Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strength of the engine in chess programs (My theory against machine)

Author: Uri Blass

Date: 01:27:37 05/26/02

Go up one level in this thread


On May 25, 2002 at 20:59:41, Rolf Tueschen wrote:

>On May 25, 2002 at 19:48:59, Uri Blass wrote:
>
>>
>>The fide rules are for games of humans against humans.
>>They are not relevant for computers.
>
>Thank you for insisting on that special point.
>
>>
>>If you talk about fairness it is possible to say that every competition
>>human-computer is unfair because humans also have no chance to run faster than
>>cars.
>>
>>This does not mean that the car cheats when it is faster than humans.
>
>Ok. I've heard that sort of argument for many times now, also Bruce for example
>always puts very nicely this sort of comparison. But what would be if we all
>would join and talk the same old arguments over and over again. Because I want
>to demonstrate here in this posting for the first time, I hope I can have the
>copyright for it, that this argument and also similar with horses and man - is
>a fallacy. Excuse me, Uri, but I think it is a nice present for the weekend for
>you. I hope you won't refutate my theory. :)
>
>>
>>Fide forbid to use opening book but fide do not forbid to remember opening books
>>and humans remeber opening book(understanding does not change the fact that they
>>remember)
>>
>>If you do not allow computers to remember previous analysis that they did then
>>you give humans unfair advantage.
>>
>>If you allow computers to remember previous analysis then you should allow them
>>to use book that is based on analysis of database of a lot of games(humans also
>>learn from analysis of a database of games).
>>
>>saying that humans can play moves in the opening in less than 1  second based on
>>previous analysis when computers cannot do it is unfair.
>>
>>If we think about fairness it is more logical to limit the speed of the computer
>>and not to prevent it an opening book.
>
>For the sake of the argument let's not debate about speed today, please. Let's
>only concentrate on books. Since I saw that we often thing similarly let me try
>to clarify what I mean with my continual claim of unfairness with the "socalled
>GM books" in computer programs. Since there's only one important point, it's
>very easy to follow, but usually people talk what others talked before. THere
>are even people who copy and past what others wrote. Also journalists work like
>that. It's a terrible mess.
>
>Let me try. Why GM books are unfair and against FIDE rules?
>
>Is it because computers have machine-like, perfect memory? - No!
>Is it because of the large scale of the memory? - No!
>Is it because it's data from GM? - No!
>Is it because it's taken for free from large databases? - No!
>Is it because masters and experts and amateurs have a weaker memory? - No!
>Is it because computers don't understand chess like man does? - No!
>
>Because it becomes to be boring, I stop it here.
>
>It's unfair because the GM books contain data, computers would not understand or
>find even if they would analyse the problem for a long time. The material is
>simply human stuff and not for computers. If a computer is allowed to use these
>parts of the GM books, then it's cheating.
>
>Now, the direct attack against my theory comes from Dr. Hyatt. He said, GM do
>the same and me, Bob, too. We learn by heart certain line, do not understand
>them and of course play them if necessary and bingo, that's the same thing.
>
>Now, this looks rather strong, no?!
>
>But I oppose this argument. But now it becomes a little bit difficult.
>Is that what Dr. Hyatt is talking about fitting with the ideals of real chess,
>human chess? No! Because, excuse me, Bob, only stupid and very weak players play
>like this. Good players play what they can understand, also with longer
>extensions into the middle game, where they could find similarities with
>patterns they know of. Of course a really weak player looks into a book and
>plays a line as far as he can memorize it and then he starts to gamble because
>he doesn't know what to do next. etc.
>
>Why is that forbidden by FIDE? No, of course not! Could also GM play like that?
>Yes, if they were stupid enough, but they won't do this. Normally they play
>_their_ openings no matter if the opponent can prepare years in advance. But
>that doesn't matter. Important is only that you stay in the scale of your
>mastership=knowledge. If you play like Kasparov against DB2, you give away your
>advantages.
>
>Now, I ask you, Uri, and it's not a really scientifical question, it's ethical.
>Why man, the human chessplayers should allow 'machines' to use GM books from man
>with parts no computer could find. Because here is the point where Dr. Hyatt is
>badly wrong! The best human players are of course able to analyse the whole
>theory and also discover all the tricks on their own, _in principal_!
>But the computer can not do this!
>
>Therefore it must be forbidden. And here we need not debate that a weak amateur
>couldn't find the tricks either like the machines. And he could still take it
>from the books and play it. Yes, he can! But still I say, doesn't matter, man is
>man, and machine is machine. If the machine side is interested to play with man
>then man should set the rules. Period. Or do we like suicide and let the
>machines do what they want in compertition with us? No! Of course not.
>
>So, Uri! The machine can take as much it wants from the books, but only that
>part the mahine can analyse itself. The rest remains magic for the machine and
>belongs to the human race. Period! :)
>
>Rules in sports with the participation of two different species, are defined by
>man or the event won't happen. Period!
>
>I don't see where such rules could be unfair for machines because machine don't
>even understand the concept of unfairness. But man does!
>
>So, please. Do you understand now, why it was so difficult for me to explain my
>theory? Because it is not 'fair' for machines and man likewise! But what we have
>seen in the past events isn't fair either. It's scandalous! I repeat, it's an
>offense against the human dignity what happened to Kasparov there in 1997. And
>we must guarantee that this won't happen again. Not that a machine wins, but how
>unfair the event was designed!
>
>Are we so stupid that we take our best ethical concepts and give them away to
>machines and their operators? Of course not. We are not crazy! The machines have
>so many advantages, the are never tired etc, but we won't give our typically
>human discoveries away, only to weaken us.
>
>Now, finally your car metaphor. It's not the right comparison for the chess
>machines.  The speed of the machines is not the top opponent of man. Because
>what is speed if you run against a wall in the end? That is not a threat against
>the intelligence of man. But it's a threat, if human operators or programmers
>come and steal the parts of chess theory only man can understand and no machine.


I do not believe that there are part of theory that no machoine can "understand"
Machine can understand everything that the programmer explain them.

I also believe that machines can find theory moves by search in big majority of
the cases.

I do not expect a big difference in the level of programs against humans if the
program does not play theory moves that it cannot "understand" because I believe
that programs can "understand" almost everything without a work of the
programmer to change the evaluation function.

The main problem is that program neeeds a lot of computer time to "understand".

The computer time is needed for analyzing every book position for a long time
including extensions of lines that were played in human-human games but if you
consider the fact that there are a lot of computers then it is possible to do it
if the customers give the programmer a lot of computer time and they may do it
for getting an upgrade.

The easy solution is to save this time by adding the program an opening book
but I do not think that it is the best solution because I believe that computers
may find good novelties by the lot of time that may be used to "understand".

Part of the novelties may be bad novelties but I  believe that the chances of
humans to find a line when the computer plays a bad novelty is small if the
program is not public at the time of the game.

I also believe that a good program may learn to change it's evaluation function
after losing a game in position that it does not understand because if you get a
good position based on your evaluation and lose then you should learn that your
evaluation is wrong or that you do not know to play that position and this is a
good reason to evaluate it as bad for yourself.


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.