Author: Jesper Antonsson
Date: 17:11:57 06/06/01
Go up one level in this thread
On June 06, 2001 at 06:31:03, Gian-Carlo Pascutto wrote: >On June 06, 2001 at 04:22:34, Tanya Deborah wrote: > >>Too, i believe that Deep Fritz is a bit better, (running in 8 procesor machine) >>than Deep Blue 97 version...(the same that won against Kasparov) Is this >>right? > >The truth is the no-one _really_ knows, and everybody who posts >a claim here is just guessing. > >I do think it's reasonable to say (i.e. I'm just guessing too) >that Deep Blue had more knowledge, although that of Fritz is >probably better tuned. Tactically DB was extremely sound because >they did no pruning, but Fritz will be outsearching it in raw plies, >but those are riskier. > >The conclusion of that is that well, I don't know. And neither >does anyone else. Well, IBM had a a couple of really bright academics that worked many years full time on the problem, with additional support from IBM staff and a good grandmaster to do opening preparation. They could do complex eval essentially for free in hardware and still reach 200M nps average case. Sure, they did not do null-move (I assume that this is what you refer to as "pruning"), but they probably had good reasons to. I think it is safe to assume that an 8M nps (or whatever they get from the 8-way combo) Fritz machine can't touch DBs strength. I doubt Fritz would be better even given equal NPS, which is perhaps 6 years away or more. So, in 2007, when 10 years has passed since DBs last match, we *might* have a new computer combo that matches DBs strength, but probably not before that. I agree we don't *know*, but this is a reasonable guess. Jesper
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.