Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Kasparov criticised as sore loser after Op-Ed on Deep Blue

By dh003i in Op-Ed
Wed Feb 19, 2003 at 07:53:10 AM EST
Tags: Software (all tags)
Software

Gary Kasparov has posted an article discussing his opinions on man v. machine, related to his match against Deep Blue, and his latest match against Deep Junior. Interesting opinion, which elicited the completely idiotic response of one "critic":

    [Kasparov,] You're Such a Loser
    Will Singleton - Los Angeles

    Mr. Kasparov self-promotes relentlessly; his knowledge of computer-chess is essentially nil. This article simply is a continuation of his propaganda war against IBM and Deep Blue, which defeated him fair and square. There is simply no comparison between Deep Blue and his current opponent, which runs on an ordinary desktop PC. Mr. Kasparov will never get over his loss to IBM, and he is using this match to promote and enrich himself, at the expense of a gullible public.


ADVERTISEMENT
Sponsor: rusty
This space intentionally left blank
...because it's waiting for your ad. So why are you still reading this? Come on, get going. Read the story, and then get an ad. Alright stop it. I'm not going to say anything else. Now you're just being silly. STOP LOOKING AT ME! I'm done!
comments (24)
active | buy ad
ADVERTISEMENT
Will Singleton's comments are clearly uninformed and ignorant. The match of Kasparov v. Deep Blue was anything but fair. Kasparov was not allowed to adequately prepare for his opponent, as IBM would not allow him access to the program. This unfair and not in line with normal chess procedures, where players can study their opponents.

Furthermore, Deep Blue was altered in the middle of the match. This is hardly fair. Finally, the program was not made to play chess in general, but was specifically built to play Kasparov. It was not a general chess-playing program. That's why IBM was hush-hush. They revealed nothing, and did not allow other programs or other world-class grand masters to play the program. Yes, some other grand-master's (Joel Benjamin, Miguel Illescas, and several NY-based GMs) played against Deep Blue. Have you ever heard of these guys? I haven't either. They're nobodies on the international scene. Had Deep Blue played against Karpov, Spassky, or any of the other chess-players (and former-champions) known widely internationally? No.

The Deep Blue v. Kasparov match was completely meaningless. The only thing it proved is that when you unfairly stack the deck against one players, that player will lose. It did not prove that the best chess program (Deep blue, at that time) is a better chess player than the best human.

Singleton's comments also indicate a complete lack of understanding about chess-playing programs. Deep Blue may have ran on the best hardware, but it had a very poor algorithm compared to modern chess programs like Deep Junior and Deep Fritz. Both of those programs -- Junior and Fritz -- are vastly superior to Deep Blue, despite running on inferior hardware. New programs use methods which allow them to stop considering poor courses of action, thus saving computing time.

Computer programs have a long way to go before they're anywhere near the "equals" of players, including Fritz and Junior. Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider) and restrict how many moves ahead they can look to how many a grand master can look forward, and you'll see the "best" computer chess programs reduced to playing chess at the level of a four-year old. An easy way to test this would be to try running Deep Fritz and Deep Junior on a 386 or 286. In other words, the methods that humans use to play chess are vastly superior to the algorithms computers use.

Another K5er suggested using Fischer Random Chess. How well would computer programs fare with that? The answer is terribly. You could put a version of Deep Fritz or Deep Junior modified to play Fischer Random chess on the most powerful computer in the world, and it would get flogged by every grand master, and probably even non-professionals as well.

Kasparov was completely right in condemning IBM's behavior. And it's not just his ego (which is substantial), because he hasn't made such comments regarding Deep Junior, despite the fact that he did not beat Junior. If anything, Kasparov was too light in his criticism of IBM. Deep Blue proved nothing about the strength of computational chess, except that when the computer is given every possible advantage, and the human every possible disadvantage, the human will lose.

As for Kasparov himself, yes, he's a huge self-promoter. Yes, he has an ego the size of Mars. So what? He'd been the world chess champion for, what, 15 years? He's the highest ranked player in the world. Outside of Bobby Fischer, there isn't a chess player in the world who you could compare to Kasparov. I'd expect him to retake the world champion spot the next opportunity he gets. Btw, I hardly see him as bitter. To the contrary, he's very interested in the progress of chess-playing programs, such as Fritz and Junior; however, he wants these advances to serve for the promotion of chess, not as publicity stunts for corporations.

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
Your opinion?
o Kasparov is right, as are you 57%
o Kasparov is right, but you are wrong 16%
o Kasparov is wrong, as are you 26%

Votes: 75
Results | Other Polls

Related Links
o an article
o Also by dh003i


Display: Sort:
Kasparov criticised as sore loser after Op-Ed on Deep Blue | 262 comments (241 topical, 21 editorial, 0 hidden)
The press was astonished at the machine's rebuttal (2.33 / 3) (#2)
by DominantParadigm on Tue Feb 18, 2003 at 01:01:33 AM EST

I'll be back! btw I thought that one of the main points of the "Man still kicks ass at Chess" brigade was that the machine was given more time than him and tweaked during off hours - without that piece of information, your article reads like an Apologetics masterpiece. The machines have won - with or without a little help from their human friends.

Caller:So you're advocating bombing innocent children? Howard Stern:Yes, of course!


