Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strength of the engine in chess programs (Critical points on memory)

Author: José Carlos

Date: 10:24:19 05/22/02

Go up one level in this thread


On May 21, 2002 at 19:15:58, Rolf Tueschen wrote:

>"I would bet a new Porsche vs a Mattel Hot Wheel toy..."
>
>(From a reliable source,
>wasn't it Janis Joplin?
>Read below for the solution.)
>
>                         =============
>
>I will always respect you for your willingness to discuss, because only this way
>we can clarify things right now, and it shows that things are not what you'd
>thought in the last three decades. Today this will be proven, just below in this
>posting. But again due to your openess this was possible.
>
>
>On May 21, 2002 at 11:40:07, Robert Hyatt wrote:
>
>>>Nothing of pre-recorded stuff? Mhmm. Or do you use a different wording?
>>
>>
>>Nope.  It is "memory".  Just as _I_ recall specific sequences of moves
>>(or entire games for that matter).  I recall sequences of moves from _my_
>>memory.  The computer recalls sequences of moves from _its_ memory.
>>
>>Seems simple enough to me...
>
>Too simple and false to me.
>
>You don't deny pre-recorded stuff. You say that it is MEMORY. Ok.
>Now the little and final question for this chapter:
>
>Whose memory? Read what you wrote yourself! You recall YOUR memory. And the
>computer what does he recall? HIS memory? HIS material? Nah.
>It's the memory of GM chess of the last 100 years. Ah, you might say, and the
>haman GM, what hey are doing? The same! I say No! The human GM has understood
>what he has in memory, but your comp has in memory most of all stuff he cannot
>understand, because he would be terribly lost if the memory would be taken away.




  You seem quite confused with computer science terms. Every piece of sotfware
lies in memory. There's RAM, there's cache, there's processor registers, there's
hard disks, floppy disks, tapes, ... All of that is virtually the same, except
for speed and capacity.
  An "opening book" is just a piece of software, a series of information bits
stored in a memory. No matter if in RAM, chache or floppy it's the same. Just
software.
  Programs don't "understand" anything. Period. A program is a black box with
two ends. You enter data in one end and get the result in the other end. What
the program does inside the box is not relevant, unless for the programmer.
Imagine this piece of pseudocode:
  if (move=="e4")
    reply("e5")
  else if (move=="d4")
    reply("Nf6")
  else if...

  Whether the program does its calculation this way, or with an AlphaBeta search
is irrelevant for the user. Same if these "if's" are in RAM or in hard disk.
Actually, the whole program is usually in hard disk or CD before it gets loaded
to RAM for performance reasons.




>But the memory of a GM, now this is crucial and you should note that, wonders me
>that GM Roman did never tell you, does not primarily contain opening moves. The
>GM is thinking in 'Ganzheiten', patterns. That's why I am saying that against a
>GM your computer memory book out of 100 years of chess is not the decisive tool
>to take advantages over the GM but it's a toll for defense to not being busted
>too early in the game. Against players like us the tools become decisive and
>winning because we do not master opening theory in total. Of course we have our
>variations too up into the twenty moves but it's by far not total knowledge. Why
>for a GM it's different with the Ganzheiten? Because, and that is just chess,
>the single opening moves or full lines do not matter much without the connection
>to the middle-game and even endgame. So a GM has, depending of the opening
>choice, a direct view on Ganzheiten for more than opening lines of 20 or 30
>moves. THAT is memory! But "your" memory argument of computerchess is simply a
>false understanding and a consequence of the different situation in the
>beginnings in the '60s. Computerchess then needed a couple of foreign moves in
>memory of the machine, because otherwise it hadn't played correct chess at all.
>But if you want to claim that your computer's memory would exactly mean the same
>than the memory of the chessplayers, then you have simply not understood the
>chess memory of human chessplayers. It's funny, because you are a chessplayer.
>And you did never observe that you didn't simply play move after move by heart/
>by memory without having the whole idea/picture/Ganzheit of the variation before
>your inner eyes?




  When you see a program play a "book move", how do you know whether the
programmer inserted 5000 lines of code in eval to handle that position or not?
You simply don't know. And you don't need to know it. How the program works is
not your bussiness as a user.




>I repeat. A computer is only functioning you say, when he has memory. Either for
>some code with specific computer related stuff or chess content related. Both is
>memory for you which you equate with the memory of a chessplayer.
>
>I repeat the cheat.
>
>A human chessplayer uses his memory where he has moves of course but most of all
>Ganzheiten of ideas, concepts etc.
>
>A computer needs memory to function, but what you do so that he can play chess
>is you take the best theory of 100 y. of GM chess, enter these foreign moves
>into the memory of the machine and you are thinking that this is only fair
>compared to the human chessplayer. But it's not.




  If the "book moves" are "foreign moves", then also evaluation is foreign
evaluation (the programmer implements some rules humans have discovered over the
years), and the search rules are foreign (invented by humans, the computer
doesn't understand them), etc. You're logic is completely flawed.





>As others like Martin and Torstein have said, we could try to let the machine
>analyze the openings itself. Fine by me!




