Conference 7.286::digital

    
    The odds are not even as you say, Kasparov has the advantage.
    
    I firmly believe that Gary threw a couple of the first games, so that
    interest would peak, and the prize money would increase.
    
    Deep Blue, is only as good as the programmers program.
    
    The fact that it can process 200,000 or whatever moves per second is 
    in material to whether or not the moves are "good" or "bad" moves.
    
    Good or bad moves are defined in the context of the Chess 
    masters/programmers/advisors perspectives, not by whether or not
    they are in fact the "best" moves.
    
    I submit that Gary would beat all of the chess masters collectively,
    and thus would still beat their collective Deep Blue effort.
    
    Of course, it will be a close match, sweat will appear on Gary's
    forhead, and in the end he will pocket another $500,000.
    
    Yahoo, for Gary!
    
    /art

    It's 200 million positions a second for Deep Blue; I'm not sure what
    kind of systems Kasparov's using. DB's current incarnation is an MPP
    (512-processor?) RS/6000. Evaluating lots of chess positions
    parallelizes well, so he with the most processors wins.
    
    Re: .1: As for "throwing" games, I tend to think the human grandmasters 
    watching (and commenting on) the match would pick up on any questionable 
    strategy. These games are probably analyzed as much or more than any in
    tournament chess history. I do agree that Kasparov has the advantage, 
    though.
    
    Those of you with a spare PC can watch the match at www.chess.ibm.com -
    I've been glancing over at the game now and then this afternoon.
    
    RE: .0: 
    Sure, it'd be nice to have Alpha systems doing stuff like this, but
    IBM has been in this one from the beginning... we're certainly not
    getting in on it at this point. 
    
    IBM knows how to use those R&D and marketing dollars...
    

> DB's current incarnation is an MPP (512-processor?) RS/6000.

IBM tries to make this look like an off-the-shelf supercomputer, but it's not.
Attached to each processor is a bunch of specialized hardware for evaluating
chessboards.  If it were done purely with software and general-purpose
processors it would be a couple of orders of magnitude slower.

Just tried http://www.chess.ibm.com, and it timed out with no response. 

May be they need some fast AlphaServers to handle their web traffic :-)

>IBM tries to make this look like an off-the-shelf supercomputer, but it's not.

Actually, they are making it look like a computer.  Thats all that matters to
most of the population.  IBM continues to grow its name recognition, and it 
doesn't matter whats in the box.  Its another piece of marketing, something
we really don't understand.

    >>  I firmly believe that Gary threw a couple of the first games, so that
        interest would peak, and the prize money would increase.
    
    This was my suspicion last time, although I would simply say there is a
    conflict of interest which does not make the table level. However, we
    now apparently have a case to work on. Smart Gary apparently resigned a
    drawn position, so he either shot himself in the foot or he was
    subconsciously planning to lose. Also, Deep Blue may have blundered by
    offering him the stalemate. There is a strong smell of fish, in any
    case.
    
    After the last match an article appeared in Digital Today reporting on
    a match between an Alphastation 500 and Judit Polgar, the world's top
    woman chess player. The article stated that Digital Israel was working
    with the developer's of the chess program Junior to improve it and take
    advantage of 64 bit technology. That was in October 96. I wonder if
    there has been any development. Also, wasn't there some kind of "chess
    software" match held recently?  Apparently it was another chess program
    which discovered that Gary had 'thrown' the match against Deep Blue.
    
    Big Blue has not won this one yet.
    
    ..Kevin.. 
                                      

    according to Newsweek (May 5, 1997) VLSI Technology Inc. provided the 
    (IBM co-designed) custom chess processing chips integrated into the IBM
    Deep Blue machine. 

> (512-processor?)

Yesterday's Globe had an article reporting DB to be 32 node machine.
I think they called it an RS/6000 SP2CP.  The physical dimentions
and weight (1.4 tons?) made it sound about the size of a TurboLaser.

    Sooo,  would it be easy, difficult, or near impossible to have an Alpha
    machine eclipse the IBM # of moves/second processing time???
    
    Ken

>                     <<< Note 5277.9 by DEMAND::KAMINSKY >>>
>                              -< Still not sure >-
>
>    Sooo,  would it be easy, difficult, or near impossible to have an Alpha
>    machine eclipse the IBM # of moves/second processing time???
>    
>    Ken
    
    First you'd have to have a microchannel interface into which to plug the
    specialized chess analysis boards.

>    After the last match an article appeared in Digital Today reporting on
>    a match between an Alphastation 500 and Judit Polgar, the world's top
>    woman chess player. The article stated that Digital Israel was working

Yeah, don't tell the outside world about this or anything ... let's just keep it 
between us.

re .6

>    Big Blue has not won this one yet.
 
