Author: Uri Blass
Date: 02:12:32 04/30/05
Go up one level in this thread
On April 29, 2005 at 20:31:08, Rolf Tueschen wrote:
>On April 29, 2005 at 09:51:04, Thorsten Czub wrote:
>
>>On April 29, 2005 at 08:34:38, Rolf Tueschen wrote:
>>
>>>On April 29, 2005 at 08:27:14, Thorsten Czub wrote:
>>>
>>>>The programmer and operator did not believed into the situation and continued
>>>>and continued the game. he hoped that CSTal would FORGET about the law.
>>>>but it could not. it was in the static knowledge of CSTAL.
>>>
>>>Isn't it a good compliment for Chris Whittington and his work? Now the million
>>>dollar question: why nobody else did continue this paradigma? Other question:
>>>was that game you quoted the exception of a normally losing trend? Or still
>>>another question: would such a program, like CSTal, if it would have been
>>>improved, be still a good rival to the search monsters today? Just for
>>>reminiscence...
>>
>>so many questions.
>>i doubt i can answer them all.
>>
>>The game is an example for the 2 fights of ideology: search versus knowledge.
>>
>>in this game CSTAL KNOWS something that dark thought is NOT knowing.
>>therefore CSTal can draw because it knows the simple rule:
>>a king supports the pawns.
>>without king support, you cannot do anything.
>>therefore the first thing CSTal is doing is to seperate the opponent king
>>from the own pieces.
>>
>>this is especially interesting because the opponent was a very very fast
>>chess program on a very very fast machine at this time.
>>during the game Ernst A. Heinz often impressed us with telling the search depth
>>of Dark Thought. CSTals search depth was very low.
>>
>>In the endgame, in the position, Ernst STILL told us that dark thought COULD
>>win.
>>Ernst believed in the main lines of his program. Dark Thought searched so deep
>>and the problem was that in EACH of these main lines, CSTal let the opponent
>>king come to its pieces. But of course these main lines were pure fantasy. but
>>very deep fantasy !
>>
>>we told him that CSTal will not, never , behave as in Dark Thoughts main lines.
>>Because it found the right moves without even doing a search, just by static
>>evaluation !
>>
>>No search !! versus very deep search.
>>
>>I guess this must have been shocking (for Ernst).
>>He was so shocked that although he had the faster hardware, although he had
>>the better position , although his program searches better, he CANNOT win.
>>
>>Knowledge can substitute search.
>>
>>therefore i do not really understand what chandler wants to tell us.
>>
>>todays strongest chess program rely of course heavily on a good and deep search.
>>
>>
>>IMO the paris championship showed that chris was able to do a good job.
>>also the paderborn championship.
>>
>>on the other hand i think that the very tough work can only be done in a team.
>>One person alone cannot stand the difficult years or development by tuning such
>>an approach. you would resign to early if you get beaten by the other programs .
>>
>>those ideas sometimes develop into a working program. in the early years of
>>chess programming there was e.g. Thomas Nitsche and Elmar Henne who tried a
>>commercial approach with Mephisto III. this was also a very radical program
>>doing only
>>3 NPS and much via static stuff.
>>
>>And this approach also was stopped not because it was maybe not succesful but
>>because the programmer, at a certain point of development was so much
>>disapointed about the prejudices he had to fight ("it cannot be done this way!")
>>that he lost interest in programming.
>>
>>
>>Thomas Nitsche and ELmar Henne won the championship in Glasgow 1984 together
>>with
>>other chess programmers with their new concept.
>>
>>it would have been a much more difficult task to continue the race.
>>in this moment H+G changed horses to Richard lang.
>>
>>And Nitsche and Henne were in a dead end street.
>>
>>
>>>Or still
>>>another question: would such a program, like CSTal, if it would have been
>>>improved, be still a good rival to the search monsters today?
>>
>>I think so.
>>
>>but the question is not how much potential is in the aproach.
>>it is THE PROGRAMMER who loses interest in the more and more difficult idea.
>>
>>CSTal was very close. but it would have been much more difficult to continue the
>>race.
>>
>>we worked arround 4 years and more to reach the level of the others with a
>>different program.
>>
>>chris tried a complete rewrite of the program.
>>the idea was that the old program who grew over the years in a kind of chaotical
>>way (chris programming :-))
>>had to be structured new to make progress.
>>
>>it would have been a difficult task to continue.
>>
>>I would like to see them all come back, Nitsche + Henne, Dave Kittinger, Marty
>>Hirsch, Martin Bryant, Chris Whittington, the Spracklens, Julio Kaplan.
>>
>>But i think this is utopia.
>>
>>they have lost interest. they have other things to do.
>>or they think that they would have no chance.
>>
>>IMO this is a wrong thinking.
>>
>>If you consider the strongest dedicated chess computers, e.g. Maestro 10 mhz
>>with D++ (Kaplan), or Super-Forte C with 6 mhz (Kittinger), or Mchess8 (Hirsch)
>>or
>>Sparc Module (Spracklen/Kaplan), ... it would be possible to come back.
>>
>>on todays hardware the old Wchess still plays good.
>>And Mchess8 on todays hardware still has enough AI in it to kill a top.
>>
>>But the programmers need IMO support. they will not do it by themselves.
>
>
>Well, then we are without hope. Because you can't organise the motivation for
>another human being. Either someone has it or not. And for this job here you
>must have an outstanding high motivation.
>
>But let me make a prophetic statement. For all what I know from chess and
>computerchess I am convinced that the brutal search approach - on the base of
>always stronger hardware - will be condemned to come to a dead situation where
>no longer effective progress in chess will be possible. And if the programmers
>don't understand how to improve the chess knowledge for the machine, then they
>will never overcome human top chessplayers. In special if certain limiting rules
>will be found for the attending machines. At that moment then people will be
>sorry that in the years around 1997 those who followed the other paradigma
>(knowledge) like Chris and his team had no better support in their peer group!
>
>It's a pity for that waste of creativity because of silly flame wars on the
>internet...
>
>Hold open your eyes and inform the community about any small development when
>someone new or old gets into the knowledge issue again - for the best of
>computerchess! And don't let you become depressed. You once were part of the
>better issue in computerchess. All the best to you in future!
I think that the example is not example when special knowledge is needed.
Even stupid Movei can play better than Dark thought and find Rxb2
Cheap knowledge is enough for it and I think that most programs of today have no
problem to play the endgame better including the amateurs not because of deep
search.
Uri
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.