Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Bob, do You finally accept, that PCs are playing at GM level?

Author: Robert Hyatt

Date: 18:01:54 06/21/99

Go up one level in this thread


On June 21, 1999 at 14:23:28, James Robertson wrote:

>On June 21, 1999 at 13:58:32, Robert Hyatt wrote:
>
>>On June 21, 1999 at 12:44:25, Paul Richards wrote:
>>
>>>On June 21, 1999 at 09:29:53, Robert Hyatt wrote:
>>>
>>>>1.  Pick any of the 5 programs that played the GM players.  I will find a
>>>>game where they played so badly that if you look at _that_ game no one would
>>>>consider that program to be a GM.  For example, take the winner and look at
>>>>the playoff game.  Three different GM players commented that they had _never_
>>>>seen white screw up the opening so badly...
>>>
>>>True, but GMs make terrible blunders too.  The difference is that the
>>>program will make the same sort of blunder until you fix it.
>>>
>>
>>
>>You missed my point... tactical blunders are not uncommon.  But this game
>>was _not_ a blunderful game...  It was just positional mistake on top of
>>positional mistake...  IE no one move led to that position around move 25,
>>it took _several_...
>>
>>
>>
>>>
>>>>3.  I've been working on chess programming for a long time.  And regardless of
>>>>how they 'seem' to play in many games, I still know just what they can and can't
>>>>do.  And they are nowhere near a GM's level in 'knowledge'.  They are still
>>>>surviving on tactics.  And there are plenty of GM players that know how to
>>>>squelch tactics and make the game hinge on positional play.  And there the
>>>>programs simply don't measure up.
>>>
>>>True, but the only real measure of strength is in the result.  The relative
>>>strength of a human GM is knowledge, the strength of the computer is
>>>tactics.  You posted a quote from a GM observing a game who admitted that
>>>in complex tactical positions Crafty was much stronger than he was.  In
>>>other words it's common knowledge what the relative strengths and
>>>weaknesses of the two species are.  They are two different animals with
>>>a different approach to the game. But just as we don't dismiss human
>>>GMs for making tactical blunders, we can't say programs are "weak"
>>>because of their lesser knowledge.  Sometimes DB played like a non-GM,
>>>other times it clearly out-thought Kasparov.  So what?  He lost.  The
>>>sum of DB's strengths minus its weaknesses was greater than Kasparov's
>>>total for the match. What matters where ratings and titles are concerned
>>>is the final result.
>>
>>
>>Note that computers are better _in some types of tactics_.  But there are
>>positions where a computer has no chance.  IE Shirov's Bh3 sac.  It is not
>>impossible to solve with the right extensions, but no one does yet (perhaps
>>excepting DB as I haven't asked Hsu if he has tried it).  But there are
>>still _plenty_ of places where a human GM can tactically blow away a computer,
>>because in some cases, the tactics occur after a 30 ply forcing line that the
>>GM can follow but the computer can't...
>>
>>IE computers are tactically strong, but not invincible.  Crafty still loses
>>blitz games to GM players.  Not real often, but enough to see where it has
>>tactical problems even at 1M nodes per second...
>>
>>
>>
>>
>>>
>>>>5.  GM players exhibit a consistency in quality that computers don't.  A
>>>>computer will play like a GM for 5 games, and like a beginner for 1.  What
>>>>happens when the GM players learn what the computer can't do and then
>>>>exploit that game after game?.
>>>
>>>The consistency issue is debatable.  GMs play well until they make their
>>>next blunder.  Computers are obviously completely consistent, it's just
>>>that their weaknesses are only exposed when certain positions crop up, so
>>>it has the appearance of a sporadic phenomenon.  But if a computer plays
>>>like a GM a good percentage of the time, it's a GM.  Once a human earns
>>>a GM title, it can't be taken away, so you don't have to have a great
>>>performance every game or every tournament.  Once you earn that title
>>>with a few good performances it's yours, so by that measure I think the
>>>programs would have easily earned their titles by now.
>>
>>
>>
>>Computers are not consistent at all...  From something I have told before:
>>Before Jakarta, Roman was playing lots of games vs Crafty to help me tun it.
>>One day he called and started on the 'bad bishop' thing once again....  and
>>said that he had found it serious enough that he was able to make it screw up
>>fairly frequently.  A week later he called back and said "much better...  it
>>is not hemming in its own bishop any longer...  good work."  I didn't have the
>>heart to tell him I had made no changes as this was only 2 weeks prior to
>>the tournament.  :)
>>
>>IE it looks like a genius in some games, like an idiot in others.  As do all
>>programs...  A computer might play like this over 8 games:  2500 2500 2500 2500
>>1800 2600 2500 2400.  A GM won't have that 1800 game.
>
>Sokolov just did. :)
>
>James


Not quite.  He did a single 1800 _move_.  Big difference from an 1800 _game_
if you know what I mean...



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.