Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Let's define GM

Author: Mark Young

Date: 14:33:09 07/15/98

Go up one level in this thread


On July 15, 1998 at 16:42:43, Robert Hyatt wrote:

>On July 15, 1998 at 15:50:59, Don Dailey wrote:
>
>>On July 15, 1998 at 13:30:09, Robert Hyatt wrote:
>>
>>>On July 15, 1998 at 11:57:01, Don Dailey wrote:
>>>
>>>>On July 15, 1998 at 11:48:59, Fernando Villegas wrote:
>>>>
>>>>>Hi Danniel:
>>>>>In this, like in any other field, a tittle is always the result of many
>>>>>accumulated results, some explicits and some not. It's clear that just to be
>>>>>part of a GM tournament an so to be capable of becoming a GM you must be at
>>>>>least an IM invited to be there. And to get 2500 elo you need to win many times
>>>>>people with or around that rating. So, in the case of Fritz, he will deserve the
>>>>>tittle if he can obtain the same kind of steady, regular results any human
>>>>>player need to be tittled as GM. To win here or there is not enough and at most
>>>>>is a sign that it could be a GM, If he try to be. In other words, we must not
>>>>>mix the meaning of being GM, a formal tittle given according certain scores
>>>>>under certain rules, and being capable of being a GM. I think last version of
>>>>>Fritz -stronger than the first by a wide margin, I believe- could be capable at
>>>>>least to be IM.
>>>>>Fernando
>>>>
>>>>
>>>>It would be nice if there was such a title of computer GM.  It should
>>>>be based on games against humans and not be subject to too much
>>>>manipulation.  There are so many logistical problems involved I don't
>>>>know if this could happen.
>>>>
>>>>- Don
>>>
>>>
>>>what would be the point of this?  Because then you still have to compare
>>>"GM" to "CGM" and they wouldn't be the same...  We need fewer classifications,
>>>not more, particularly when the criteria for earning equal-sounding titles
>>>(GM vs CGM) would be different.
>>
>>I don't know what you mean by fewer classifications since I'm not aware
>>of any classification system for computer chess unless you count the
>>rating system which is not a title.
>>
>>In my opinion, certain programs stand out as being the cream of the
>>crop.  There is no formal system of titles that make this clear at
>>all.
>>
>>I did say "it would be nice."   I'm not expecting to see such a system
>>but "it would be nice" if there were.
>>
>>I think I am no longer interested in comparing to humans anyway.  Since
>>computers seem to be improving rapidly there proably should be no
>>real attempt to integrate the two.  It may not make sense for computers
>>to limit themselves to a scale that they are WAY on the high side of.
>>
>>- Don
>
>I am still after humans, at present.  I believe it may be another 10-20 years
>before micros catch the upper end of the GM's with any regularity.  I now
>have hope that this is doable, because of DB.  They are roughly 1,000 times
>faster than we are.  if we generously consider doubling in speed every year,
>the log2(1000) = 10, which means 10 years will have us at 200M.  I don't
>believe that, because I believe that (1) doubling won't continue to happen
>every year, but is going to spread out more over time.  (2) I believe that
>DB is realy maybe 10,000-100,000 times faster, when you factor in their
>hardware eval that they can do at 200M nodes per second.

> But we at least
>know that the upper reaches of the GM range are attainable now, which we
>didn't 10 years ago.
>

I always thought everyone agreed it would be attainable, even 10 years age. Just
from predictions of how computers played at that time. From what I can remember
reading most people predicted that a computer would beat the world champion
sometime between the years 1995 and 2000, and micros would follow about 10 years
after that.

>So while computers may one day be supreme in chess, it is far enough away
>(20 years minimum) that there's plenty of work to do, and things to be
>tried...
>
>As far as ranking computers, I consider that hopeless.  Because playing
>against other computers and playing against GM players is currently two
>different projects.  I wish we had an easier way to get a USCF or FIDE
>rating, but it's not to be.  Crafty is a member of USCF, but I have been
>unable to find a tournament without the (NC) classification, so that it is
>not really "useful"...



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.