Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A 2005 Appeal to Bob Hyatt, the Scientist! Tell us the Truth About DBII!

Author: chandler yergin

Date: 09:01:26 04/26/05

Go up one level in this thread


On April 26, 2005 at 11:36:12, Robert Hyatt wrote:

>On April 26, 2005 at 11:01:11, chandler yergin wrote:
>
>>On April 25, 2005 at 20:46:36, Robert Hyatt wrote:
>>
>>>On April 25, 2005 at 19:05:57, Rolf Tueschen wrote:
>>>
>>>>On April 25, 2005 at 15:31:02, Robert Hyatt wrote:
>>>>
>>>>>On April 25, 2005 at 14:39:40, Rolf Tueschen wrote:
>>>>>
>>>>>>On April 24, 2005 at 22:56:48, Robert Hyatt wrote:
>>>>>>
>>>>>>>On April 24, 2005 at 18:59:35, Rolf Tueschen wrote:
>>>>>>>
>>>>>>>>On April 24, 2005 at 15:42:54, Robert Hyatt wrote:
>>>>>>>>
>>>>>>>>>On April 24, 2005 at 13:13:22, Rolf Tueschen wrote:
>>>>>>>>>
>>>>>>>>>>On April 24, 2005 at 10:13:18, Robert Hyatt wrote:
>>>>>>>>>>
>>>>>>>>>>>On April 24, 2005 at 05:14:48, Peter Berger wrote:
>>>>>>>>>>>
>>>>>>>>>>>>On April 23, 2005 at 23:52:36, Robert Hyatt wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>>Several of us have looked at the logs for the games, and game 2 looked perfectly
>>>>>>>>>>>>>normal and the program even reported a fail low and "panic time" where it
>>>>>>>>>>>>>searched longer than normal because of the fail low.
>>>>>>>>>>>>>
>>>>>>>>>>>>>This is a red herring and crap.
>>>>>>>>>>>>
>>>>>>>>>>>>When was the first time someone independent had a look at these logfiles? Have
>>>>>>>>>>>>you and the several others had a chance to look at the logfiles right after the
>>>>>>>>>>>>games took place, say May 1997?
>>>>>>>>>>>>
>>>>>>>>>>>>The logfiles IBM published eventually mean nothing at all. This was more than a
>>>>>>>>>>>>year after the games, wasn't it? Even I could produce most impressive logfiles
>>>>>>>>>>>>given that much time ..
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>Yes.  Several looked at the log right after the event.  I believe that Ken sent
>>>>>>>>>>>me the section from the game although I don't remember whether it was the Qxb6
>>>>>>>>>>>(not played) or the Be4 position.  I believe that Amir posted something about
>>>>>>>>>>>the position early, but his comments were based on either not understanding what
>>>>>>>>>>>DB's log output meant, or something else.
>>>>>>>>>>>
>>>>>>>>>>>This was about the "fail low (panic time)" that caused DB to search much longer
>>>>>>>>>>>than normal and may have been on the Be4 move although I simply don't remember
>>>>>>>>>>>much about it since it was not a particularly significant event in my mind
>>>>>>>>>>>because at the time I posted an excerpt from a Crafty log that looked
>>>>>>>>>>>_identical_ in concept.
>>>>>>>>>>>
>>>>>>>>>>>Nothing ever looked strange about the log stuff to me...
>>>>>>>>>>
>>>>>>>>>>The point is that you explained at the time that it would cost almost no energy
>>>>>>>>>>to fake the output so that it looked koscher. So - what do you want to say with
>>>>>>>>>>the above. In truth the later outputs mean nothing in a judicially relevant
>>>>>>>>>>sense. The question is why they didn't discuss the thing with Kasparov in time
>>>>>>>>>>before the following game so that the event could have been continued as
>>>>>>>>>>planned. If they didn't have to hide something.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>OK, we go full circle.  I could certainly produce a fake log to support any move
>>>>>>>>>that anybody would want to see Crafty play.  As I said back in 1997, it is
>>>>>>>>>_impossible_ to prove they didn't cheat.  But, the other side of the coin is
>>>>>>>>>that it would be possible to prove that they did, all one needs is to know the
>>>>>>>>>methodology or catch them in the act, or whatever.  But this was never proven.
