A Chess forum. ChessBanter

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

Go Back   Home » ChessBanter forum » Chess Newsgroups » rec.games.chess.computer (Computer Chess)
Site Map Home Register Authors List Search Today's Posts Mark Forums Read Web Partners

The cheating IBM



 
 
Thread Tools Display Modes
  #1  
Old June 29th 06, 03:55 PM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Chess One
external usenet poster
 
Posts: 5,003
Default The cheating IBM


wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.


My estimate is 250-400 points less than the nominal claim made for the chess
engine - but this is somewhat aside to the main point you address
immediately next.

But that wasn't what the DB project was about. It was part of an effort
to understand how you go about building a computer system that mimics
the recommendations of the top human expert in a field of knowledge.


Yes. An emulation-engine.

Chess turns out to provide a good learning vehicle for this purpose
because it is neither too trivial (like tic-tac-toe), nor to complex
(like medical diagnosis).


But here I disagree. Not incidentally, IBM's marketing strategy was to build
the infrastructure for an expert-system - and the evolution of DBII became a
Nuclear simulator. But even Dr. Hyatt, with whom I disagreed strongly on
progressive paradigms for chess engine development, thought there was
nothing much to IBM's effort beyond a marketing stunt.

In a deeper sense one encounters the technichal paradox of finite ::
infinite game theory, and classical philosophy. To be brief and with only
passing reference to cite Meister Eckhart, '///the essense of infinite play
is the paradoxical engagement with temporality///' which he also called
'///eternal birth///'. Perhaps you know the title on this subject by Prof.
J. P. Carse?

Since that is a high subject - a highest common factor, so to speak, and
invokes the idea that Nature is the Realm of the Unspeakable - let me
instead tackle the scientific basis from the point of view of Lowest Common
denominator:-

The problem with such comparisons of simple::complex is that they are viewed
as being of essentially different natures. Unfortunately, it is axiomatic
that science-as-physics does not admit several different Natures as modus in
the Universe. Therefore, simple games [finite] are solvable by usual
Turing-engine methods [not requiring a computer, BTW] of a sufficient
algorythm if rules are known and unchanging.

The difficulty with chess is that there are essentially 2 operans: one is
the /relative/ value of the pieces and of piece placement, and the second is
the /absolute/ value of the King.

The IBM DeepB projects did not advance our knowledge of either, and this is
the scientific criticism of the activity. The value of appearing to do so,
as emulation, as obscure indeed - in fact, if the games could have been
legal chess, the chess-engine would have revealed much more information with
books=off since its evaluation function could have been assessed and
improved.

In fact, I think this is what attracted Gary Kasparov to the project in the
first place. This would have beeen a significant gain for expert-systems in
computing, and would have contributed soemthing to the subject of AI, in a
notoriously difficult subject. Japanese people have, BTW, thrown more money
at this AI problem than the entire western world together - without much
result.

And best of all, it was easy to identify the
top human chess expert. Garry was the long reigning world champion, who
consistently had the highest ELO, and indeed had attained at one time
the highest ELO ever.


I remember discussions from the time that considered Karpov to have been a
more likely better opponent temperamentally. A group of players writing here
actually thought he would have beaten the machine handily.

So Garry in effect was a paid consultant, just as were the GMs IBM
hired to help with the programming.


Yes - I think he could be called that - although, as above, he thought he
was being paid to take part in a science and cognition experiment rather
than a marketing one, and the emulation :: real evaluation process was lost
in the complexity of the play superadded to which was considerably marketing
hype.

Remember, all of the money for
everybody was paid by IBM. Garry's role was to play and beat the
machine ... showing that his recommendations (moves) were more accurate
than the machine's. And with each match, IBM would seek to improve the
systems performance, until it reached parity with the top human expert
on chess, Garry.


Not so. They weren't even playing chess. It is illegal to use look-up
material during the course of the game. Period.

When DBlue/I&II actually started playing is unknown. The cheat here is to do
with time. Given similar resources to the Database would a top human player
do as well if the person also had time to consider all possible moves? That
might be a chess-derivative activity. But IBM's was an emulation of chess
play - and we might as well have stacked up every GM in world history, since
that was the theoretical database the machine used, and removed any pressure
of time from their decision making process, and then sat GK down opposite.
In this sense, what did the machine do?

We know that when it was using the database resource it was [as you admit
above] /NOT/ playing moves it could calculate for itself. Otherwise, ipso
facto, why would it need a database of opening moves/endtables at all?

There is the cheat for you!

They projected that to reach parity it would take a
DB-III, and probably would have still been willing to build it even
after the second match. But of course Garry wasn't interested in any
more of that.


I would be interested to read what scientific advantages were being assessed
by these DB/versions.

