Author: Robert Hyatt
Date: 19:02:46 01/29/99
Go up one level in this thread
On January 29, 1999 at 20:21:44, Matt Frank wrote: >>The Deep Blue machine had 256 Chess CPU's. By the term "Chess CPU", that means >>the native instructions were things like "Nxb4" >>These CPU's calculated 256,000,000 NPS. About a factor of 1000 faster than the >>best available current systems, give or take a factor of 4 (depending on your >>budget and the kind of searching you do). > >I have no question about the speed of Deep Blue, or the research budget. > >>The Budget for the Deep Blue project must have been orders of magnitude greater >>than any resouces available to the best equipped PC programming team. People >>have an image of IBM as some stodgy old guys standing around in black suits, but >>you had better believe that they have some of the best research and development >>teams in the world. Just do a web search for "IBM fellows" and you will find >>some pretty impressive names. > >The resources that were given to the Deep Blue project were considerable, yet >that doesn't necessarily translate into a added value over software makers >dedicated to their craft. Surely I don't have to provide a list of software >wizards that destabalized markets and terrorized companies many times larger. >Reputation lasts as long as it works in this field. Don't let resource >allocation and reputation cloud your perception. The leap forward in playing >strength by the IBM team is almost totally attributed to faster processing. I do >not underestimate the work Mr. Benjamin and the programmers, yet I am telling >you that the top flight software competitors are in a lot better shape in terms >of producing strong programs with what they have to work with compared to IBM. > >Furthermore, how is this for a challenge: Give me a Pentium 2 450MHZ with 256 >ram using Hiarcs 7, and I will play Deep Blue as configured for it's win >against Kasparov in 1997, with these time limits::: Deep Blue 5 minutes for the >game, me 40/2hrs 20/1hr. That is only 24 to 1 time advantage. That should >indicate that Deep Blue does not have nearly as big of an advbantage over the >present programs. Indeed it would indicate that the programs are much more >efficient using their native hardware compared to the IBM program. After all >even 24 times the present speed of a 450 MHZ Pentium still leaves you behind the >speed of Deep Blue. > >I think you don't know where the micros are now because the GMs are hiding. >Anand said to a reporter after his match with Rebel 10, "No more, with >computers." > >Matt Frank think realistically. The DB guys have more experience than any commercial programmer by a _wide_ margin. Murray Campbell was working on computer chess in the 1970's... for example... Don't underestimate their skills when comparing them to commercial programs... they are _not_ behind at all, quite the contrary. and no micro has come close to beating a kasparov in tournament play with a million dollar prize on the line. Nor can one come close today. I don't think they could beat any GM in a match at 40/2hr at present, although they would likely win some games... and this _particularly_ evident when the GM is 'computer savvy' and knows how to 'play the machine' which is becoming more common...
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.