>>>>>>>>>In fact, no methodology for cheating was ever mentioned.  Normally when you
>>>>>>>>>accuse someone of cheating, you state the violation clearly.  "He spit on the
>>>>>>>>>ball".  "His raquette head is too big".  "His golf club face is improperly sized
>>>>>>>>>or weighted"  and so forth.
>>>>>>>>
>>>>>>>>
>>>>>>>>Exactly, and that does prove at least to me that Kasparov didn't exactly accuse
>>>>>>>>them of chesting. How could or should he? All he did was making public his
>>>>>>>>astonishment, his disbelief. Wouldn't it have been ok if they had cared for him?
>>>>>>>>In what style or with what material they should have done this, that is a
>>>>>>>>different question. But why a player like Kasparov shouldn't have been allowed
>>>>>>>>to express his confusion? Exactly with your explanation from that time Kasparov
>>>>>>>>could have been given back his peace and new motivation to continue the fight.
>>>>>>>>Without any humiliations. And with the hope to get a valid result of the match.
>>>>>>>>(Perhaps I'm missing a specific lingual undertone in your messages to me, so
>>>>>>>>that I'm still missing your exact position in the debate, but I can't figure out
>>>>>>>>why it should be so difficult to get the sense of what I'm saying. Couldn't you
>>>>>>>>explain, why in your eyes my science argument isn't good in respect to the
>>>>>>>>original question of the research, namely could a machine play better than the
>>>>>>>>best human? Why do you throw it into the bin? Are you really believing that a
>>>>>>>>win, also a win with certain unfair methods, could prove anything at all?)
>>>>>>>
>>>>>>>
>>>>>>>As I have said before, it proved just one thing.  That the human world champion
>>>>>>>could be beaten by a computer in a standard time-control match on that
>>>>>>>particular day.  There are plenty of other questions that could be asked now.
>>>>>>>But the question back then was asked, and answered...
>>>>>>
>>>>>>I disagree. That wasn't proven at all. If you psyche out a human chessplayer
>>>>>>with denying talks then you can't claim that your chess was better and therefore
>>>>>>you've beaten him. This is impossible to conclude at least in my books. Remember
>>>>>>that we are talking about machine vs human chessplayer chess, we are NOT talking
>>>>>>about normal chess. Of course there the psychology is part of the game. But NOT
>>>>>>between machines and humans. Why? Because machines have no psychology and a team
>>>>>>of human operators is making the psycho job. Which is fair enough in
>>>>>>championships but NOT in science events... Simple, no?!
>>>>>
>>>>>Nope.  When you play chess, you have to play no matter how you feel on the day
>>>>>of the match.  Personal problems?  Deal with them.  Illness?  play or forfeit.
>>>>>You simply have to play, period.  And that is a part of chess.  I've played
>>>>>tired, because I had to drive a long distance prior to the start of round 1.
>>>>>I've played tired because of staying up too late the night before the round.
>>>>>I've played with a fever.  That's just a part of chess.  Not a part that the
>>>>>computer has to deal with of course, but then computers lose due to remote power
>>>>>failures, or communication failures, or outright program bugs, etc...
>>>>
>>>>Of course - I agree with you. This is chess and also computerchess of machines
>>>>against human chessplayers if it's for a form of championship. Of course then a
>>>>human player would never accept 6 games in 8 days. -
>>>>
>>>>But NY 1997 between IBM/Hsu and DBII and Kasparov, that was NOT about a
>>>>championship. Therefore Kasparov didn't take it too seriously, simply because he
>>>>KNEW for sure that this machine couldn't be stronger than him.