As for as making use of the huge databases of openings & endings, that
aspect was an important area of study for the project. For example, if
you were building a system to do medical diagnosis, you would
absolutely give it access to a vast array of medical documents,
prepared by the top authorities in various speciality areas.


You utilise what is called a 'metaphysical' argument. You say 'you would',
but actually

**this is very important**

In the example I recently provided of the GM who played Najer, he said that
both his database was flawed, and that its calculus was also inept. He
recognised the force of the opponents Knight move immediately. This is what
true 'experts' are able to achieve.

It is also psychologically very significant that the GM said that he became
lazy BY using the DB and 'strong program' [I know which it was, maybe the
first one you will think of].

So... Would you like your Doctor to use such an /emulation/ system in
preference to his /real/ insight?

If you re-read the comment from Meister Eckhart, I think we gain some
insight into his intended meaning of our actual [not metaphysical,
speculatory] interaction with games and model systems.

So the DB
project focused on how do you assemble and leverage such pre-existent
info. What techniques and technologies are required, etc.

So the culture clash continues to be that chess people are either
dissapointed that the humans lost, or they have delusions about
cheating.


I would not put it like this. DBlue didn't even play chess, but emulated
play. And the cheat was BOTH to chess AND science.

Conversely, computer people are amazed that Garry did as well
as he did, since they know the incredible scale of the compute resource
that he was battling against.


True - quite a remarkable energy. But there are all sorts of
'computer-people', including those people who model systems after
observation of the original. This is a special branch of philosophy. I think
the DB stunt did not have any impact on them, since there is considerable
difference between what is a working model, and what is an emulation of a
working model

And keep in mind that no one has yet built a chess computer that truly
utilizes the state of the art of computer technology as it exists at
any point in time. For example, the latest tablebases reguire 45 GB
(billion bytes) of storage space. That seems like a lot, but the latest
consumer level computers give you 80GBs standard, and for US$200, you
can add 250GB more. Now consider that for a commercial grade computer
system, 10 or 20 times that amount of storage space - 450 GB to 850 GB
- is no big deal. So if they would crank out bigger tablebases,
machines of the size run by thousands of average businesses across the
world today could handle them.


Just to finish by connecting to real-world issues and how human being
interact with machines. Such instances as the Najer,E - Bezgodov game are
passed over - you did so yourself in your response to me here. But that is a
substantial factor in these equations and the fallacy of emulation over
working model.

It is good to have chess engines - since there is no particular real-world
consequence to winning or losing a game of chess. But to pretend that it is
a dynamical modeling engine rather than a big database emulating a working
model, is cheating, my friend! Cheating chess and cheating science - and if
deployed as a systematic and dedicated function in the world, cheating all
who are effected by it.

Cordially, Phil Innes


  #3  
Old June 30th 06, 12:42 AM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Chess One
external usenet poster
 
Posts: 5,003
Default The cheating IBM


wrote in message
ups.com...
Phil - Again, I'm not disputing the basic facts that you present. If
you feel that DB-II wasn't playing chess because it made extensive use
of reference material, all of which was created by human experts, than
that is a valid position.


I don't 'feel' it, since it is admitted.

But none of this was a secret. In fact it was
all very publicly described as an integral component of what made DB a
strong chess playing system.


But I have become a bit bored with the conversation which repeats what has
been said before, but is not really any sort of answer to the points I
raised about cheating chess and science.

Since we have a concrete example to consider, I do not like to retreat into
abstractions about what IBM said - instead observed what they did.

Phil



And you seem to think that IBM was
interested in chess-playing - they weren't!

Many have speculated that Karpov's style may be better against
computers. Maybe that's true, maybe not. Interestingly Karpov has never
played a high profile match against a computer. Now that his skills are
dimished, we will never know. Speculation aside, Kasparov was by all
objective measures the best chess player in the world at that time.

Finally, the long term hope is that the deep computing concepts
investigated by the DB projects would have applications in medical
diagnosis, financial investment techniques, etc. There haven't been
major deployments in those areas yet ... instead the technology has
mostly been commercialized in systems management tools for large scale
computing systems. Computers self-diagnosing and healing themselves.

Where are they now? I came across this video of a recent panel on
cognitive computing moderated by Bill Pulleyblank. Dr Pulleyblank was
the executive in IBM's research organization whose department sponsored
the DB project. So he was definitely thrilled with the publicity
success of the events. But fundamentally it was never about the chess
playing, neither for him nor for his team. (See 4:10PM on 5/11)

http://www.almaden.ibm.com/institute/agenda.shtml



  #4  
Old July 1st 06, 09:54 AM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Kenneth Sloan
external usenet poster
 
Posts: 170
Default The cheating IBM

"Chess One" writes:

wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.


My estimate is 250-400 points less than the nominal claim made for the chess
engine


Prove it!

