Author: Robert Hyatt
Date: 13:43:38 07/22/00
Go up one level in this thread
On July 22, 2000 at 13:17:57, blass uri wrote: >On July 22, 2000 at 12:49:49, Peter Hegger wrote: > >>On July 22, 2000 at 00:39:09, blass uri wrote: >> >>>On July 21, 2000 at 16:48:09, Randall Shane wrote: >>><snipped> >>>>Using the above statement to claim that Deep Blue had a simple evaluation >>>>function is a clear misunderstanding of the paragraph's internal and external >>>>context. From that statement, all that one can reasonably derive is that Deep >>>>Blue has a simpler eval function than the human brain >>> >>>I think that commercial programs has more complicated evaluation function than >>>the human brain. >>> >>>Humans cannot remmeber many numbers to calculate the evaluation of the position. >>>The advnatage of humans is tha ability to use selective search and the ability >>>to learn from search to change their evaluation and not more complicated >>>evaluation function. >>> >>>Uri >> >>I for one believe that the human brains evaluation function is so many orders of >>magnitudes more complex than DB's that it is totally unfathomable to suggest >>otherwise. >>Take any program you like and put it on a PC that is so slow it only computes >>3 NPS such as a human brain does. Then play a games of chess against it and let >>me know who won. >>Of course DB gave an illusion of complexity because of it's 200,000,000 NPS But >>it is an illusion none the less when compared to good ol' grey matter. >>Regards, >>Peter > >1)Humans know better which lines to search and know better rules to learn from >search about evaluation so your test proves nothing. > >2)More complex does not mean better evaluation. > >3)From the human's point of view the evaluation of chess programs is more >complex because if humans try to calculate the evaluation function of crafty in >a game of chess they will lose on time and will not be able to search 3 nodes >per second or only one node per minute. > >I am a human so I look at the facts from the human's point of view >Uri You are wrong there. I am a human. I play chess. And I use every evaluation term in crafty in my chess. I just don't always use them "consciously". And my neural networks have learned to recognize chess positions, weight the various eval terms accordingly, and decide between two moves. The brain just does it _differently_. Not necessarily better, because the goal is to do some task and he who does it fastest and most accurately is the best. But I don't think _any_ program knows 1/10th of what I know about chess. Or 1/1000th of what a GM knows about chess. They know enough positional stuff to get by in many cases, and the tree search is so good that they can simply calculate the tactics where we often rely on tactical intuition.
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.