Au contraire.  Whether the human or the computer wins the match, Big Blue has 
won this one by a landslide...

A company can not pay for this sort of publicity...
                                      

    I was listening to Howard Stern this morning (no comments please on my
    "taste" (or lack thereof)) and he said:
    
    "God, I love IBM.  I love that IBM invented a computer that can
    beat a world chess champion."
    
    Now of course we understand that it is more complicated than that.
    Its more than just a computer, but here's the "perception is
    reality" concept again.
    
    				-John

    Now this is interesting. I'm wondering if the AP wire service
    still has this wrong. I remember reading earlier (when Deep Blue
    was first announced), when the publicity first came out on it 
    playing chess vs a human, that it was either a radically modified
    version of some production IBM machine, or that it was totally
    custom. The mention of the Rs./6000sp in the previous note seems
    to suggest that it isn't a custom machine. I think this is 
    mis-representing the facts. 
    Can anyone set this straight once and for all? Is Deep Blue a
    spin-off of a production machine (Rs./6000sp), or is it a
    totally customized machine built for only one purpose? I was under
    the impression that it was the latter.
    
    Bob

The RS/6000 SP2CP would seem to imply that it is a modification of
an RS/6000 SP2.  Each processor is certainly connected to a special
chess position evaluation board.  That's probably the CP part.
Within the past 2 years there was a Scientific American article with
a photo of a custom board built by IBM.  The photo in yesterday's
Boston Globe certainly looked like a regular RS/6000 SP2 box, but
boxes are boxes -- Digital reuses its boxes for many things too.

    Go to http://www.chess.ibm.com and click on "the players" to get a
    description of the hardware. It is a combination of a standard
    32-processor RS/6000 with a set of chess-specific option boards (so
    that there are 8 chess-analyzing VLSI chips for each CPU in the
    system for a total of 256 chess chips running in parallel). 
    
    Seems to be sorta like adding a video board with special MPEG
    support in your PC -- you don't have a made-to-order MPEG processor,
    but you do have a PC that does MPEG a whole lot better than the
    average PC.

    
    Well,
    
    one of our customer has a brand new RS6000 (it's been dubbed "the Black
    Venus" by the local DECcies ;-) in his computer room, sitting back to
    back to a bunch of TurboLasers.
    
    It happened that the Black thing was so slow in processing requests
    coming from our systems that our programmer had to put a "sleep (1)"
    inside the main processing loop.
    
    Maybe the box is OK for playing chess and counting torchlights only ?
    
    			Ezio
                            

    >>Au contraire.  Whether the human or the computer wins the match, Big
    >>Blue has won this one by a landslide...
    
    At the present time no doubt IBM is getting a lot of mileage out of
    this. What I am suggesting is that computer chess technology is in a
    primitive state at present and that is why they are looking good. Plus
    they are playing Garry a lot of money and he is obliging them by
    throwing in the towel and resigning a drawn position. 
    
    However, the publicity IBM got from the Atlanta Olympics was not
    exactly bril was it. And it was another chess software that
    discovered that both Garry and Deep Blue blundered.
    
    My point is that the question remains unanswered: would a vlm
    application running on alpha give us a competitive advantage in this
    area and can we leverage it?  To quote Shy Bushinsky, who is working on
    this: "This summer we hope to work with Digital Israel to take full
    advantage of Alpha's 64 bit architechture."
    
    I could imagine the following situation: chess grand masters using a
    vlm application on alpha to provide them with support for analysis in chess
    competitions might have a competitive advantage. That would be a story
    which would be nie to read, so please pull down the white flag and
    let's see what we can come up with. 
    
    ..Kevin..
             

    re:.19
    
    >My point is that the question remains unanswered: would a vlm
    >application running on alpha give us a competitive advantage in
    >this area and can we leverage it?
    
    I am thinking of a similar spin:
    
    Digital helps Kasparov even the score.  Digital's new Alpha technology
    can analyze 400 million moves per second, doubling the existing
    analysis speed of the IBM machine.
    
    Says Kasparov,  I never realized the IBM machine was so slow.
    
    Ken
       

    	I had resisted this thread until now, but now you have provoked
    me to reply.
    
    	Chess playing programs have been around for a long time.  I can
    remember some of the early ones (mid 60's), and they were quite
    remarkable.  The rules are simple enough so that any competent student
    should be able to write a program which plays legal games.
    
    	Then you just strap a "reasonable" static evaluation function on
    it, add some code to look ahead a few levels, and wait for the
    horsepower to improve to the point where the machine plays a good
    game, right?  Wrong.  The big problem is that while all of these
    areas have improved (dramatically) over the years, the stupid 
    machine still has no clue about strategy.
    
