Author: Matt Frank
Date: 14:27:05 02/28/99
Go up one level in this thread
>I think you are absolutely correct Matt. But be careful, this is a >dangerous conversation. I'm sure you are about due for the "you don't >know what you are talking about speech" or something similar. > >I don't see your statement as taking anything away from the massive >effort the talented Deep Blue team put into their project. I see >your statement as just plain common sense and logical. > >Bruce is right too, it's not fair to try to scale our programs >to their hardware. I don't think the learning thing is very >relevant but I understand what you are saying, I believe there >are dozens of little micro-engineered things (like this) that micro's >almost certainly do better. I think you hit the nail on the >head when you say Deep Blue is not as far ahead as most assume. > > >- Don Thanks for the cautionary advice; I know who I was talking to. And yes, the multiple micro-engineered factors (learning being one of those factors--don't underestimate it) provide advantages that the Deeper Blue team did not have for their system. Of course, many factors were to the advantage of the Deeper Blue team (e.g., endgame databases, etc.) yet the present state-of-the-art in micros are quite advantageous compared to the Deeper Blue software technology. Incidentially, a secondary issue refers to the previously believed argument about a leveling effect (i.e., as you get to master, IM, and now GM level) in which the increase in speed (doubling) for a GIVEN PROGRAM will result in diminishing elo returns; seems to be a case of whisting in the dark. This appears so, because any given program seems to have a moderate leveling effect the farther the program probes the search tree (see Newborn, 1997 pp. 121-125), not the flattening out expected by some in the late 70s and early 80s. All in all, this portends that present day micro sofware programs are close to reaching super GM level performance (i.e., = Anand or Kasparov) with increased hardware performance (i.e., hardware available in PC format in the time frame 2001-2003). Bold predictions, but IMHO reality. Regards, Matt Frank
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.