  Why? The "knowledge" the program would use for such analisys is "human
knowledge", so same cheating then...





>But Martin will understand in near
>future that chess is not like checkers. And for all, he has confirmed me that he
>won't change the choices of the machine. Then Aloha I would say, just smiling.
>The machines of the next 10 years and more won't find the best opening concepts
>on their own. Man had to intervene. But this is exactly what would destroy your
>equating concept of the two different "memories".
>>
>>I think the _intent_ of the rules was quite clear.  The "player" has to rely
>>on his mind and memory, exclusively, when playing a game.  The computer does
>>_exactly_ that...
>
>Yes, yes, _his_ mind and _his_ memory and _his_ work, but not the data taken
>from human masters. Period.




  All the chess knowledge in chess programs is take from human masters. Period.




>>>It's funny, yes, GM with eidetics have a database in their head.
>>>But all the rest of us is arguing against your forbidden use of 'books'.
>>
>>Sorry, but _I_ play chess also.  and _I_ remember specific sequences of
>>opening moves, from studying them in books like MCO, etc.  I don't know how
>>_you_ play chess, and I don't care.  But I do know how _people_ play chess,
>>and they _definitely_ rely on memory as one component of their skill...
>
>Uhmmm. For human chessplayers you are right. They can 'read' in books. A
>computer can 'read' too? No? You must take the whole theory of human players and
>putting it into the "memory". What does it mean human 'reading'? Thinking,
>understanding? Yes. But the machine just is instructed to copy the book lines
>the way they were imported into memory by YOU, the programmers or special book
>doctors.
>
>All I'm saying is let the machine find its own openings




  No. The machine doesn't find its own anything. It finds what the programmer
tells it to search.





>and let it play along
>these line but don't use that wrong consept anymore that came from the '60s,
>where the machine couldn't play sober chess without some moves. You simply must
>forget this praxis, because it's not fair and against the FIDE rules. But I'm
>_not_ saying, and that is what you are trying to tell the people, that machine
>should work without memory. That would be indeed a contradiction in itself. So,
>no need to smirk about the totally illegal engine. You know, we come out of
>different disciplines but we ain't therefore beginners. The funny circles
>argument doesn't become true by mere repetition if you are unwilling to stand up
>and watch the third dimension!
>
>I is also a bad argument to quote from a debate in R.G.C.C. where we had no
>exchange of arguments last year.
>
>
>>>Are you referring to the long-distance connections to the computer? BTW you're a
>>>chessplayer too, and not GM, what is your opinion in that question? Are you
>>>happy with the comps to read in books during the game? Wouldn't you be much
>>>better in chess if you could do the same?
>>>
>>
>>
>>I am referring to the phone connection, the space required by a computer,
>>the noise it makes, the confusion it adds, the interest in its games and
>>screen output by spectators, the rule that says that players can "opt out"
>>of playing the machine prior to the first round, which makes pairings very
>>complicated in the last few rounds.  Etc...
>
>Yes, very true, but then you can't accept the topic here, which is intended to
>show how "we" could and should change some traditional errors or birth defects
>so to say, to be better presentable for human tournaments?
>
>>
>>_My_ "comp" doesn't "read in moves during a game".  It "studies" the games
>>prior to playing just like a human.  It analyzes who won, who lost, what the
>>various positions looked like.  It then plays those moves, and then remembers
>>how the resulting positions actually worked out in a game (learning).  It does
>>_exactly_ what I as a human do...  _exactly_...
>
>I read that before, but isn't it a weak argument? You know too well that GM
>don't do that, that's a typical toy in commercial chess programs. There you are
>talking to real GM and you want to tell us and more so yourself that chess is
>that stupid? Nothing against your fine idea, better than nothing, but you know
>quite well that you can never expect the program to play good enough with mere
>counting percentages - automatically because you don't have the time to do it
>yourself (your own words). Just tell me, do you really believe in such bogus
>stats? In the ChessBase books you could find nice stats too. With some few (two)
>categories. Elo and performance. Now, that could be a good example for the
>beginners in statistics classes. Also the question of the choice of the
>population.
>
>>
>>Just better in some ways.  But the same nonethelsss..
>
>Against FRITZ I play 350-400 points above my normal Elo. Of course I play with
>books too. :)




  Your choice...



  José C.




>>It is all available in r.g.c.c archives.  You have brought this up several
>>times.  Same arguments.  Same acrimony.  Nothing new.  So yes you _did_ do
>>that.  This is just the N+1'th instantiation of the same old stuff...
>
>Yes, the archives contain many things from our "childhood and youth...".
>Who do you think wrote this here:
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>"I would bet a new Porsche vs a Mattel Hot Wheel toy that what has been written
>in R.G.C.C was _not_ written by Rolf. (1) it doesn't sound like him at all;
>(2) the vocabulary/style is wrong; (3) the posts were from aol, which was
>_never_ used by him (he used t-online.de exclusively)."
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
>Of course, the author was no other than you, Bob.
>
>Let's find a consense for defining computer chessplayers, memory and stuff like
>that.
>
>Rolf Tueschen



This page took 0.01 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.