    	The biggest advances (and there HAVE been advances) have been in
    endgame (small number of pieces) studies where the processing power
    finally proves its worth.  In fact the rules have been modified (what
    a concept!) to allow for certain endgames which are now known to be
    winnable, but NOT within the older limit of 50 moves.  Anyway, the
    point is that there have actually been some advances in knowledge
    due to some brute force analysis made possible by current programs.
    
    	But the lack of strategy is such a handicap that even with special
    hardware and the fastest of machines, at the highest levels the machine
    is still woefully inadequate.  Look at the kinds of comments that are
    being made by those in the know - the machine just "does not play like
    a human, and makes shortsighted mistakes".
    
    	So, you might counter, add a strategy routine.  Easier said than
    done.  Much easier.  In fact, this is where the best programs are
    really better.  This is also where the best humans are MUCH better.
    That's why all the serious programs have professional grandmasters on
    the programming teams.
    
    	Even with the fastest machines, one MUST limit the search space
    (for it grows real fast in the early stages of the game).  Of necessity
    the program selects some number of moves to examine, and discards
    others.  Just like humans.  But not a lot of time can be spent deciding
    which ones to examine more deeply and which to discard.  And one invariably
    needs to do this WITHOUT examining the following lines.  This is exactly 
    where the human wins.
    
    	If you don't believe this, consider the following.  I was watching
    Grandmaster Tal (a truly marvelous tactical player) in a simultaneous
    event some years ago.  He was playing about 40 opponents, moving from
    one board to the next fairly rapidly.  I was watching him play.  There
    was one position in particular, and I was familiar with the line of
    play that had been used.  Ths position was very interesting.  Tal
    stepped up to the board, spent only moments looking at it, and played
    a move that I had not considered.  Not at all.  And it was a very, very
    interesting move.  So I stared at this position for quite some time,
    asking myself why this move had been selected, and what it
    accomplished.
    
    	To make a long story short, the move was really quite brilliant.
    It had a long term effect that I had not considered.  Not at all.  And
    of all the moves available, this was a move that I was not ever going
    to consider making.  I'm certain of that.  That's the difference
    between how a grandmaster views the possibilities within the position
    and how an expert (which I am) views them.  This was simply a move I
    would not consider.  And I would submit that the SAME thing would be
    true of even the most sophisticated programs.
    
    	The bottom line is that humans think in terms of themes and ideas
    and long term strategies.  Machines tend to be just great at
    calculation of long lines in limited positions and lousy in early
    evaluations of the current position.  Now all you artificial
    intelligence advocates will reply that it is merely a matter of
    programming, and you will continue to lose money to those with
    the common sense to bet against your machine...
    
    	The effort to make even the most modest improvements in how the
    machine looks at the game is huge.  Gigantic.  Staggeringly
    difficult.  Many have tried.  The programs are good, but not great.
    Great play is still a human thing, and I expect this to be true
    for the forseeable future.
    
    Cheers!
    Dave Eklund              
    

    Hey Dave,
    
    What's the difference between a Chess Master and a Grandmaster? 
    Headwear? ;-)
    
    Thanks for the insight, btw...
    
    

    re .21
    
    >But the lack of strategy is such a handicap that even with special
    >hardware and the fastest of machines, at the highest levels the
    >machine is still woefully inadequate.
    
    Well that was my understanding of the conventional wisdom, but how is
    it then that Deep Blue has beaten one Garry Kasparov twice, given that
    Mr K is not exactly a beginner at the game. If the above statement is
    true then the world is being hoaxed, isn't it? That has been my
    suspicion all along. Kasparov 6 Deep Blue 0 is no story, but Kasparov 4
    Deep Blue 2, thanks for the cheque, see you in 6 months,  is nice work
    if you can get it. 
    
    
    ..Kevin..

    In an interview after game 4, Garry said that he was getting tired
    after several hours and his concentration was getting poor.  He
    admitted he had an advantage late in the game, but was unable to
    capitalize on it.  Deep Blue doesn't have the problem of getting
    tired...  For Garry against the computer, I would like to see the
    time limit lifted or extended.
    
    				-John

I've often heard Howard Stern mention IBM.  The rumor is that IBM 
gave him a free laptop, free internet service, etc.  IBM knows how to 
play to marking game!

Of course, I not sure the value of having your company mentioned in 
the same breath a naked-mud-wrestling-lesbians.


                                                -Paul

    re .22
    	There are two international titles conferred by FIDE, the
    most recognized international chess body - these are International
    Grandmaster (IGM) and International Master (IM).  Both of these
    are "earned" by participating in recognized tournaments and
    achieving certain "norms".  These tournaments are rated by "category"
    (strength of players).
    
