Author: KarinsDad
Date: 11:28:43 03/07/00
Go up one level in this thread
On March 07, 2000 at 13:06:01, Jorge Pichard wrote: [snip] >> >>I believe that a 386 can beat kasparov if people develop the right program >>because of the fact that 386 is more than 100000 times faster than all humans. >> >>It is more than 100000 times faster than humans in adding big numbers or >>multiplying big numbers and the only reason that it is not better than kasparov >>in chess is that humans do not know to develop the best possible programs today. >> >>Uri >Sorry Again but the Human brian has not been match with any supercomputer yet >and it will not be in the near 10 years. In calculating number there is no >questions that a Simple 386 can beat any human, but the computer is only a >device or a tool. There is no computer capable of building another computer yet, >that is where a supercomputer with the capability of our Human brain come into >place. That is why we have computer engineers and computer programmers that make >this wonderful computers easy for operator like yourself to interact with them >without spending 6 months programming the statements. > >Ps: Knowledge is good but it has to be combine with strength. example an 8 years >old kid could be a martial artist with a black belt, but he could never beat an >18 years healthy boy who has never had a lesson in martial art or boxing or >wretling etc... Another example take the early fritz 4 and use it with an AMD >ATHLON 800 mhz Vs the latest Fritz 6.a and use it with 386 which you mentioned. >you will be surprise to find out that Fritz 4 would beat it 60% of the time. I have to agree with Uri on this one. As programmers, we are limited in our thinking of HOW to write chess programs. Easy things such as "how to castle" are simple since the algorithms for them can be intuited by most humans. Complicated things such as "selecting promising lines to search" are complex due to the inherit difficulty in duplicating the human brains capability to filter out miscellaneous information, to recognize patterns, and to focus on pertinent information. If the idea is complex, it is often easier to code an algorithm that approximates the desired behavior than it is to actually code an algorithm to perform the desired behavior. A good example of this are opening books. Humans understand not only the lines of the opening books, but why they are good and bad, and what types of positions that given opening variations lead to. Computers only understand to select one of the next moves in the book. They have zero understanding (unless you consider some of the learning functions to be a type of understanding) of the positions within those books. Another example is the search engines themselves. Most engines will happily search any position given to them without having any idea as to why they are searching, and what criteria they should use to filter. Killer moves, capture moves, null moves, etc. are only simple programmatic tools to assist in coming up with such a criteria. But REAL criteria are not there, just approximations based on previous searches or based on a limited understanding of the current position such as which moves put the king in check or which moves capture a piece. A REAL simple program that would play perfect chess is one which accesses a complete EGTB of all 32 pieces. If such a database existed, you could write a chess engine in less than 50 lines of code and it could run on a 386 (with some form of extension to actually access the database, the file structures in current OSs could not handle the database itself) and it would beat Kasparov every time. KarinsDad :)
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.