Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Resumee: DB2 was designed without sound science

Author: Robert Hyatt

Date: 10:18:24 07/23/02

Go up one level in this thread


On July 23, 2002 at 11:58:42, Rolf Tueschen wrote:

>
>It's indecent and insulting to use metaphors of being able to follow a cow
>walking down the street _in case of a physically handicapped_! You haven't
>learned your kindergarden lesson good enough yet. You're simply not a gentleman,
>neither in the case of Kasparov vs DB2 team and IBM in 1997 and here in our
>debate. Period.
>
>

You brought up the word "follow".  I pointed out that I don't think you
know what "follow" means.  Particularly when you can't follow the technical
discussion about proving that axb5 and Qb6 are considered equal by more than
a couple of programs now...



>
>You don't get the chess relevant and scientific question! A typical type of
>answer from you. The wanted to win at all costs and therefore it's ok that
>something is hidden and therefore the impossibility to do research doesn't
>matter. And you want to insult me for being stupid several times in your
>posting? You leave science although being a scientist, you leave fairness in
>sports and then you insult the critic of such a strange happening and let me
>follow in vain a cow walking down the streets. This speaks for itself.
>
>

As I said, their primary _goal_ was to build a fast machine.  One of the
things they gave up as a result was the ability to back up a PV from the
hardware to the software part of the search.  That was simply a constraint
they had to work within.  The DB circuitry was _very_ close to not fitting on
the chip as it was, Hsu's book will tell you just how close it was when it
becomes available.  So they lived within the constraints they had, and made it
work as it did.  As I said, the _goal_ was to "go fast.  As fast as possible."
Nothing more, nothing less.  They made compromises to make that happen.  Same
thing happens _everywhere_.  Wouldn't it be better if airplanes had steel frames
and skins?  It is _much_ stronger.  Of course, you might only carry a dozen
passengers on a 747, but that is not important compared to making it as safe as
possible, right?  Another "compromise".  It happens daily.  It happened to them
and for completely non-sinister reasons.




>
>No need for the long affirmation. You should better find some explanation for
>for the tradition of CC, of not allowing exact research on the thought processes
>and output of the machine. I'm not a programmer, I can only make the
>scientifically correct analyzes on the base of your statements.
>
>Let me ask a theoretical question: if someone, say Murray Campbell came forward
>and explained that they played a dirty psycho game from the beginning on when
>they designed the first machine, would you still be happy about the factual win
>in 1997? And if no, by chance, could you give us some less serious faults which
>would influence you to change your position. From when on you would no longer
>defend the DB2 team and IBM? Just asking you as expert and chessplayer.
>

I am neither happy with the win, nor disappointed.  I wasn't involved.  I was
surprised, but nothing more.  There are constraints they lived with.  Parallel
search is but _one_ of them.  However, ask _anybody_ working on a parallel
search about "can you get rid of the non-deterministic behavior?" and see what
answer you get.  Perhaps by the 10th or 15th "no" you might begin to notice
a trend and finally conclude that "this is normal"...




>
>Are you serious? I told you that Kasparov didn't mean "no computer ever and no
>matter how tweaked". You are simply making a classical mistake. You can't prove
>"axb5" with such testing. Are your machines independantly finding axb5 or not?
>How could you say yes, if you had 5 years for the creation of the machines? If
>you know what I mean. This isn't sound science and logic.


kasprov said "no computer can make this move."

Deep Blue's log shows that Qb6 was best at first, but dropped each iteration
until axb5 was slighly better and was played.  Crafty and Fritz are _both_
showing the same thing and _neither_ has been "tweaked" to do so.  Just standard
crafty and standard fritz.  Again, _you_ want to make the conspiracy deeper by
suggesting that we "tweaked" the programs to produce DB's move.  There is no
way to "tweak" fritz that I know of.  And anybody can try the test with any
version of Crafty they want to see if the two moves are really +3.0 apart...





>
>
>>
>>Whether deep blue had human help or not is not the issue for _this_ test.
>
>Who had said that? I didn't.

kasparov did...

>
>Objection. Nor has Kasparov meant what you want to imply, nor could machines of
>today prove that DB2 could have found it too, the move axb5. Or are you
>correcting your earlier positions?


Nothing can prove that DB _did_ find it.  But _any_ program can prove that it
is _possible_ that DB found it with no help.  +that+ is the point.  "no computer
can ..."  -> "a computer can" means the "no computer" is wrong.

That is _all_ we are looking at.





>
>
>Again, you miss the point. Many of the moves criticised are of that sort.
>Kasparov wanted to say that other moves are not better, but it doesn't mean that
>g5 is a good move. It's a bad move, but the only one for K. Is that ok for you?
>Logic isn't so simple, I agree.

Kasparov said "g5 is the _only_ move black can play and maintain any hope
of staying in the game".  That is _clear_ as to its meaning.  It does not
mean it is bad.  It does not mean other moves are not better.  It means that
move is _the_ move to play...  He said he would play it himself.  What more
can there be???


>
>Then let me say a simple truth. No chessplayer would behave like that. If CC
>members behave like that it might be a special hybris but it's wrong and very
>strange. We must spend more thoughts for the many CC lovers. Human chessplayers
>should be treated with respect and the CC fans should be treated with respect
>and gratitude. You did so many goods but sometimes you behave like the famous
>elefant damaging china. If you know what I mean. Let's not escalate this. Just
>try to respect my questions from science.

Sorry.  _I_ behave like that.  In fact, I'll bet that _nobody_ saves the logs
from every game they play.  Had I done that I would now be the owner of over
1,000,000 log files.  I don't even know how to manage that many, much less find
something important in one of them.  I _always_ have the most recent 300 logs
from Crafty.  But no more.  Sometimes this is 1-2 days worth of logs, sometimes
1-2 weeks.  But I keep what I can work with, what is important at the time, and
I cull the rest.

No mystery about how or why.





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.