Author: Dave Gomboc
Date: 17:15:16 07/04/02
Go up one level in this thread
>Instead of doubling the number of processors for example, they could have >designed a better search. This would have been much more efficient. > >There are two possibilities: > >1) either it was useless, given the NPS they already had, to try to search any >deeper, and it was better to invest time exclusively on the evaluation function. >In this case, why did they invest any effort in doubling the number of >processors? NOTE: doubling the number of processors, given the big number they >already had and the poor scalability of parallel search, would not even give a >1/2 ply improvement. Maybe not even 1/4 ply. This is the correct explanation. The chess chips were redesigned to allow for a better evaluation function. All of their work in the time between the first and second matches was to improve position evaluation. That the number of chips were doubled was a fringe benefit, and if they would have had to invest significant effort to do this, then they wouldn't have bothered. Computer chess software developers sometimes forget that the purpose of Deep(er) Blue was not to win computer-vs.-computer matches, but to beat the human world champion. Sure, the search could have been made more efficient, but it was outsearching Kasparov anyway, so why bother with it? Assessing a position as well as Kasparov did, that was the problem they tried to solve between the matches. And it is clear that they had some success. ;-) Dave
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.