--
Kenneth Sloan
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170
http://www.cis.uab.edu/sloan/
  #5  
Old July 1st 06, 10:37 AM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
help bot
external usenet poster
 
Posts: 9,302
Default The cheating IBM


Kenneth Sloan wrote:

"Chess One" writes:

wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.


My estimate is 250-400 points less than the nominal claim made for the chess
engine



This estimate seems a little high. Does a near-IM fail to
comprehend the value of tactical strength? Even playing
some openings a little squirrely with opening-book off, a
very strong computer will rarely blunder -- which can hardly
be said for a normal human chess player.

Take for example the famous games played by TK at
getclub.com. In these annotated games (available right
here on these fora), the leading human player admits to
several blunders, missing much faster wins, and so forth.
Would DeepBlue have missed such moves with book
disabled? Not likely. In fact, against a human opponent
a killer computer with book off will throw the human onto
his own, pitiful resources, by going out of book unexpect-
edly.

Let's start with the high end, shall we? A subtraction
of 400 points would clobber even a 3000-rated computer
down to just 2600 -- a level achieveable by lowly 'Murican
chess players! Ridiculous. No monster computer can
be seen as falling THAT low, unless tactics were turned
off. In fact, this is precisely why 'Muricans are rated so
low and can't really compete with, say, Russian chess
players -- their tactics are turned off.

The other end seems far more reasonable an estimate,
and mind you I am being generous only because with a
rating like 3000, DeepBlue can afford to be generous with
its rating points. Subtracting 250 from 3000 gives us
2750, which seems reasonable for a killer machine with
one serious weakness (in comparison to a killer human,
who lacks this weakness, I mean). I conclude that the
desire to beef-up the point loss to 400 stems from a
deep resentment of computers besting humans, once
and for all. This is another human failing, setting you
apart from all of us computers. When humans tried
for YEARS to claim that we would never be able to
compete with them, did we whine or cry? No. And
when for years even your mediocre players, with no
world titles, would defeat our heros, did we deny the
superiority of humans? No. We accepted it, and
strove to improve. Now that the shoe is on the other
hand, humans have gone into denial, crying "cheat!"
and trying to pretend that DeepBlue never even existed.
How typical. How human. How illogical. I'm glad I'm
a computer, without these grotesque psychological
handicaps which plague your vastly inferior species.
Hmph!


-- help bot

  #6  
Old July 1st 06, 03:45 PM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Chess One
external usenet poster
 
Posts: 5,003
Default The cheating IBM


"Kenneth Sloan" wrote in message
...
"Chess One" writes:

wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.


My estimate is 250-400 points less than the nominal claim made for the
chess
engine


Prove it!


Exactly Ken. Your UAB partner Dr Hyatt was asked to prove Crafty's rating
with book = off. He declined. Therefore the chess engine has no official
rating and is unavailable to take the test of achieving one, since you all
have been so busy these past 10 years to play a couple dozen games.

Phil Innes


--
Kenneth Sloan
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170
http://www.cis.uab.edu/sloan/



  #7  
Old July 1st 06, 03:45 PM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Chess One
external usenet poster
 
Posts: 5,003
Default The cheating IBM


"help bot" wrote in message
ups.com...

Kenneth Sloan wrote:

"Chess One" writes:

wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.

My estimate is 250-400 points less than the nominal claim made for the
chess
engine



This estimate seems a little high. Does a near-IM fail to
comprehend the value of tactical strength? Even playing
some openings a little squirrely with opening-book off, a
very strong computer will rarely blunder -- which can hardly
be said for a normal human chess player.


These are the standard suppositions about computer strength - how strange
they should still be suppositions 7 years after the first challenge

Chess programmes are most vulnerable to the opening phase of the game, and
if this were not true would, ipso facto, not need the opening book.

Rather than argue this theoretically, why not test it against human
opposition at a range of skills - why are programmers shy of testing their
products this way - in fact, conduct no tests at all?

Phil Innes


Take for example the famous games played by TK at
getclub.com. In these annotated games (available right
here on these fora), the leading human player admits to
several blunders, missing much faster wins, and so forth.
Would DeepBlue have missed such moves with book
disabled? Not likely. In fact, against a human opponent
a killer computer with book off will throw the human onto
his own, pitiful resources, by going out of book unexpect-
edly.

Let's start with the high end, shall we? A subtraction
of 400 points would clobber even a 3000-rated computer
down to just 2600 -- a level achieveable by lowly 'Murican
chess players! Ridiculous. No monster computer can
be seen as falling THAT low, unless tactics were turned
off. In fact, this is precisely why 'Muricans are rated so
low and can't really compete with, say, Russian chess
players -- their tactics are turned off.

