Author: Amir Ban
Date: 10:07:03 01/20/00
Go up one level in this thread
On January 20, 2000 at 08:34:21, Graham Laight wrote: >Suppose that IBM could not get DB to work, so they decided to put a humming >black box in the Equitable building, and transmit the moves in from outside. The >moves would have been chosen by a team of IMs working with other chess >computers. > >Would the match referees have picked this up? > >As I understand it, there was only 1 referee, and he couldn't have been >everywhere all the time. Was there even a referee in the IBM team room? > >Let me state for the record that I do not believe that, in reality, IBM cheated. > >-g > Ken Thompson was the observer at this match. He never made any comment on this that I was aware of, and I found that odd. So, a year ago, I wrote him and asked what his view of this was, plus some specific questions that I had on the printouts. I got this as reply: amir, my job at the tournament was to be a neutral observer. as it turned out, i represented gk more than ibm. for example, i had access to gk's private dressing room behind the stage, but never exercised it. i spent every game in the control room talking to the ibm operators - all of the ones not on camera -- and watching the multi-camera views of the playing room. i saw the live printout of the screen of every move of every game. i asked many questions and had forthcoming answers. i was quite familiar with the printout and operation of db over the years. it really hasnt changed much from cmu days. i had long technical discussions with the programmers. in my opinion - based on a large number of inputs - there was absolutely no underhanded dealings with the conduct of the computer. i got the printouts to game 2 sometime near game 4. i was disappointed about the delay by ibm in producing the printouts, but they were as i remembered them. i studied them for a day before i replied to gk. i then studied them for a total of several days after that. i reported to gk that i thought the evaluation and intention for 35 Bd6 was clear. the evaluation for 36 ab5 shows a clear horizon effect - delaying something bad by throwing in an exchange. the evaluation for 37 Be4 shows an alpha cutoff and a scramble for an alternate move. the pattern is quite consistent. i had seen similar patterns in nearly every game. the magnitude of the values were not unexpected. ken Since this reply didn't answer some of my questions on the printout, I tried a follow up question, and ultimately got this: >> What about my other point ? i'm not going to comment on ibm's printout. it is not something i would do. i've told you my conclusions. ken I have immense respect for Ken and I think his conclusion as he stated it carries a lot of weight. Still, I was disappointed by some things in his answer. It seems he looked at the printout from the purely technical point of view, and didn't really get into any problem related to the position on the board. For example, his conclusion that axb5 is played due to the horizon effect, delaying Qb6 for one move, does not agree well with the position or with the printout analysis (and is not flattering to DB either). He says about evaluations that their magnitudes were not unexpected, and that I don't agree with either. I don't understand what he means by his comment on move 35, because IBM did not deliver printouts for this move. Most of all, he didn't address the question that Kasaprov was really asking, of why DB passed on winning a pawn in move 36. It would have been natural to ask the DB team for explanations. Maybe he did that, but he doesn't tell it. I don't know why he feels bound not to discuss the printouts. Amir
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.