DEEP BLUE did it...
Ed Schroder
My congratulations to IBM and the Deep Blue team!
As the first chess program in history Deep Blue has beaten the human
world champion, a historic moment.
No matter all the speculations they did a fantastic job.
The dream of Hsu and co has become reality.
My warm sympathy goes to the Deep Blue team, IBM amd of course Kasparov
who made this all possible.
I am already looking forward to the "revenge match".
- Ed Schroder -
My "special" congrats goes to Bob... :)
chrisw
Ed Schroder <rebc...@xs4all.nl> wrote in article
<5l6rsr$14i$1...@thor.wirehub.nl>...
I agree, but something seems not right:
Some of the PC programmers talk to each other, I'll use the word 'we',
although it doesn't cover everybody.
We know that the 4-5 best PC prgs are searching 30,000 to 200,000 nps.
We know that DB is searching what, 1000x faster than this ?
We know that the 4-5 best PC prgs are going 9,10,11 ply plus various
extension tricks in tournament times.
We think that DB is doing 15,16 ply plus various extension tricks in
tournament times.
So far, so good, it all makes sense: 1000x faster gives another 5-6 ply
and/or extension trickery equivalant.
We suspect that DB has an evaluation function not as stupid as the fastest
PC prgs, but not as knowledge packed as the slower PC prgs - but who knows.
I personally suspect that they don't have anything dramatic in the eval,
just normal stuff.
We know from AEGON and elsewhere that GM's and IM's who have bothered to
understand prgs behaviour win just about all their games. And that all the
evidence says that top PC prgs are around 2400 or so.
We know that 'our' PC prgs have a very long way to go before they are going
to seriously challenge the good IM's and GM's. We don't think that nps is
going to help that much, that basically knowledge is needed.
We also have the idea that a speed doubling gives 60 ELO or so at lower
plies, but that this ELO increase falls off at deeper plies. We had assumed
that this fall off meant that the ELO increase topped out BELOW Kasparov
grading - here we may have been wrong.
Nevertheless this is what doesn't add up for me:
1. I don't see the 1000-fold speed increase and the extra 5-6 plies being
enough to raise a standard from 2400, and difficult to fight booked-IM's
and GM's, to beating Super-GM, World Champion Kasparov.
2. I see questions about Kasparov's choice of moves, and his style of play,
and his resignation of a drawn position, and his failure to win from a
powerful position (Game 4?), and his playing of a losing book line. He was
not playing as a super-GM.
It's disturbing.
Chris Whittington
>
Ingo Althoefer
Ed Schroder <rebc...@xs4all.nl> wrote
>My congratulations to IBM and the Deep Blue team!
>As the first chess program in history Deep Blue has beaten the human
>world champion, a historic moment.
>
>No matter all the speculations they did a fantastic job.
>The dream of Hsu and co has become reality.
>
>My warm sympathy goes to the Deep Blue team, IBM amd of course Kasparov
>who made this all possible.
>
>I am already looking forward to the "revenge match". - Ed Schroder -
My congratulations to IBM and the Deep Blue team, too !
You have made big steps for chess ( although they may be only small steps
for the pawns ...). Thank you.
Especially I like the imaginative defenses of Deep Blue in games 3, 4, and 5.
Kasparov had some advantages in all three games but DB was strong enough to
prevent him from transforming these in full points. Very nice defense.
Kasparov made one interesting comment on Deep Blue's style of play: "...The
opponent was constantly changing...". Maybe the Deep Blue team had prepared
this before the match. When you have a chess program with many parameters
there is typically not only one "quasi-optimal" setting, but several rather
different ones ( almost same strengths, but completely different playing
styles ). Now collect for instance 6 such settings before you let your
computer play a 6 games match versus a human player. Use setting 1 in game 1,
setting 2 in game 2, ..., setting 6 in game 6. This "moving target" approach
makes it very difficult for the human to find out weak spots of "the" program
and to concentrate on them.
1996, during the AEGON tournaments there were rumors that GM Seirawan was
offering a 10,000 US-$ bet that he would not lose a 24 games match in blitz
chess ( 5 minutes per player for the whole game ) against any commercial
chess program. We discussed and arrived to the following computer strategy
for such a match:
Find 24 different good settings for the parameters of your commercial
program, and use them, one in each game, in this match. I don't know if
someone is going to realize it ...
Back to the DB-GK match.