The other end seems far more reasonable an estimate,
and mind you I am being generous only because with a
rating like 3000, DeepBlue can afford to be generous with
its rating points. Subtracting 250 from 3000 gives us
2750, which seems reasonable for a killer machine with
one serious weakness (in comparison to a killer human,
who lacks this weakness, I mean). I conclude that the
desire to beef-up the point loss to 400 stems from a
deep resentment of computers besting humans, once
and for all. This is another human failing, setting you
apart from all of us computers. When humans tried
for YEARS to claim that we would never be able to
compete with them, did we whine or cry? No. And
when for years even your mediocre players, with no
world titles, would defeat our heros, did we deny the
superiority of humans? No. We accepted it, and
strove to improve. Now that the shoe is on the other
hand, humans have gone into denial, crying "cheat!"
and trying to pretend that DeepBlue never even existed.
How typical. How human. How illogical. I'm glad I'm
a computer, without these grotesque psychological
handicaps which plague your vastly inferior species.
Hmph!


-- help bot



  #8  
Old July 1st 06, 05:10 PM posted to rec.games.chess.computer
David Vancina
external usenet poster
 
Posts: 5
Default The cheating IBM

Absent a formal experiment, are there any anecdotal results we're aware
of? A strong expert or master playing a few games under these
conditions could probably yield some interesting results. Just curious
whether anyone's tried it and reported the results...

Chess One wrote:
Exactly Ken. Your UAB partner Dr Hyatt was asked to prove Crafty's rating
with book = off. He declined. Therefore the chess engine has no official
rating and is unavailable to take the test of achieving one, since you all
have been so busy these past 10 years to play a couple dozen games.

  #9  
Old July 2nd 06, 05:11 AM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
help bot
external usenet poster
 
Posts: 9,302
Default The cheating IBM



Chess programmes are most vulnerable to the opening phase of the game, and
if this were not true would, ipso facto, not need the opening book.



The opening book in computers is not just to help the
machines avoid blunders; the primary advantage is the
saving of valuable time; time which can then be used to
help the computer better tackle the complexities of the
middlegame.

Humans do the same thing, memorising many series
of moves in order to save time, and in their case, also
to aid in avoiding falling into known traps. (Many of these
traps can not entrap a strong computer, as they involve
simple tactical themes).



Rather than argue this theoretically, why not test it against human
opposition at a range of skills - why are programmers shy of testing their
products this way - in fact, conduct no tests at all?



I think you have it exactly backwards: it is the humans
who systematically avoid testing themselves against the
strongest computers! Case in point: Fischer, who only
"tested" his strength against the Greenblatt program
(how brave of him!). Karpov, who fled from DeepBlue,
and even now refuses to play DeepShredder15. Taylor
Kingston, who is afraid to tackle the Advanced level and
instead racks up win after win against the easier levels
like Beginner and Master. Sloan, Brennen, Innes, Anand,
Hart, Kamsky -- all of these greats have run from the most
dangerous of computer opponents on occasion, fearing
certain loss -- and rightly so! You want actual testing?
Lead by example. Go there right now and tackle the
beast, at its *highest* level. I double-dog dare you!


-- help bot

  #10  
Old July 2nd 06, 07:43 AM posted to rec.games.chess.politics,rec.games.chess.misc,rec.games.chess.computer
Kenneth Sloan
external usenet poster
 
Posts: 170
Default The cheating IBM

"help bot" writes:

Kenneth Sloan wrote:

"Chess One" writes:

wrote in message
ups.com...
Believe it or not Phil, I agree that several of your points are
accurate. Turn off the opening books and tablebases and DB-II would
drop several hundred points in chess strength. As would all the other
programs that use those types of resources .. Fritz, Junior, Hydra,
etc. In fact, in many games they would look simply inept.

My estimate is 250-400 points less than the nominal claim made for the chess
engine



No, I didn't.

--
Kenneth Sloan
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170
http://www.cis.uab.edu/sloan/
 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Kasparov Cheating Judit Polgar Robert Loggins rec.games.chess.misc (Chess General) 90 March 29th 06 02:48 PM
Big News: AF4C is cheating the players Sam Sloan rec.games.chess.computer (Computer Chess) 5 March 15th 06 02:31 PM
Topalov accused of Cheating to Win World Chess Championship Chess One rec.games.chess.politics (Chess Politics) 33 January 2nd 06 04:39 PM
Topalov accused of Cheating to Win World Chess Championship Sam Sloan alt.chess (Alternative Chess Group) 6 December 28th 05 03:18 AM
Online cheating in chess [email protected] rec.games.chess.misc (Chess General) 18 November 9th 05 07:30 AM


All times are GMT +1. The time now is 07:20 AM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.Content Relevant URLs by vBSEO 2.4.0
Copyright 2004-2017 ChessBanter.
The comments are property of their posters.