    	For example, in a category 14 tournament, one might need to
    score 10/14 to get an IGM norm, and 6/14 to get an IM norm (not the
    real numbers).  Then it takes a certain number of norms to be awarded
    the title.  I believe that it takes 3 IM norms to become an IM, and
    don't remember how many to become an IGM.
    
    	Then there are titles from other "national" groups, like the
    USCF (United States Chess Federation).  For example we have titles
    like master, senior master, expert, and various classes (a,b,c,d,...).
    These titles are based only upon US ratings, which get adjusted by
    playing in USCF sponsored tournaments.  Generally speaking, these
    rating groups are 200 points wide with master from 2200-2400, expert
    from 2000-2200, senior master from 2400 on up, class a from 1800-2000,
    etc.  Bobby Fischer in his prime was about 27xx, and received his IGM
    title somewhere around age 14.  Roughly speaking there are not too
    many players rated below about 2300 in this country who have IM titles,
    and not too many below about 2450 with IGM titles.
    
    	Grandmaster Tal was one of the very best (in many ways), a world
    champion at age 20 (the youngest), but with poor health over many
    years.  His games are a testimonial to his genius - many astonishingly
    brilliant moves to entertain the crowds, sparkling combinations, and
    few dull games.  His play was largely intuitive in the purest sense.
    He openly criticised the "children of the Informants", those who merely
    memorized published lines of opening analysis.  He was certainly
    capable of eating computers/most humans for breakfast, lunch AND dinner...
    although he might also lose games through holes in his intuitive
    analysis (or weariness).
    
    Cheers!
    Dave Eklund
    

Re .21:

>        Chess playing programs have been around for a long time.  I can
>    remember some of the early ones (mid 60's), and they were quite

Early? In the mid 60s? You're forgetting the 18th century "Turk" automaton (but 
you'll probably disqualify that. ;-)

BTW no prizes for guessing what comes up if you do an AltaVista search for 
"chess".

PJDM

    re .21
    
    Well the postulate you propesed in .21 Dave, that the computers are
    "woefully inadequate" is looking a bit windy this morning. Deep Blue
    beat Kasparov, after he lost in 19 moves, from what I heard on the
    radio. Also, there was  an interview with the chess correspondent of
    the New York Times who said that computers have been able to beat 99%
    up to now, but that now they have got to 100%.  A lot of questions
    arise out of this. Kasparov apparently said that Deep Blue was getting
    "human help."  He would "tear it apart" if Deep Blue played competitive
    chess. Precisely what was the role of the grand master consultant who was
    working for Deep Blue? Did he "coach" Deep Blue to take stategic
    decisions ? 
    
    The first computer chess player beat Napoleon three times. It tuned out
    there was a dwarf inside the box ? What is going on inside Deep Blue?
    
    ..Kevin..

    Actually, they haven't reached 100%. I believe he won a match and 3
    were draws out of the 6. 
    
    Chip
    

Forget the result, just keep watching the publicity role in. The UK Sunday Times
had a front page article on Kasparov claiming deeper blue was 'cheating' and he
wanted to see why certain moves were played. In the Inovation section of the
paper was an explanation of how the technology behind deeper blue was going to
aid the drugs developers half time to market from average 12years to 6yrs.
Message of the article - if you need a superpowerful computer for fast multi
computation simulations - IBM are your people. The IBM person said deeper blue is
showing off technology they plan to release soon. Electronic copy of article is
on the ST web page - www.sunday-times.co.uk

Danny.

    What about an "SP2 against TurboLaser" competition ?
    Do we really need Kasparov after all ?

    
    
    Ha ! nice one :^)
    
    A loooong time ago, in 1984, I had an HP PC (150B it was called) play
    chess against a Macintosh (first version) just for the fun of it. The HP
    salesman told hid management about it, and I was called in for an
    interview, and that was how I got hired as a salesman in HP ! (sounds
    ridiculous but it is the real truth). I lied. I told them the HP PC had
    won !
     I later left for a much better company called DEC (at that time it was..
    ..I mean called DEC)
    
    Michel.

>                      <<< Note 5277.31 by BIS1::WAUTERS >>>
>                              -< Why Kasparov ? >-
>
>    What about an "SP2 against TurboLaser" competition ?
>    Do we really need Kasparov after all ?
    
    First you have to get someone to design and build the custom chess
    analysis boards for TurboLaser (as mentioned several places, this
    was not an off-the-shelf SP2).

And then the software....

		Steve

> And then the software....

This is the key.  The analysis of chess tends to be of exponential
complexity, unless careful algorithms can prune this to be more manageable.
In other words, a well designed chess program running on an Intel 8086
will likely outplay a brute-force program running on a 466 MHz Turbolaser.

Ram

    >> And then the software....
    
    Exactly, and this is not an easy part...
    
