Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The failure of validation with DEEP BLUE 2

Author: Robert Hyatt

Date: 09:10:03 07/18/02

Go up one level in this thread


On July 18, 2002 at 07:57:15, Rolf Tueschen wrote:

>On July 18, 2002 at 01:00:11, Robert Hyatt wrote:
>
>>On July 17, 2002 at 17:30:01, Rolf Tueschen wrote:
>>
>>>On July 17, 2002 at 09:23:53, Robert Hyatt wrote:
>>>
>>>>On July 17, 2002 at 05:24:50, Rolf Tueschen wrote:
>>>>
>>>>>On July 17, 2002 at 00:03:24, K. Burcham wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>>In game 2,  Kasparov thought that there was human interference with this line.
>>>>>>He requested the logs to see for himself that these two moves were actually in
>>>>>>the Deep Blue eval.
>>>>>>Kasparov did not think any program would play 36.axb5 avoiding 36.Qb6 or the
>>>>>>move 37.Be4.
>>>>>>
>>>>>>But it seems that todays programs will accomplish what Deep Blue was trying to
>>>>>>do in the game.
>>>>>>Deep Blue opened up the a file and blocked Kasparov's play with 37.Be4, limiting
>>>>>>blacks mobility.
>>>>>>
>>>>>>Below in the examples you will see that two of todays strongest programs will
>>>>>>also accomplish this same objective. Both Fritz7 and Chess Tiger 14.0 will open
>>>>>>the a file and control the a file. also both Fritz7 and Chess Tiger 14.0 will
>>>>>>play Be4 limiting Kasparov's mobility with black.
>>>>>>
>>>>>>All three programs, Deep Blue, Chess Tiger 14.0 and Fritz7 put the
>>>>>>squeeze on black, blocked with the Be4 move, opened the a file, threatened to
>>>>>>capture blacks bishop, forced black to protect the loss of pawns, etc.
>>>>>>
>>>>>>I started each program after Kasparov's move 35...Bxd6.
>>>>>>after  each program analyzed for several hours, I took the line from each
>>>>>>program and played it through to the position after blacks move 40.
>>>>>>This way we can cover both controversial moves>
>>>>>>
>>>>>
>>>>>What is the error in such experiments?
>>>>>
>>>>>Answer: You can't _prove_ something as authentic with repetitions on different
>>>>>machines built-up _after_ the event. History of CC has shown that we could never
>>>>>exclude special preps right on to the point. Therefore, logically, we cannot
>>>>>accept such "proofs".
>>>>
>>>>You _can_ disprove Kasparov's main "claim".  That "no computer would play ..."
>>>>By demonstrating that at _least_ one computer _would_ play that move, his
>>>>statement is disproven for all time.  And the rest of his claim can therefore
>>>>be taken with a mountain of salt.
>>>
>>>Not that I _ever_ was or would be your teacher of English resp. American
>>>English, but I must insist that Kasparov did _never_ say that no computer would
>>>_E V E R_ play these moves.
>>
>>First, let's get the quote right.  He did say "no computer _could_ play this
>>move.."
>>
>>
>>> What he meant was at the time being and he was
>>
>>
>>And here you go on a tangent.  Not saying what he _said_ but what he
>>"meant"...  judge says "inadmissable, that is a conclusion, not a fact."
>
>No, I do not go on a tangent, because always we must interpret language. It is a
>tour de tangent if you pretend to be able to prove exactly what a certain phrase
>means out of itself. You are right, he said it. But what it means, what he said,
>is a question of interpretation. Now, you have a _little_ problem.

I don't have any problem at all.  "no computer could play that move" is not
open to "interpretation".  Each word has a precise meaning.  The
"interpretation" that he later emphasized was "no computer could play such
a move, so DB had human help in the background, somehow."





>
>>
>>
>>>talking about DEEP BLUE 2 in the first place. Now back to the many tries to find
>>>the moves with our commercial or amateur programs. Here I must insist.
>>>Afterwards (with so many possibilities to interfere) you can never prove that
>>>such progs could have found the moves at the time being. But with respect to the
>>>position of Ed Schroeder I would say that even if some alien prog could have
>>>found the  moves, we are talking about DEEP BLUE 2 and we always were. Now - the
>>>deconstruction was the worst thing that could have happened. Because now we
>>>don't have any possibilities at all to corroborate or reject Kasparov's thesis.
>>>This alone was and still is the biggest violation of all ethics.
>>>
>>>Just another thought to think about.
>>
>>See my comment to Amir.  DB2 searched Qb6 and _every_ iteration the score
>>dropped.  Until it dropped all the way to +48 and the program decided that
>>axb5 was better by about 1/10th of a pawn.  No mystery.  No magic.  Just
>>the opinion of the program, right or wrong.
>>
>>It's _all_ in the logs, if you just look.
>
>You know the answer! You yourself explained what "logs" really could mean at
>best. I don't want to re-open that can but I want to mention that after all what
>happened IBM and DB2 team _still_ are in a state of emergency concerning
>"explanations" and evidence. Ok, you can hold this open until the final
>solution, but for me, excuse me, this is not ok. You are in a double bind. DB2
>team has to offer evidence, in special after the deliberate deconstruction of
>the machine. Now we'll never be able to repeat some thought processes. Bad luck
>for IBM/ DB2 team. They didn't pass the doping control. Their World Record can't
>be accepted as valid.
>

It can by those of us that know them.  Otherwise _no_ computer vs human
game/match can _ever_ be trusted.  As I have said many times, there is _no_
way to be 100% certain there is no outside intervention...



