Author: Robert Hyatt
Date: 08:35:42 11/19/00
Go up one level in this thread
On November 19, 2000 at 11:08:53, Bob Durrett wrote: >On November 19, 2000 at 10:00:07, Robert Hyatt wrote: > >>On November 19, 2000 at 00:13:53, Pete Galati wrote: >> >>>On November 18, 2000 at 23:57:40, Robert Hyatt wrote: >>> >>>>On November 18, 2000 at 22:47:33, Bruce Moreland wrote: >>>> >>>>>On November 18, 2000 at 21:01:56, Robert Hyatt wrote: >>>>> >>>>>>On November 18, 2000 at 17:30:37, Garry Evans wrote: >>>>>> >>>>>>>On November 18, 2000 at 16:54:01, Robert Hyatt wrote: >>>>>>> >>>>>>>>On November 18, 2000 at 15:45:32, walter irvin wrote: >>>>>>>> >>>>>>>>>if you had a cray super computer in your house , so your time would not be >>>>>>>>>limited .what kind of chess program would you create?? another Cray Blitz or >>>>>>>>>would it be more like crafty .Also if you played it on ICC what blitz rating >>>>>>>>>would you expect that you would get ? how many nps do you imagine you would get >>>>>>>>>if it were the 32 processor Cray ??? I think it would be about as interesting as >>>>>>>>>the second comming of deep blue .Also what do you think your chances to be world >>>>>>>>>champion with that beast be?? >>>>>>>> >>>>>>>> >>>>>>>>The program would be a hybrid approach. To run fast on the Cray requires >>>>>>>>a lot of vectorization throughout the program. Bitmaps are cute, but they >>>>>>>>don't run fast on a Cray without some vector stuff as well. >>>>>>>> >>>>>>>>NPS? CB could do 7M. Crafty should be able to do the same. It would >>>>>>>>be tough... but it wouldn't be Deep Blue by any stretch. >>>>>>> >>>>>>> I think Deep Jr. and Deep Fritz are performing nearly as well as Deep blue >>>>>>>has. NPS is not everything, I think the Software is just as important? Am I >>>>>>>wrong? >>>>>> >>>>>>Short answer: yes. :) >>>>>> >>>>>>DB had both good software _and_ good hardware, contrary to what some would >>>>>>have you believe. >>>>> >>>>>Whether or not the software was good is a point of religion, and regarding this >>>>>issue, I am agnostic in the strictest sense. What you say may be true. It may >>>>>be false. I don't think it's possible to know. >>>>> >>>>>bruce >>>>> >>>> >>>>If you know the guys responsible, then you _know_ they didn't do anything >>>>"half-way"... >>>> >>> >>>Considering the brutal amount of computer power behind the Deep Blue program, >>>could they have acomplished basically the same thing _without_ a good programs? >>>Would the hardware have gotten around "half-way" efforts in programming, or >>>would the whole event have fallen on it's face with a "half-way" programming >>>effort. >>> >>>Pete >> >> >>If you follow their evolution closely, the answer becomes apparent. When they >>first burst onto the scene in 1986, their hardware was quite fast, but their >>software was fairly poor. But over the next 10 years, there were advances >>on _both_ fronts. The main difference between the DB that lost to Kasparov >>and the DB that beat him was (a) new software and (b) new hardware to support >>the new software ideas. > >If I recall correctly, Kasparov was defeated because DB whipped him, in the >critical game of the match, while both DB and Kasparov were in their opening >books. Kasparov forgot a transposition and that's what killed him in that game. > So you need to add that the new DB had a better opening book, too, in addition >to the other differences. I don't think an opening book can compensate for, nor expect, the opponent choosing a risky line to try to 'trap' the computer. I think he underestimated the computer's ability to win a line that commercial programs had no chance in. It turned out to be a bad gamble. In any case, the DB guys published several papers from 1986-2000, describing what they were doing. They didn't do null-moves, but they did lots of other things like singular extensions and the like. They weren't exactly doing nothing, in terms of software... which was my point. Folklore would suggest that they did everthing based only on super hardware. However, folklore would be wrong, here.
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.