Author: Christophe Theron
Date: 09:50:24 04/20/01
Go up one level in this thread
On April 19, 2001 at 18:55:47, Duncan Stanley wrote:
>On April 19, 2001 at 18:32:20, Christophe Theron wrote:
>
>>On April 19, 2001 at 18:06:37, Duncan Stanley wrote:
>>
>>>On April 19, 2001 at 17:14:11, Christophe Theron wrote:
>>>
>>>>On April 19, 2001 at 14:29:42, Duncan Stanley wrote:
>>>>
>>>>>On April 19, 2001 at 14:05:14, Christophe Theron wrote:
>>>>>
>>>>>>On April 19, 2001 at 13:54:30, Duncan Stanley wrote:
>>>>>>
>>>>>>>On April 19, 2001 at 13:01:53, Christophe Theron wrote:
>>>>>>>
>>>>>>>>On April 19, 2001 at 12:55:47, Duncan Stanley wrote:
>>>>>>>>
>>>>>>>>>On April 19, 2001 at 12:50:12, Christophe Theron wrote:
>>>>>>>>>
>>>>>>>>>>On April 19, 2001 at 12:46:45, Duncan Stanley wrote:
>>>>>>>>>>
>>>>>>>>>>>On April 19, 2001 at 12:43:05, Christophe Theron wrote:
>>>>>>>>>>>
>>>>>>>>>>>>On April 19, 2001 at 12:37:12, Dan Andersson wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>>It would do to have a settings file or somesuch. And Switch it to the optimum at
>>>>>>>>>>>>>once close to the match date. Or A gradual normalisation till the match takes
>>>>>>>>>>>>>place.
>>>>>>>>>>>>>
>>>>>>>>>>>>>Regards Dan Andersson
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>Of course. And can it be forbidden in the contract?
>>>>>>>>>>>>
>>>>>>>>>>>>Of course not!
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>Oh dear. Even the idealists accept it to be "sneaky and underhand" :-(
>>>>>>>>>>>
>>>>>>>>>>>Can't you stay idealist just a little longer?
>>>>>>>>>>>
>>>>>>>>>>>You don't have to be like "them", you know.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>Do you feel like you need to behave in a ideal way when you are faced with a
>>>>>>>>>>dishonest condition?
>>>>>>>>>>
>>>>>>>>>>I don't.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>Nail, head, hit.
>>>>>>>>>
>>>>>>>>>Nor did I. Nor did any young programmer who saw what was going on.
>>>>>>>>>
>>>>>>>>>But, if you then "behave in a (less than) ideal way" you join the corrupt
>>>>>>>>>establishment. And the younger ones see you, and they copy that too, and so it
>>>>>>>>>continues.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>Amateur programmers did not have to wait for me to find ways to kill the big
>>>>>>>>ones with cooked lines in the official tournaments.
>>>>>>>>
>>>>>>>>Not that I have anything against amateur programmers. I was one of them not so
>>>>>>>>long ago...
>>>>>>>>
>>>>>>>>That's life. That's the way it is.
>>>>>>>>
>>>>>>>>If you want to succed, sneaky tricks will never do it for you. But if you don't
>>>>>>>>know the sneaky tricks, you might well never succeed.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>Hence the mess you see now. All the 'players' were idealists once. Now they are
>>>>>>>>>merely corrupt. Don't join them.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>I think that some people need to learn that chess computers and chess computers
>>>>>>>>programmers are not little puppets.
>>>>>>>>
>>>>>>>>Well... At least some of them are not. ;)
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>I didn't explain myself properly.
>>>>>>>
>>>>>>>Ok, try again.
>>>>>>>
>>>>>>>"Do you feel like you need to behave in a ideal way when you are faced with a
>>>>>>>dishonest condition? I don't."
>>>>>>>
>>>>>>>The statement is a universal one. Almost everybody thinks it. And acts on it.
>>>>>>>
>>>>>>>But it has a snowball effect.
>>>>>>>
>>>>>>>If one thinks the consensus behaviour is 'dishonest', then it's ok to be a
>>>>>>>little 'dishonest'. More than ok, one has no choice.
>>>>>>>
>>>>>>>Then the consensus behaviour becomes more dishonest, and so on. Whether this is
>>>>>>>in actual chess game play, off the board play, newsgroup behaviour, commercial
>>>>>>>behaviour, whatever.
>>>>>>>
>>>>>>>Why I said nail, hit, head, was because I believe this is what happened in
>>>>>>>computer chess. Maybe the snowball now reached the bottom of the hill.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>The purpose of the thread I have started is not to promote unethical behaviour.
>>>>>>
>>>>>
>>>>>Clear.
>>>>>
>>>>>>The purpose is to show that the condition "Kramnik must have the program 3
>>>>>>months before the event" is not a show stopper.
