Author: Robert Hyatt
Date: 12:26:49 08/29/01
Go up one level in this thread
On August 29, 2001 at 14:52:31, Joshua Lee wrote: >On August 29, 2001 at 14:07:20, Robert Hyatt wrote: > >>On August 29, 2001 at 13:59:01, Joshua Lee wrote: >> >>>>Unfortunately I found only one volunteer to help me to >>>>analyze the positions that deeper blue pondered and >>>>I am not going to use more than 20 hours of computer time >>>>per week to analyze. >>>> >>>>Uri >>> >>>I would be glad to help as i said before but there is one problem, I know that i >>>have seen the analysis of different positions in which a commercial will find >>>the move played however we would have to go through the log files alot closer >>>than you think as some moves are given in the mainline of others. >>>For example i posted recently about Deeper Blue playing h5 when it saw this at >>>11/6 from 3 or 4 moves earlier so this will probably show to require much more >>>time than we think...........But when you think about the draw missed by >>>Kasparov this should've come up in the pv from earlier moves but i didn't see >>>it, It is supposed to be too deep for DB so when micros find this move can they >>>be picking it for the wrong reasons ? If not then clearly Deep Blue searched >>>positions differently and better, for example several times a top commercial >>>will find the correct move in the pv for test sets yet change it's mind. >>> >>>If we are going to do this it should be under strict rules agreed apon by people >>>who think deeper blue is stronger and those that do not. I don't know if >>>Prof. Hyatt would participate or not but i think if he would this would bring >>>closeure for the time being ...atleast untill either the DF-Kramnik match and or >>>DB being ressurected to play weather that be against DF or any other program. >> >> >>I think it is a pointless experiment. For those that don't want to accept DB >>for what it was, _nothing_ is going to change their mind. They will keep >>finding new excuses as to why some finding is in error. >> >>Any program that could produce a 2655 rating over 25 consecutive games to win >>the Fredkin GM prize, and then have a successor with 25X as many circuits and >>100X the speed, _must_ be a strong chess player. There is no way on earth I >>can see how it could be weaker than today's best programs. That statement >>boggles the mind, IMHO. > >I agree, but atleast doing this would give sceptics one less way to say DB is >weaker! >Ofcourse someone here must have connections to the team so i don't see why they >haven't tried to convince the DB team to run some public games against Deep >Fritz??? Also what about the DT code that's out there you don't have the >hardware but it's in "C" so i don't know why it hasn't been discussed here. >Just because you don't have the hardware doesn't mean that a glimpse into what >the machine was thinking can't help. > >Or am i missing something. Thankyou > >BTW how about those Cray Blitz-Crafty games already :) Or maybe if you would >explain how different the pv is between the two and how much ply crafty needed >to win or draw instead of lose. And ofcourse what kind of hardware Crafty needs >to compete equally with CB on it's cray. The problem with cray has _always_ been getting information "back". IE I ran the games inside a facility that is very security-conscious. They greatly limit what gets in there, and they allow practically nothing back out. I didn't give any thought to this because I was simply curious about how the current 1M NPS crafty would compare against a program with a significantly different search approach. I should have tried to have someone up there ship me the output on a tape or CD, but I didn't think about it. By the time I went back up there, the files were long since cleaned out... I was hoping I might have captured the PGN files if nothing else, but I can't find anything here although I am not 100% convinced I have searched _everywhere_ yet. When you have a hundred gigs of stuff sitting around, it is easy to lose a dozen PGN games. :)
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.