Author: Bruce Moreland
Date: 17:27:05 11/07/00
Go up one level in this thread
On November 07, 2000 at 18:59:57, michael mayfield wrote: >I remember reading somewhere that deep blue's software wasn't very strong >compared to programs like fritz and that it was the hardware speed that was the >main factor to account for its strength....Anyone have an info about this? This question is guaranteed to start an emotional argument here. The question is complicated by the fact that DB is a hardware/software hybrid, and so the design tradeoffs they made are different than a PC software engine would make. The skill set exercised when a PC programmer builds a program is different than what you use when building hardware. The amount of competition that exists in the PC world is much greater than that that exists in the world of supercomputers, and in fact there is no credible supercomputer competition for DB. They can adopt whatever religion they prefer, and there is nobody that can tell them that their religion is worse than some other religion in practical cases. A PC programmer has to adapt when confronted with failure, which is met with more often. The amount of time finishing the project is also not as high as that spent by PC programmers. In one of the two matches they barely got hardware before the match started, and I would bet that the amount of testing and refinement they were able to do is much less than what a micro programmer can do. What it really comes down to is a question about whether Hsu and Campbell are for real. Given standard PC resources, could they make a strong program? Given the backing of a hardware company, could the rest of us make a stronger program than they did? I give them credit for positioning themselves to build such a thing, and for building it. Nobody else has done this. And I give micro programmers a lot of credit for having built algorithms that are very strong on much weaker machines. It isn't possible to provide a good answer to your question, until they build a micro program or someone else builds a strong supercomputer program. I am not willing to say that they are dumber or smarter than micro programmers. I've spoken with a lot of people involved in this field, and while it is hard to assess someone's personal capabilities without getting closer than most computer chess programmers get to each other, a lot of the people I've met seem pretty damned smart. And I'd include Hsu and Campbell in that group. bruce
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.