>>>>>>
>>>>>>It is a stupid condition that can at best only backfire against the organizers
>>>>>>and Kramnik, and thus should be removed as soon as possible.
>>>>>>
>>>>>>I think it is now clear that:
>>>>>>
>>>>>>1) it is interpreted as an attempt to cheat
>>>>>>
>>>>>>2) it will FAIL and not help Kramnik at all
>>>>>>
>>>>>>It is a DOUBLY STUPID requirement.
>>>>>
>>>>>I don't agree. It was obviously an attempt to get round the earlier objection
>>>>>raised by Kasparov and others that he needed to have some prior knowledge of the
>>>>>play style he would be facing at the time of the DB match. That is not an
>>>>>unreasonable requirement - after all, in high level human chess, players study
>>>>>each others games, work on opening preparation and so on. The concept of
>>>>>'playing the opponent' may not be one recognised in computer chess, but it is of
>>>>>vital importance at GM level.
>>>>
>>>>
>>>>
>>>>I agree with you. Asking for games in correct.
>>>>
>>>>No problemo.
>>>>
>>>>
>>>>
>>>>
>>>>>Where the so-called organisers failed was in demanding an unchangeable program
>>>>>copy - because of deterministic factors this gives the human player an
>>>>>unreasonable advantage, almost equivalent to a brain-scan and carbon life-form
>>>>>disassembley, if that were possible.
>>>>>
>>>>>Kramnik could reasonably ask for a large selection of games played by a single
>>>>>copy of the program, preferably against humans, for study beforehand,
>>>>
>>>>
>>>>
>>>>This is OK.
>>>>
>>>>
>>>>
>>>>
>>>>> and also
>>>>>insist that the program executable was not modified in any way from that point
>>>>>on.
>>>>
>>>>
>>>>
>>>>This is not OK.
>>>>
>>>>The argument that Kramnik itself cannot be frozen as been used against me
>>>>recently, and it is valid point.
>>>>
>>>>If Kramnik cannot be frozen, why should the program be frozen?
>>>>
>>>>
>>>
>>>It depends what you mean by 'frozen'. Of course the programmer would be free to
>>>develop his program. I would argue that the executable used to create the
>>>example games was frozen, in a safe or whatever, until the match.
>>>
>>>If you were able to argue that program development was only bug fixing or
>>>'minor' improvements that did not affect play style away from the example games,
>>>then I guess nobody would complain - but, how would that be policed?
>>
>>
>>
>>Why should this be policed at all ???
>>
>>
>
>Simply because, if there was an agreement to not change style and a commercial
>contract and a seven figure sum, there's the usual problem of trust (lack of).
>
>>
>>
>>>Do you think a sudden and dramatic style change from the example games would be
>>>acceptable? Me not.
>>>
>>>I take your point about 'freezing' Kramnik. However, Kramnik is not going to
>>>change as dramatically as a program could over the time scale. Is he?
>>
>>
>>
>>Of course he will. If you want a brilliant example, look at GK-DB match II.
>>
>>If is he allowed to have the program in advance, he would play ANY playing style
>>that leads to a won game learned by heart.
>>
>>So now you are in favor of book or program cooking by humans?
>>
>
>Of course not. How can you suggest that?
>
>Couldn't you create a Tiger that didn't play the same game twice? There's a lot
>of opening moves aren't there?
The problem is that a chess program is not an unique entity.
The copy of the chess program that the human player is going to test for three
month is going to be different than the one that will play the real match.
The program that will play will not have the "experience" of the one that the
human player will have used.
So the programmer will have no idea about the bad opening lines the human player
will have discovered.
Under these conditions, it is extremely difficult to have a book wide enough to
prevent the human player to reproduce at least one won game.
From a practical point of view, no book will be wide enough and there is going
to be massive book cooking from the human player.
>>>Isn't the point to operate in a fair way and to the 'usual' human methods.
>>>Preparation by game study - to make this work in human-computer, the programming
>>>side simply has to accept an executable freeze for N weeks before the match. How
>>>long N is, is up for discussion.
>>
>>
>>
>>We can discuss this, but if I make a program that changes its playing style
>>automatically all the time:
>>
>>1) how are you going to prove it ?
>>2) how are you going to forbid me to do so ?
>
>Probably the only way would be an industry code of conduct. That's why I would
>ask programmers to sign up to make programs that would not do dynamic style
>changes. Of course you would have different style options, but once a 'style'
>was selected, the program stayed with it.
It is impossible to enforce and unfair for the software players.
>>Are you going to insist on DETERMINISM from chess computers? This is going to be
>>a problem for SMP programs, you know.
