Author: Matthew Hull
Date: 06:09:07 08/22/02
Go up one level in this thread
On August 22, 2002 at 08:36:07, Uri Blass wrote: >On August 22, 2002 at 08:10:27, Matthew Hull wrote: > >>I've been following these DB threads for a long time now. It appears to me that >>there is a mental block for people east of the Atlantic that does not allow them >>to accept some basic facts: >> >>1. Everything Hsu ever made (chess machines) was brilliant. >>2. Hsu's creations _slaughtered_ the computer competition...ALL OF THEM! >>3. Kasparove didn't watch what he was doing, and got his helmet handed to him >>by DB2, plain and simple. >> >>And you guys can't take it. You cant' stand it. You can't accept it. Why? I >>can't figure it out. Is it jealousy? Envy? Sub-concious anti-Americanism? >>CONCIOUS anti-Americanism? European ego-centrism? European inferiority >>complex? Something in the water? The air? The food? Religion? EU mind >>control? WHAAAAAAT!!!!!!! >> >>Most of you all seem like nice guys most of the time (excepting maybe Vincent >>"aspergers" Diepeveen). But when the topic of DEEP BLUE comes up, it's >>get-out-the-knives time. >> >>What have you to say for yourselves? > >I could not agree about claims that seemed to me illogical about deeper blue. > >It is clear that it was better than the machines of that time but based on >looking at the games it is not better than the chess programs of today. > >The claim that they search 18 plies with no pruning except futility pruning does >not make sense and finally they say that it was wrong. > >It seems to me that there is a mental block that does not let people to accept >the simple fact that deeper blue could not do the impossible(branching factor of >4 together with no pruning and no use of hash tables in big part of the search) > >some people assume that deeper blue was better than the programs of today only >because of the number of nodes and we discover that > >1)The number of nodes was a lie(IBM said 200M nodes when it was only average >number of 126M nodes). >2)The lack of efficiency means that the eqvivalent number of nodes on single >processor is clearly smaller than 126M nodes. > >It clearly make sense to believe based on these observation and the progress in >software that deeper blue was not better than the programs of today(No doubt >that it was the best at it's time). > >They had more knowledge in the evaluation but bigger is not always better >and tuning the evaluation is also important and I believe that their evaluation >was not better than the top programs of today because they had not enough time >to tune their evaluation. > >They searched more nodes than the programs of today but the difference is not so >big when you consider the lack of efficiency of deeper blue and the programs of >today compensate for it by better search algorithm. > >Uri Almost all of what you say here seems like speculation and blind assertion. What is motivating this kind of scoffing is what I don't understand. And it tends to be very vehement. Why? You say above they did "no forward pruning" which Dr. Hyatt has already said that Hsu said they did some kind of futility pruning. Why do you just contradict that? You say that bigger eval is not always better. Agreed. But the tone of your argument seems to assume that theirs definitely was not better. How do you know? It just looks like you guys are continually probing for ways to discredit the machine, the team, and especially the _big_, _rich_, _American_ company. Most of this stuff looks like axe grinding to me. What's the big deal 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.