>>>
>>>I don't know if I'd take the "not too seriously" very seriously.  There was
>>>$1,000,000.00 at stake.  Speaking only for myself, I'd take that _very_
>>>seriously... :)
>>>
>>>
>>>
>>>> It was a science
>>>>experiment. Kasparov figured as the client from the human side, just to guaratee
>>>>a high level of human chess skills. Suddenly Kasparov felt confused because in
>>>>his opinion the machine played something strange. Kasparov demonstrated his
>>>>disbelief.
>>>
>>>The problem here is that nothing "strange" happened.  The moves were reproduced
>>>by other programs, in 1997.
>>
>>You keep using this as an 'explanation'; as some type of justification,
>>and it is not justified, nor defensible.
>>
>>The Programs that 'reproduced' the move did not do so under the same conditions
>>or time control used in the match.
>
>You are really _way_ out of your "depth" in these discussions.  You do realize
>that in 1997 DB was searching 200M positions per second?  And that for example
>Crafty was searching 80K positions per second on a pentium pro 200, or 320K on a
>quad pentium pro 200?  So of course no program of 1997 could find those moves,
>the hardware was too slow.  Programs of _today_ can find those moves in a minute
>or two, as we are nearly 10 years down the road in hardware evolution and speed,
>and today on my dual xeon I'm hitting 2M+ nodes per second, on the quad opteron
>I used last year, I hit 11M+ nodes per second, etc.
>
>You _do_ realize that this is all about speed?  you _do_ realize that in 1997 no
>program was within 2 orders of magnitude of the search speed of deep blue?
>
>I didn't think so...
>
>
>
>>
>>The problem here, is that something "Strange" did happen; which completely
>>and totaly refutes your position.
>>
>>No Top commercial program at the time ever played that move.
>> Some Programs were specifically 'prohibited' from playing that move.
>
>What are you talking about?  We are not talking about Nxe6, in game 6, which was
>a pure _book_ move from deep blue.  We are talking about the Be4 rather than a
>move that appeared to win a pawn outright.  And in 1997, Crafty found that move
>after a search of more than 12 hours.  Today it finds it far faster.  Someone
>even reported Shredder finds it in a minute or so, not on the fastest hardware
>available.
>
>So I have no idea what "prohibited" means.

Hmmm that should give us pause for thought.
Beginning Programmers know.
Ask one.




>
>
>
>
>>
>>All Kasparov had for his pregame analysis was use of these Programs.
>> He gambled thet Deep Blue would retreat the Knight.
>>It's as simple as that
>
>
>So?  In 1997, during the game, Crafty predicted Nxe6.  It was in my book of the
>time, as it is a known book move.  What is your point?  Or do you even have a
>point?
>
>
>
>>
>>Rolf is correct
>>
>>
>>
>>>I remember Crafty finding Be4 for example, although
>>>it took an overnight search on hardware of 1997 to find it.
>>
>>Then why do you keep criticizing Kasparov for believing a 'move' was strange?
>>
>>No Computer before had ever made those moves.
>
>
>So what?  That was the point of the contest.  There had never been a machine
>built that could search that fast.  Wonder if he would have said "they cheated"
>when the Bell X-1 first broke the sound barrier?  After all, no other aircraft
>could reproduce that.  Did they cheat?  That's a totally stupid way to reason
>this out...
>
>
>
>
>>
>>
>>
>>There simply was
>>>not anything that looked wrong to knowledgable observers.  In fact, more DB
>>>moves were criticized during the match, although the moves were sometimes later
>>>found to be the best.  h5 (DB was black) was one such move where later Kasparov
>>>said "that was the only viable try there..."
>>>
>>>I think he felt that this thing was not going to roll over like 1997 Fritz did
>>>when he practiced, and a loss of touch with reality led to the "OK, it can't be
>>>me, and the machine can't possibly be that good, after all Friedel said it was
>>>no better than Fritz, so they must be cheating somehow."  Of course they never
>>>explained who could have possibly given the machine any advice, in real-time,
>>>when they were already playing against the strongest human around.
>>
>>IM Joel Benjamin, that's who.