>>
>
>I didn't see a deterministic program yet. But these are random move selection
>changes, often inexplicable by the programmer even. Not the same as a whole
>style alteration.
>
>>
>>
>>
>>>>> Such a condition simply places the match on a normal strong human-strong
>>>>>human level of preparation. The opposition program is of course also able to
>>>>>have prepared on Kramnik's game history.
>>>>>
>>>>>>
>>>>>>If it is not removed, then who cares? It is not a valid reason for a chess
>>>>>>programmer to withdraw.
>>>>>>
>>>>>
>>>>>Which brings me to the second point; namely that your initial post throws up how
>>>>>easy it is to prepare a random-style cyber opponent - one that a human cannot
>>>>>make normal preparation against.
>>>>>
>>>>>In the spirit of how chess games are played, I believe that the computer chess
>>>>>community, if it wishes to be above-board in its behaviour would make a rule
>>>>>that outlawed such a beast. If it doesn't it can probably forget matches against
>>>>>strong humans.
>>>>>
>>>>>Proposal for rule 1 of computer-human chess matches: "no play style changes".
>>>>>
>>>>>Interesting to know which programmers would 'sign-up' to such a charter.
>>>>
>>>>
>>>>
>>>>I will never sign up on this.
>>>>
>>>>I think one important improvement in the future will be polymorphic programs.
>>>>
>>>>I would agree on a limitation on the number of processors and speed of the
>>>>computer, no problem.
>>>>
>>>>But I want to be free about the playing style of my program, until the last
>>>>minute.
>>>>
>>>
>>>Then you probably disqualify yourself from playing strong opposition in
>>>exhibition matches. The human players simply won't accept it, and there is no
>>>reason why they should. They'll argue that you take from them their ability to
>>>'prepare' as per normal before the match.
>>
>>
>>
>>Poor little things, those GMs...
>>
>>
>
>Unnecessary comment.
Probably means that you do not understand it.
GMs are so fragile that they need their software opponent to be handicapped?
>They aren't robots, their acceptance of terms is needed. Hence some attention
>needs to be paid to their views.
And no attention needs to be paid to the views of chess programmers?
>>>One day, it is possible that programs will be able to outplay the best humans.
>>>If, during this process, programmers generate these polymorphic monsters as a
>>>means of confusing humans; then humans will simply refuse to participate. Why
>>>weightlift against a fork lift truck arguments, et cetera. Any victory will be
>>>tainted by counter accusations of winning by confusion, et cetera.
>>>
>>>Either the current AI techniques are going to work in computer chess, or they
>>>are not. Don't try to make additional 'progress' by confusion techniques. It is
>>>not fair, and it won't be seen to be fair; however appealing it may be as a
>>>technical can-do. Basically it will come across as a cheat.
>>
>>
>>
>>
>>OK, so in order to be fair I want human players to be disqualified if they play
>>the Stonewall against my program.
>>
>>I have already told you that IMO a limitation on computer power is fine with me.
>>
>>But please let me do what I want to do with the software part.
>>
>>
>
>Ok, your position is clear. You require total freedom of action up to the last
>minute to present whatever software engine you wish.
>
>Why?
>
>Or, I rephrase the question. You are one side of the equation. A human is on the
>other. He may argue for some conditions, which he considers totally fair and
>reasonable. You are going say "I do what I want"? At the expense of no match?
>
>No compromise?
I happily accept to compromise on the power of the computer that will be used.
That must be the third time I am going to repeat this in this thread:
Hardware limitations are fine with me.
I would accept a rule that states that the computer must not have more than X
processors, each processor being no faster than Y MHz, the opening book being no
bigger than Z Mb, and so on.
The hardware limitations are fine with me because the human player has hardware
limitations too.
The software limitations are not fine with me because the human player does not
have any.
The human player, even if he has a limited "computing" power, is allowed to be
as creative as possible. Why would the programmer be forbidden to be?
I consider a creativity contest as fair and interesting, so I do not accept any
limitation of creativity (what I put in the software).
Tell me, how could possibly the other side of the equation refuse the creativity
contest? Is it insulting for the human players if I refuse to be limited in what
I put in the software?
If they want to limit creativity, then no match.
>>>>>Chris Whittington
>>>>
>>>>
>>>>
>>>>You got me. I did not recognize you this time. Sure, a new name coming out of
>>>>nowhere looked a little bit suspect to me, but I did not guess it was you.
>>>>
>>>>
>>>
>>>I wasn't trying to 'get' you.
>>
>>
>>
>>I know, it's just a private joke in this forum. The game is called "Catch the
>>Whittington". You earn points, then when you have enough points you receive a
>>CS-Tal II.
>
>The idea is to turn me into an object, I suspect. Objects are easier to deal
>with.
No, the idea is to turn an annoying thing into fun.
Christophe
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.