    A fast computer (and custom hardware) is not the only thing that
    is needed to play chess well. Don't forget that the commercial chess
    programs that run on PC (Fritz, Genius, VirtualChess, etc...) already
    beat 99% of the chess players on this planet, with hardware that is
    far slower than what Deeper Blue has.
    
    Someone wanting to beat Deeper Blue with Alpha + custom hardware would
    also need to find out some people who have the same experiences with
    chess programing that the Deeper Blue team has.
    
    Michel.

    re .36
    
>   Someone wanting to beat Deeper Blue with Alpha + custom hardware would
>   also need to find out some people who have the same experiences with
>   chess programing that the Deeper Blue team has.
    
    I thought we first need Marketing   ;-) Sorry, couldn't resist.
    
    The question I try to answer is how you can be sure that Mr. Kasparow
    doesn't get paid for losing the battle? 

Kasparov got $400K for losing.

			Steve

DB got $700K for winning.

Re: .31

That's the ticket!

Why doesn't DS challenge IBM?

If DS loses, who cares?...the company is now on the short list of 2 for major
wins....If this company is afraid of the good/bad/? publicity, or if it isn't
up to competing with IBM in a game scenario, how can they hope to compete in the
real world?

Do you think that words to this effect don't roll off IBM sales' tongues daily?

$.02
gc

What's the point?  There would be little interest in a computer-computer
challenge, and if we weren't willing to invest the years and millions of
dollars it would take to develop the software and expertise IBM has, we
might as well not bother.

				Steve

    And Deep Blue wins the match...

>>What's the point?  There would be little interest in a computer-computer
>>challenge,

What's the audience for the ACM computer chess championship?

						Brian

I'd rather see an ad that showed an office full of people obviously wasting time 
playing chess, with the caption "Wouldn't you rather have your computer systems 
doing real work?"

PJDM

By Fred Kaplan, Globe Staff, 05/12/97 

                                       NEW YORK - The 
                                       science-fiction nightmare came 
                                       true yesterday. Garry Kasparov, 
                                       the best chess player in the 
                                       world, perhaps of all time, was 
                                       beaten at his own game by a 
                                       computer. 

                                       The IBM RS-6000
                                       supercomputer, known as Deep
                                       Blue, not only beat Kasparov in
                                       the last game of their six-game
                                       match - it slaughtered him...

[of course, the RS-6000 is not a supercomputer,
and Deep Blue is not an RS-6000.]

    Why dont we do a Steve Jobs, Turbolaser plus 6 year old kid beats Deep
    Blue, actually piloted by Kasparov? Just joking, IBM has this marketing
    hook cold, you need to find something else to capture people's
    imagination.

     It is clear that a master chess player should not have lost game 6 as
     Kasparov did.
    
    Secondly, the computer has been taught to play by a grand master,
    it did not defeat him through mathematical analysis. It would seem,
    (but we can't be sure) that Deep Blue recognised the position from
    memory and played the move it had been taught to trap Kasparov with a
    sucker punch. The suspicion is also that the computer had been taught
    stategic rules by the grandmaster coaching it, that is why it kept attacking
    and did not try to regain material. So what we have is computer/human
    assisted chess v. Kasparov.
    
    The future may be that we have a special form of chess developing,
    namely computer assisted chess, just as we have simultaneous
    exhibitions and 5 minute games.  We might see players sponsored just as
    racing cars are, and that would be an opportunity for Digital. I would
    agree though that IBM has quite a lead on this, but in a competitive
    context, other computer assisted players might find ways of catching
    up.  
    
    ..Kevin..
    

>     It is clear that a master chess player should not have lost game 6 as
>     Kasparov did.

"Oh, I'm sorry - I should have caught that ground ball and thrown
that player out at first to win the game, but I made a mistake.  Can we
do that batter over, please?"
He lost.  If a player of Kasparov's ability makes a similar bad move one 
in a thousand times, that's the way humans (at that level) play.
He lost.  If this was his 1 in a thousand moves, that's what cost him 
the game and match, just like Bill Buckner's one-in-a-hundred error
cost the Red Sox the 1986 World Series (not that it was ALL his fault).
    
>    Secondly, the computer has been taught to play by a grand master,
>    it did not defeat him through mathematical analysis. It would seem,
>    (but we can't be sure) that Deep Blue recognised the position from

I'm not sure we know how Deep Blue was trained.  Yes, there are Grandmasters
on the programming staff, but one very common method of chess programming
is self-teaching, that is, the computer plays its own games and learns
what works and what doesn't.  The job of the GMs in that case is to analyze
the self-taught play of the computer and explain it to the rest of the
staff and to talk the computer out of strategies that look good 
but are known to have deficiencies the computer hasn't found yet.