>
>>
>>
>>>
>>>We all know that DB 2 did not use the typical nullmove strategies of today's
>>>chessprogs. It would be the least what you and your collegues could do, that you
>>>elaborate what this could mean for the question of Qxb6 vs. axb5. Just for the
>>>sake of our own class of debate. If we could show that DEEP BLUE 2 would have
>>>been a much more difficult task to reject Qb6
>>
>>
>>Just look at their log file, and the variation they produced (note that this
>>is never the full variation on deep blue since the hardware does not supply
>>a PV, meaning you only get the software PV and since it is from the hash
>>table it is not completely reliable nor always complete.
>>
>>The log shows why they rejected Qb6 quite clearly..
>
>Maybe, maybe. Without doping control, not valid unfortunately.


Based on that logic, _nothing_ dealing with computers would be valid, because
ensuring no outside influence would be impossible in the context of the match
between a human and a computer.



>
>>
>>
>>> we would at least be able to
>>>understand why our logic is shaky when we conclude that if some PC progs could
>>>find the solution that then it might be possible for DB 2 as well. You know, we
>>>must not take Kasparov as our scientifical God but we should take him as the
>>>honest reporter from the chess angle of the problem. In other words. Even if you
>>>were right, that his "claim" could be refutated, he could still be right.
>>>Look, if that would come out in 40 years when you and me are dancing in
>>>paradise, it would be a pity if the security patrol would catch you for being
>>>guilty of supporting the confusion to the disadvantage of Kasparov, just because
>>>you are a friend of many of the DB 2 team.
>>
>>
>>I would defend deep blue whether I was _friends_ with the team or not.  It
>>is their _reputations_ that cause me to defend them, not the fact that I have
>>been on a first-name basis with them for 20+ years.
>
>Yes, but it's five years ago now that I told you that they probably violated all
>ethics of science, no matter if they were working for IBM or not. The details
>after game two speak against them. They should have known better. Alas, when Hsu
>Feng tried to get in touch with Kasparov we saw what his reputation was worth.
>He left the field for good. He will never more be accepted from a top
>chessplayer. What did Kasparov say? "You must qualify yourself, then come back."

I wonder why Kasparov would not say "yes, I will play your machine for the
easy opportunity to pick up a million dollars."???

I can think of several reasons why he wouldn't.  And the main one _did_ have
something to do with Hsu's reputation.  Because Hsu told him he would have
a machine another 30X faster.  I don't think Kasparov wanted _any_ part of
that...




>What does that mean? Oh, nothing but the _fact_ that only with the friendliness
>of Kasparov your friends got the chance to play. And then they violated the
>silent mutual acceptance that "normally" a show like that would never happen,
>simply because it's unfair to the bones.

You don't think the chance to win $1,000,000.00 had _anything_ to do with
Kasparov's decision?  It had _nothing_ to do with friendship.  It had
_everything_ to do with $$$...

Otherwise don't you think Friedel would have set up a match between Kasparov
and Fritz?  Eh?



>
>You know, Bob, it's kind of strange, that you had to come forward with the
>"contracts". Of course you are right with the contracts. But at that moment you
>lost the track of the silent mutual acceptance. Their is a logic beyond
>contracts. In special with contracts for simple show events!

I was taught _long_ ago.  "if it is important, get it in writing.  If it is
not important, then a wink and a nod is perfectly acceptable."  Kasparov signed
a contract after negotiating the details of that contract.  If you sign a
contract to buy a Rolls Royce, and after you get it, you decide "this thing
isn't worth this much money" it is _too late_.  You negotiate _before_ signing,
_not_ after...




>
>I think that this is the best possible description of what has happened. Without
>insults. The facts alone speak their verdict. Against one certain side.
>They changed a show act into a World Championship and then denied to pass the
>doping control. The End.
>
>
>Rolf Tueschen
>

That is just your characterization.  Not the general opinion.



>>
>>
>>
>>
>>
>>> Friendship is one thing but Science
>>>and Ethics is something more important. And I tell you, the truth will come out.
>>>Now or in future. And you won't have a good excuse with the statement that
>>>Kasparov made a scientifically wrong thesis. We all know that he's not a
>>>scientist. But he's one of the best chessplayers we ever had.
>>>
>>>Rolf Tueschen
>>>
>>
>>
>>
>>And one of the biggest sore losers as well...
>>
>>
>>
>>>>
>>>>
>>>>
>>>>>
>>>>>The deconstruction of DEEP BLUE 2 right after the event, in special with the
>>>>>knowledge of the prior attitude of the DB team, which was one of secrecy (not a
>>>>>single game score existing!), speaks against the validity of DB2 output.
>>>>>
>>>>>THe deliberate deconstruction invalidates DB2 results. Just compare it with the
>>>>>refusal of passing the doping test directly _after_ the race.
>>>>>
>>>>>It's so basic!
>>>>>
>>>>>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.