>
>
>So Benjamin could beat Kasparov in a match, himself?  :)
>
>Been watching too much twilight zone...
>
>
>
>
>> He was in fact the one that kept 'tweaking' the
>>Program, and expanding the Opening Book for Deep Blue, based on his knowledge of
>>Kasparov's style.
>
>He didn't "tweak" the program at all.  He did do some of the opening book
>preparation (but not all, other GMs were involved as well).  He didn't touch
>"deep blue" itself as far as tweaking however.  Don't know where you get that
>from...
>
>
>> That's why he was hired in the first place.
>
>
>No, they hired him because he had a reasonable understanding of computer chess,
>and because he was a grandmaster.
>
>
>
>>
>>>That is
>>>often overlooked.  Would I want to suggest a move to DB when playing Kasparov?
>>>Could I?  Could it possibly be a move good enough to beat him?  It's not even
>>>worth thinking about..
>>>
>>>
>>>
>>>
>>>> Now how the team of scientists reacted? - They saw that their human
>>>>player was confused. "So be it!" "This is a good chance to kill the guy, the
>>>>scientists said. Let's offend him even further and deny him any talks at all!" -
>>>
>>>That is assuming things that are (a) not proven;  (b) unprovable even today; and
>>>really nothing more than "dark-side speculation".  Why do you assume IBM had
>>>dark motives here?  I know the people involved.  I played such a match against
>>>David Levy in 1984.  David would tell you today that we were never "adversaries"
>>>in that sense of the word.  We wanted to win.  He wanted to win.  We both played
>>>by the rules, and respected each other during the match.  I don't believe the DB
>>>team were the ones to "vary" from this.  After all, Kasparov took off the gloves
>>>first and accused them of cheating, in a public forum, leaving little chance to
>>>smooth things over since the Genie was already out of the bottle...
>>>
>>>
>>>
>>>
>>>
>>>>
>>>>Now that is exactly what is incredibly stupid in my eyes. Because, first, DBII
>>>>wasn't stronger than Kasparov in 1997.
>>>So why should the fathers of the machine>
>>>attempt to beat Kasparov right now?
>>
>>Why?  Why? Why?   Are you Naive? You can boldly ask why?
>>
>>Because of the FREDKIN PRIZE!
>> That's why!
>> You didn't know about that?
>>Of course you did, and you deliberatly leave that out as a motive for 'why'
>>the Team was so intent on beating Kasparov?
>>
>>That's shameful.
>>
>>The Fredkin Prize was $100,000 for the first team to build or program
>>a computer that would defeat the World Champion in a Match!
>>Big motivation I'd say, plus the prestige, glory, & PR that goes with it.
>>Why didn't you mention that?
>
>Why don't you grow up?  Would IBM spend _millions_ to make $100K?  What an
>incredibly astute financial plan.  :)
>
>
>
>
>
>
>
>>
>>
>>
>>>They would beat him in the nearer future.
>>
>>Neither you nor they knew that.
>>
>>They weren't taking any chances were they?
>>
>>
>>>
>>>
>>>You can say that all you want.  But all the evidence we have is 6 games played
>>>in 1997.  They suggest that DB was better, because DB won more games...
>>>
>>>
>>>
>>>>But no, they played dirty and suddenly the event changed from a science thing to
>>>>a killing party.
>>>
>>>
>>>Again, a wild stretch.  How did "they play dirty" when they were accused of
>>>cheating, a multi-billion dollar company with an impeccable reputation, and now
>>>they stand on the stage and are accused of cheating in front of the entire
>>>world?  And "IBM played dirty"???  That I can't reconcile with events that
>>>actually happened in 1997.
>>
>>We lots of other people can, and IBM certainly does not have an impeccable
>>reputation now.
>
>
>IBM _does_ have an impeccable reputation today.  Kasparov does not.  The
>cheating claim against IBM.  The cheating claim against him, caught on tape,
>against Polgar.  Temper tantrums multiple times.  Battles with FIDE.  Yes, he's
>a great moral example...
