do a few things today I didn't back then, I did a few things back then I can't afford to do today on a non-vector machine. But the approach of "my code was crap so everybody's code was crap" simply doesn't fly. I'm waiting for him to agree to a match against a 1997 program on 1997 hardware since he can beat them so easily... I can supply both the program _and_ the 1997-era hardware. And another ridiculous statement will simply go down in flames. Again. As usual.
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.