Note that this is the key part of the whole chess programming thing:
  The 200 million positions per second stuff is fluff, merely 
  enabling technology.  The hard part is the evaluation algorithm
  built into that hardware that says "this position is better for me than
  this other position."  We don't know Deep Blue's evaluation algorithm
  or how it was derived.
    
>    The future may be that we have a special form of chess developing,
>    namely computer assisted chess, just as we have simultaneous
>    exhibitions and 5 minute games.  

We already have this.

  There is consideration that the rules of  
  chess matches should be changed so that adjournments (suspensions of play
  that allow overnight analysis by the player, his team, and their computers)
  can be avoided.  

  Correspondence (play-by-mail) chess is intended to promote deliberate,
  analytic play, but more players are using computers to aid their research, and
  such use is becoming a matter of questionable sportsmanship.

  Every serious chess match has always involved analysis of the opponent's
  game history.  Now that analysis is being computerized to find weaknesses
  in those opponent's playing styles.

- tom powers] (a very mediocre but experienced chess player)

    re .48
    
    Dunno about Bill Buckner, but my point is that Deep Blue did not win
    game 6 because it played championship level chess, but because Kasparov
    made a blunder which many good club players would avoid. International
    Business Machines put a rather different spin on it. 
    
    ..Kevin..

    It would have only been fair if you could have found a human who
    doesn't get tired, can think at several thousand of moves per second,
    and can remember everything [s]he's been taught about chess at an
    instant.  Then, and only then, could we compare their chess ability.
    
    The only thing it proved to me is that computers are very good at
    fast data manipulation and information retrieval and processing.
    Guess what?  I already knew that...
    
    				-John

re Note 5277.49 by MKTCRV::MANNERINGS:

        > International Business Machines put a rather different spin
        > on it. 
          
        You can call them by their initials, IBM -- they don't mind.
        :-}

>>    It would have only been fair if you could have found a human who
>>    doesn't get tired, can think at several thousand of moves per second,
>>    and can remember everything [s]he's been taught about chess at an
>>    instant.  Then, and only then, could we compare their chess ability.

[NOTE: I've quoted the above purely as an example; there are many comments in
this thread that share a similar view, in my opinion.]

Oh, come on.  Are you going to claim that a game between a young prodigy with
nothing else to remember and an older wordly expert is unfair because the
prodigy can remember everything he's been taught?  Are you going to claim that
a race between a horse and horseless carriage is unfair because the horse
doesn't have a gasoline engine?

How many human players have taken advantage of superior physical endurance,
mental toughness, and memory to defeat other humans?  Is that unfair?

When computers do additions at great speed, and beat humans doing the same
task, do you cry foul as well?

Or do you want to handicap the computer so that it cannot take advantage of its
own assets (doesn't sleep, single-minded attention to a task, total recall,
etc.)?  Why is that more fair?

What is so special about chess that the fact that it has been cracked to some
high degree by a computational method gets people so upset?  It's like the
world has ended or something.  Chess is fully open to mathematical analysis;
the only reason it resists the challenge is because it's complex.  A game such
as bridge has a psychological element to it, where you try to mislead your
opponents or to communicate something to your partner with a highly limited
language, but chess is right there on the board, for all to see.  I fully
expect computer bridge players to challenge champions, psychology or not; why
not chess?

						Brian

    The chess aspects of this are interesting, but the marketing of it is
    astounding. IBM has had front page coverage on every newspaper in the
    country for days. Now its shifitng to more detailed articles like
    todays Globe Business section with a story about the market for
    supercomputers.
    
    I have no idea what it cost IBM to carry this out, but Id guess not a lot
    in the grand scale of things-a few dedicated guys woring on a fun
    project, throwing a few $K at special boards. Do you want to guess what
    the advertising is worth? You cant buy a column on the front page of
    every daily newspaper for any ammount of money, yet IBM just got that
    coverage. 
    
    Its to late to do chess, but we need to come up with some similar use
    of computers, throw the ammount of money at it that it cost to produce
    one monkey ad, and discover real marketing. 
    
    

>> IBM has had front page coverage on every newspaper in the country for days. 

Make that.. every newspaper in every other country in this small planet.

    RE: .52
    
>Oh, come on.  Are you going to claim that a game between a young prodigy with
>nothing else to remember and an older wordly expert is unfair because the
>prodigy can remember everything he's been taught?  Are you going to claim that
>a race between a horse and horseless carriage is unfair because the horse
>doesn't have a gasoline engine?

    Lets just call the comparison for what its worth, "who can do things
    faster, a human or a computer?"