I also like the offer Deep Blue's team leader C.J. Tan made to Garry Kasparov
directly after the match: "...We will continue our partnership with Garry, but
perhaps on a l e s s c o m p e t i t i v e level..."
This is exactly what chess needs: Take a strong human ( like Kasparov ) and
take a strong chess computer ( like Deep Blue ) and let them collaborate.
Shouldn't such a symbiosis be able to play chess on a fantastically high level?
Garry Kasparov's direct answer was reluctant. But such man-
machine combinations will become the dominating chess playing entities in
the future. Think of:
o Kasparov+DB versus Anand+DB
o Kasparov+"old"DB versus "new" Deep Blue
o Kasparov + Deep Blue ( without an opponent ) producing a chess game as
a piece of art
As said before, GK's direct answer was reluctant. But give him some weeks to
find in the new realities... He was one of the first GM's realizing that
small computers like Fritz are helpful in analysing. In his subconsciousness
he may see already the potential of Deep Blue as a partner ( and not as an
opponent ).
Of course, also other chess masters instead of GK should be able to perform in
such symbiotic settings, for instance experienced "old" GM's like Bronstein,
Smyslow, Fischer, Darga, .... Darga would have one special advantage: he has
been working in IBM for about 30 years.
And don't forget ambitious amateurs like me: I would like to be the Boss of
Deep Blue in a 2-best mode...
These man-machine combinations are also useful in many serious fields outside
chess: drug design, air flight planing, chip design, scheduling problems, ...
Ingo Althoefer.
PS ( for ICCA comrades only ): When David Levy became our President in 1986,
he gave us two tasks:
" Double your membership -- beat Kasparov ! "
Number 2 is solved now, so let's go for No. 1!
Joe McCaughan
chrisw (chr...@cpsoft.demon.co.uk) wrote:
: Ed Schroder <rebc...@xs4all.nl> wrote in article
: <5l6rsr$14i$1...@thor.wirehub.nl>...
: > My congratulations to IBM and the Deep Blue team!
: >
: > As the first chess program in history Deep Blue has beaten the human
: > world champion, a historic moment.
: >
: > No matter all the speculations they did a fantastic job.
: > The dream of Hsu and co has become reality.
: >
: > My warm sympathy goes to the Deep Blue team, IBM amd of course Kasparov
: > who made this all possible.
: >
: > I am already looking forward to the "revenge match".
: >
: > - Ed Schroder -
: >
: > My "special" congrats goes to Bob... :)
: >
: >
: It's disturbing.
: Chris Whittington
For this reason I'm leaning toward the "something was seriously wrong"
theory myself. I went over an old GK vs. Deep Thought game in which
Gary crushed the silicon beast. This game, (sorry, I ddont have it), was
so tactically deep that I could not understand the moves without analysis.
Game 6 in the current match really looked poor. As a 1400 player I cannot
understand why Gary locked in the KB with Qe7. T could have been
placed at c7 as another poster analyzed recently. I also dont understand
how Gary, probably rated 300 points higher than most good PC's, could
miss a draw in game 2.
--Joe McCaughan
the drowning man
Ed Schroder (rebc...@xs4all.nl) wrote:
: My warm sympathy goes to the Deep Blue team, IBM amd of course Kasparov
: who made this all possible.
Um, you may have chosen the wrong word here.
tdm
--
"You know how women always say that men aren't emotionally available.
Well, a lot of women aren't emotionally available. It's like, if you're
vulnerable, we say, 'Look, we need you to be sensitive.' So you become
sensitive, and then we go, 'You've got no fuckin' backbone,' and we kick you
in the face and run off with a ski trainer."
--Tori Amos
Komputer Korner
chrisw wrote:
> Nevertheless this is what doesn't add up for me:
>
> 1. I don't see the 1000-fold speed increase and the extra 5-6 plies being
> enough to raise a standard from 2400, and difficult to fight booked-IM's
> and GM's, to beating Super-GM, World Champion Kasparov.
>
> 2. I see questions about Kasparov's choice of moves, and his style of play,
> and his resignation of a drawn position, and his failure to win from a
> powerful position (Game 4?), and his playing of a losing book line. He was
> not playing as a super-GM.
>
> It's disturbing.
>
> Chris Whittington
>
> >
I agree except that the top PC's are a little stronger than you give
them credit for. However I too am mystified at Gary's play. His
strategy of anti computer chess was flawed from the beginning. I sure
hope IBM give Karpov a chance. He has the official title after all.
--
Komputer Korner
The inkompetent komputer.