>
>
>
>
>>>
>>>
>>>
>>>>You came and assisted them: "THat is totally conform with the
>>>>rules of the "match"!" - Yes, fine, but what is with the science research thing?
>>>
>>>
>>>There were two components to this match.
>>>
>>>1.  Hsu and company (not IBM company but his associates such as Murray, Hoane,
>>>etc) wanted to build a chess machine and beat Kasparov.  That was their goal in
>>>life.
>>>
>>>2.  IBM watched interest in computer chess over the years, and saw an
>>>opportunity for some publicity and public relations opportunities that would far
>>>exceed the potential cost of underwriting the project.  The company would prefer
>>>to win, for the marketing advantage the free publicity would generate, but they
>>>were happy to play in an event that everyone cast in a positive light, to
>>>showcase their hardware and their interest in "tackling difficult problems with
>>>innovative technologies."
>>>
>>>So Hsu wanted to win.  IBM wanted publicity.  That's all there was to it.  Oh
>>>yes, IBM wanted _positive_ publicity.  When things turned sour, as I was talking
>>>with Monty on the phone, we both said "this will be the end of these no matter
>>>what happens."  That was prophetic...
>>>
>>>There was no grand scientific experiment to beat the best human with a set of
>>>rules you or others would like to see.  For this experiment, the goal was to
>>>throw massive amounts of hardware at Kasparov to see if it would beat him.  Not
>>>beat him after he had a chance to test against the thing for months.  Not beat
>>>him after he had a chance to study games played against other opponents.  Not
>>>beat him with any kind of boundaries on the contest.  The goal was simply:
>>>
>>>"beat kasparov".
>>>
>>>Period.
>>>
>>>Later other questions might be asked, such as "could the thing play through the
>>>candidate's cycle and work its way up to challenge the world champion for the
>>>real WC title?"  And so forth.  But that simply wasn't the question, nor the
>>>rules, nor the interest at the time.  Until then, the WC had _never_ lost a
>>>serious match against any computer.  It was not expected even in 1997, although
>>>many of us knew that the DB box was very strong and it wouldn't take much to
>>>cause a human to cave in to it...  But we didn't expect it to happen in 1997...
>>>
>>>Most of us had predicted this to happen somewhere after 2000...
>>>
>>>
>>>
>>>
>>>
>>>>Ok, let's make suicide with DBII, let's win ugly the match and then leave
>>>>computerchess for good...! And Bob says (still in 2005), well, DBII was better
>>>>than Kasparov, just during that w2eek, probably not overall, but for the period
>>>>of these 6 games!
>>
>>
>>
>>>
>>>
>>>I stand by that.  Because _all_ the evidence we have supports that conclusion.
>>>Kasparov lost the match in 1997.  Hence for that 2 week period, DB _proved_ it
>>>was better OTB.  Didn't prove it was better than Kasparov at his best, or
>>>kasparov at his worst, or anything in between.  Just proved itself to be better
>>>over those 6 games...
>>>
>>>
>>>
>>>
>>>>
>>>>Bob, you as a scientist, you can't do that. Please, don't do it. Come back to
>>>>science.
>>>
>>>It is the only thing any _real_ scientist could conclude, sorry...
>>>
>>>6 games is not a lot, but if you look at what we have, the games certainly don't
>>>prove Kasparov was better.  That would be blatantly ignoring the actual data...
>>>The games don't prove DB was better than Kasparov, although there is a
>>>suggestion that is true.  But the games _do_ prove that for that two week period
>>>of time, DB was better than Kasparov, whether he was at his best or not is not
>>>an issue...
>>
>>It's exactly the issue, and you know it!
>>
>>Kasparov gambled and lost.
>>You know that.
>>Hsu knows that.
>>
>>
>>Deep Blue was not out of it's Opening Book until move 11.
>>You know that too.
>
>So what?  Kasparov made the mistakes in the game, not Deep Blue.  Of course
>Kasparov also resigned a drawn position, made other mistakes in his play in
>other games.  Etc...



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.