>What is so special about chess that the fact that it has been cracked to some
>high degree by a computational method gets people so upset?  It's like the
>world has ended or something.  Chess is fully open to mathematical analysis;
>the only reason it resists the challenge is because it's complex.  A game such
>as bridge has a psychological element to it, where you try to mislead your
>opponents or to communicate something to your partner with a highly limited
>language, but chess is right there on the board, for all to see.  I fully
>expect computer bridge players to challenge champions, psychology or not; why
>not chess?

    No strategy or psychology in chess?  Surely you don't think that, right?  
    I would think that bridge is an easier problem to solve, not a harder 
    problem...  

				-John

I think the point was that you can't "bluff" in a chess game against 
a computer that's processing every possible game outcome after every 
move.

Sure, you can "bluff" by offering the computer your queen so that you 
can move in with the rook within the next two moves for an easy mate, 
but the computer will see that very easily.  A human might miss it, 
for whatever reason.

And I don't know much about bridge [IE, how many cards are dealt ... 
all of them?], but there is a "hidden element" in bridge in that the 
computer does not know which cards are held by which players, or in 
which order they will be played.  In chess, all of the pieces are in 
plain sight, so there are no "surprises".

re Note 5277.53 by LEXS01::GINGER:

>     Its to late to do chess, but we need to come up with some similar use
>     of computers, throw the ammount of money at it that it cost to produce
>     one monkey ad, and discover real marketing. 
  
        Well, we did AltaVista Search.

        (I know, I know -- not quite the same kind of thing.  But in
        a way, it's the kind of thing we would do.)

        Bob

IBM is just doing what IBM has always done best: Marketing.
The event has almost no business merit otherwise, aside from a
rather esoteric curiosity.  They wouldn't have done it
at all, had they not had significant reason to believe
they'd win, or failing to win, at least good exposure.

And we're doing just what we've always done: talking
about how we could do that too.  If we had thought of it.

We didn't.  That has almost no business merit either.

tim

> <<< Note 5277.57 by LGP30::FLEISCHER "without vision the people perish (DTN 381-0426 ZKO1-1)" >>>
> 
> re Note 5277.53 by LEXS01::GINGER:
> 
> >     Its to late to do chess, but we need to come up with some similar use
> >     of computers, throw the ammount of money at it that it cost to produce
> >     one monkey ad, and discover real marketing. 
>   
>         Well, we did AltaVista Search.
> 
>         (I know, I know -- not quite the same kind of thing.  But in

That is true.

AltaVista Search is only of interest to those who use computers.

IBM has really trounced all competitors with Deep Blue in terms of
gaining the mindshare of those who are not potential customers.
The universality of their PR coup should consider in that light.

    Forget the chess, today we *own* the financial news airwaves.

    according to NEWSBYTES (05/14/97) American Computer Co. a maker of
    personal computers, servers, and parallel-processing supercomputers,
    has challenged IBM's Deep Blue chess computer to a match. Jack Shulman,
    president of American Computer told that his company would put up
    against Deep Blue one of its Valkyrie or Vulcan systems, with from 64
    to 16,534 Inter Corp. Pentium or Digital Equipment Corp. Alpha
    processors running in parallel.... He would say, though, that "I don't
    think that Deep Blue has the processing capacity of the machine we're
    talking about"

    >Jack Shulman, president of American Computer told that his company would 
    >put up against Deep Blue
    
    More power to that man! That is the way to go!
    
    ..Kevin..
    
     

    Yabbut,
    
    Who's gonna program it?

    >>Who's gonna program it?
    
    There must be some hungry chess brains out there we could jointly
    sponser with other entities?
    
    ..Kevin..

And what would you accomplish?  More free advertising for IBM!

				Steve

IBM is gaining even more leverage from their publicity.
They have an ad on now that pits NBA star David Robinson
against Deep Blue in a one-on-one basketball game.
Robinson wins, 21-0, and his parting comments to Deep Blue
are "stick to Web service."

- tom]

    All that ad proves is that chess is not a real sport.

    All that ad proves is that IBM is light years ahead of DIGITAL when it
    comes to marketing.

Re. Chess and bridge games

If I remember well from the game theory (this was twenty years ago ..)
Chess and bridge are different by the fact that chess is a 'deterministic' game
while bridge is a 'non-deterministic' game (I don't know if this is the correct
translation or terminology). 
This means that in chess all the future moves may be (theoretically) determined 
as there is no random factor, while in bridge the card distribution is random 
and also because the players do not see each other's cards.
Thus chess is a game where the graph (web ?) of all the possible moves 
(and situations) may be (theoretically) known by both players. In such games
there is only one winner and allways the same (as long as it follows a winning 
path), based on the fact that the starting position is in the kernel of the 
game graph or not. Currently it is not known who should be the winner (the first
or the second player) for the chess game. 
I will not go into further details but, to conclude, the day a machine will 
be able to develop the graph of the chess game, the machine should allways win
if starting as 'winner' player. Even starting as 'loser' the machine could
take advantage of the first non-winning path move of the other player and
become 'winner'.
Now, to completely determine the chess game graph (or web) .. this is really
a hard job even for today's computers ;-)
I do not know much about how bridge can be approached but I think it's more
diffcult than for chess.