Jim Gillogly
Komputer Korner wrote:
> I agree except that the top PC's are a little stronger than you give
> them credit for. However I too am mystified at Gary's play. His
> strategy of anti computer chess was flawed from the beginning. I sure
> hope IBM give Karpov a chance. He has the official title after all.
It's not obvious to me that the strategy of anti-computer chess was
flawed from the beginning. Kasparov has felt since before last year's
match that DB is the strongest tactical chess-playing entity in the
world. Are you so confident that it's not that you can say GK should
have tried it straight up in tactical positions the way (for example)
he did in the first game last year? While it's clear his strategy of
this year didn't work, what makes you think DB is a weak enough
tactical player that GK would have been better off standing two feet
from its metal face exchanging punches until one of them fell down?
Playing Irish Stand-Down with a grizzly bear doesn't appeal to me.
Now that I think of it, this <was> rather like a boxing match, with
IBM cast as Don King (MATCH OF THE CENTURY! BRAIN AGAINST BRAWN!)
and Kasparov facing the first android that looks strong enough to
give him a match. His advisors advise him that it's been improved
from last year, when it caught him once with a powerful roundhouse...
now he has to stay away from the left hook as well. The champ,
undefeated against other humans, is confident that he can win
playing rope-a-dope, letting his stronger opponent flail away until
it leaves a wide opening.
Round 1: GK floats like a butterfly, gets in a bunch of good jabs, and
scores a knockdown.
Round 2: DB floats like a butterfly, is far ahead on points, and a
groggy GK slips and falls at the end of the round. GK, punchy from
his beating, claims there's a midget inside the android controlling
its arms.
Round 3: GK wins the round on points, but can't put DB on the mat.
Round 4: See round 3.
Round 5: See round 4.
Round 6: GK, frustrated and confused by his inability to score another
knock-down and exhausted from running away from the left hook, ties
his shoelaces together before the round and is unable to float out
of the android's path. Sure enough, the android <does> have a brutal
left hook.
Post-game show: GK's team wants to see what's under the android's
skin. GK says he was robbed. GK says the sponsor wanted him to
lose. The r.g.c.c fans say -- well, they say lots of things, like:
Congratulations, Deep Blue!
GK's right -- there's a midget inside it!
It would've been different if it'd been a 20-round bout!
GK should have stood in front of it and traded punches!
GK tied his shoelaces together on purpose so he'd have an excuse!
Another boxer could have played rope-a-dope better than Garry!
Let's try other boxers until we find one that does rope-a-dope better!
It's not fair -- it didn't have to fight anybody else before Garry!
It's not fair -- there should have been a day off between rounds!
There was a PC on the grassy knoll! (nice one, mig)
Well, as far as I'm concerned... Congratulations, Deep Blue!
Jim Gillogly
j...@acm.org
Paul Rubin
In article <3378925D...@acm.org>, Jim Gillogly <j...@acm.org> wrote:
>Round 6: GK, frustrated and confused by his inability to score another
>knock-down and exhausted from running away from the left hook, ties
>his shoelaces together before the round and is unable to float out
>of the android's path. Sure enough, the android <does> have a brutal
>left hook.
Very good comparison except in round 6, I'd say GK tripped over
his own shoelaces and fell before the opponent got anywhere near him.
He transposed into a well known book trap and was busted out of
the opening. It didn't take a monster android to beat him at that
point, any other strong player would have done so also, from what
I understand.
Komputer Korner
Serge Desmarais wrote:
e Karpov a chance. He has the official title after all.
> >
> > It's not obvious to me that the strategy of anti-computer chess was
> > flawed from the beginning. Kasparov has felt since before last year's
> > match that DB is the strongest tactical chess-playing entity in the
> > world. Are you so confident that it's not that you can say GK should
> > have tried it straight up in tactical positions the way (for example)
> > he did in the first game last year? While it's clear his strategy of
> > this year didn't work, what makes you think DB is a weak enough
> > tactical player that GK would have been better off standing two feet
> > from its metal face exchanging punches until one of them fell down?
> > Playing Irish Stand-Down with a grizzly bear doesn't appeal to me.
> >
> >
> > Jim Gillogly
> > j...@acm.org
>
> Yes... Congrats! :)
>
> Serge
Well, the match proved that even if you play anti computer chess
whatever that is, you can't prevent the position from opening up.
Tactics will always be there so you might as well face the music.
If Kasparov had played his normal game, solid the way he plays against
Karpov, he would have won.