Sorin

>Chess and bridge are different by the fact that chess is a 'deterministic' game
>while bridge is a 'non-deterministic' game (I don't know if this is the correct
>translation or terminology). 

  Yes, that is the correct terminology.

    Re .70:
    
    > Yes, that is the correct terminology.
  
    I do not believe it is.  More important from the game theory point of
    view is that chess is a game of complete information while bridge is
    not -- in chess, both players know the entire state of the game, but
    that is not so in bridge.
    
    Bridge is non-deterministic before the shuffle, but the state of the
    game is fixed after the shuffle; there is no more randomness.  Does
    that mean bridge after the shuffle plays like a chess game?  No,
    because information is incomplete even though it is determined.
    
    
    				-- edp
    
    
Public key fingerprint:  8e ad 63 61 ba 0c 26 86  32 0a 7d 28 db e7 6f 75
To find PGP, read note 2688.4 in Humane::IBMPC_Shareware.

>>    > Yes, that is the correct terminology.
>  
>    I do not believe it is.  More important from the game theory point of
>    view is that chess is a game of complete information while bridge is
>    not -- in chess, both players know the entire state of the game, but
>    that is not so in bridge.
>    Bridge is non-deterministic before the shuffle, but the state of the
>    game is fixed after the shuffle; there is no more randomness.  Does
>    that mean bridge after the shuffle plays like a chess game?  No,
>    because information is incomplete even though it is determined.

Eric is right.  Though I can't speak to the game theory terminology,
there are at least three classes of games:

  Chess, checkers, go, and others are games of "perfect knowledge."
  They are not (or need not be) driven by random events or probabilities,
  all participants can theoretically known every possible path from a given
  point.

  Bridge and other card games that use the complete deck are deterministic
  but are games of "imperfect knowledge."  Inferences can be drawn
  based on deterministic evaluations of probability.  Knowledge grows
  during the course of the game, and may become perfect at some point
  before its conclusion, or may remain incomplete until the second
  to the last card is played.

  Backgammon is a non-deterministic game, driven at each turn by the roll
  of the dice.  Knowledge of the state of the game is, in one sense, perfect,
  but the course of the game cannot be known in advance.

- tom]

> Currently it is not known who should be the winner (the first
> or the second player) for the chess game. 
    
I thought that what is unknown is whether or not white has a forced
win, i.e. it is known that black does not have a forced win.

There's some proof of this roughly of the form:

    Assume black has a forced win.  For white's first move, make some
    innocuous move that has no bearing on the outcome of the game.  If
    Thereafter, play as black.  White now has a forced move.  
 
I've probably misremembered the proof.

                                                August G. Reinig

>          <<< Note 5277.73 by CXXC::REINIG "This too shall change" >>>
>
>> Currently it is not known who should be the winner (the first
>> or the second player) for the chess game. 
>    
>I thought that what is unknown is whether or not white has a forced
>win, i.e. it is known that black does not have a forced win.

I don't think that it is proven either way.
Experience would indicate that if there is a forced win, it is more likely
to be white's, but that's not a proof.
(In decided games, that is, games that do not end in a draw, white
wins about twice as often as black.  Exact statistics depend on the level
of the players involved.)

- tom]

From the Sunday Times, reprinted in the Weekend Australian:

The victory of IBM supercomputer Deep Blue over world chess champion Garry 
Kasparov two weeks ago may not be a simple triumph of machine over man after 
all. According to IBM, Deep Blue's number-crunching ability will be used within 
two years to model and test new drugs. IBM claims this could reduce the average 
12-year development period for a drug to about six years. Powerful computers now 
used to speed up development can take one or two months to determine a 
nanosecond's worth of interaction between 1000 atoms. A Deep Blue supercomputer 
could model, in one day, the split-second interaction of one million atoms.

    I'm just glad they havent started showing ads with drug researchers
    sitting around in white coats playing chess against a computer with the
    slogan: "Dr. Salk earning his next Nobel Prize", or somesuch.
    

    Well, we may have a response to this: I recall hearing at a Unix
    Symposium that the gene thought to be responsible for diabetes was
    found using an Alpha. Does anyone know the facts about this?
    
    ..Kevin..

See also http://sawww.epfl.ch/SIC/SA/publications/SCR95/7-95-63a.html

PJDM

Cartoon in last week's New Yorker magazine:

Man looking at his microwave oven and saying, "No, I don't want to play chess, 
I just want you to reheat the lasagna."