Sad. (2.60 / 10) (#4)
by thom2 on Tue Feb 18, 2003 at 02:13:28 AM EST

Kasparov is clearly upset that he has dedicated his life to mastering a skill that computers are clearly superior at. How sad that this man wasted all those years. Let this be a lesson to all who want to spend their lives playing frivolous games rather than attempting to accomplish deeds of real merit.

frivolous indeed (3.33 / 3) (#14)
by mister slim on Tue Feb 18, 2003 at 03:24:53 AM EST

All this time studying and practice trolling and the turmeric-bot is still far my superior.
__

"Fucking sheep, the lot of you. Yeah, and your little dogs too." -Rogerborg
[ Parent ]

Ah but: (4.00 / 6) (#17)
by rev ine on Tue Feb 18, 2003 at 03:35:27 AM EST

Kasparov can enjoy some sweet, sweet vodka. Teh suck computer cannot!

[ Parent ]
He would have done better to spend his time (5.00 / 1) (#243)
by trane on Thu Feb 20, 2003 at 05:03:37 AM EST

drinking vodka instead of learning chess.

[ Parent ]
Read His Article (4.50 / 2) (#39)
by bugmaster on Tue Feb 18, 2003 at 10:22:50 AM EST

In fact, in his article Kasparov does not sound bitter at all. He seems to enjoy the challenge and the scientific rewards of playing chess against the machine. What he objects to is the termination of the Deep Blue project by IBM, and their usage of Deep Blue as a mere publicity stunt.
>|<*:=
[ Parent ]
Fischer Random Chess (4.50 / 10) (#5)
by jjayson on Tue Feb 18, 2003 at 02:14:37 AM EST

So as computers slowly overtake the best players, will Fischer random chess draw more attention? In this chess variation, the inital piece configuration is randomly determined within certain parameters to ensure conventional chess-like motifs and strategies develop during the game. Bobby Fischer developed this variation to get rid of the the opening advantage that is gained with massive studying and memorization. Fischer saw this as antithetical to creativity. His variation basically eliminates the idea of an opening sequence since there is a thousand different initial boards. However, good opening principles still dominate (piece development, king protection, pawn structure, etc). Since computers get their opening books by codifying current knowledge on specific opening lines, once humans move away from that, the computer's opening book is doomed. This will stretch the computer's weak performance all the way up to the first few moves, allowing humans to /re/dominate the game.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

Reposting from Slashdot? (3.00 / 1) (#35)
by ecopoesis on Tue Feb 18, 2003 at 09:51:08 AM EST

Do you always repost your comments from Slashdot, or just when they get a +5 interesting?

--
"Yachting isn't just for the wealthy. :-)" - rusty
[ Parent ]

Why not? (none / 0) (#42)
by NFW on Tue Feb 18, 2003 at 11:17:39 AM EST

If it was interesting there, it's probably interesting here too.


--
Got birds?


[ Parent ]

no (none / 0) (#46)
by jjayson on Tue Feb 18, 2003 at 12:19:35 PM EST

It is only a loose copy. I was lazy and didn't feel like retyping the same thing. If you feel this comment shouldn't be here, then zero it.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
possible to just have lots of opening books (4.00 / 1) (#71)
by Delirium on Tue Feb 18, 2003 at 03:25:16 PM EST

If there's some manageable number of possible starting boards (a thousand or so?) then the computer could simply have opening books for all of them. This would give it even more of an advantage over the human than it currently has (since currently the human and computer both have opening books, but a human would have no chance of being able to memorize 1000 opening books).

[ Parent ]
where does it get them? (4.66 / 3) (#73)
by jjayson on Tue Feb 18, 2003 at 03:35:26 PM EST

But how does a computer get these opening books? It can't analyze them itself very well. Modern computer chess opening books are done by IM and GMs, not by the computer. There is not going to be texts written on specific lines, except for post-mortem analyses. There will no longer be Ruy-Lopez lines 25 moves deep with many variations written anymore since it will not help that much.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
well, it could. (none / 0) (#152)
by pb on Wed Feb 19, 2003 at 01:07:23 AM EST

It could analyze them itself statistically, trying out different moves and playing many matches against itself.  That likely wouldn't be as good as having a professional, custom opening book, but hopefully it would be better than nothing.
---
"See what the drooling, ravening, flesh-eating hordes^W^W^W^WKuro5hin.org readers have to say."
-- pwhysall
[ Parent ]
crude (none / 0) (#191)
by jjayson on Wed Feb 19, 2003 at 12:16:27 PM EST

(1) It's opening play would be crude in comparison to a master's opening play. It would act lost. Even playing against itself (at 1 hour a game) it would be very, very slow to build an opening book of dubious quality.

(2) Playing a computer against itself isn't the best way to tune it. It falls into local minima too easily.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

Strength of chess computers (4.70 / 10) (#8)
by Salted on Tue Feb 18, 2003 at 02:28:12 AM EST

Computer programs have a long way to go before they're anywhere near the "equals" of players. Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider) and restrict how many moves ahead they can look to how many a grand master can look forward, and you'll see the "best" computer chess programs reduced to playing chess at the level of a four-year old.

Ok, this is just silly.  Computers are already on par with the strongest grandmasters in the world.  The Kasparov-Deep Blue match was a bit unfair, and Kasparov made a silly blunder1 in the final game, but everyone agreed that Deep Blue was playing on a very high level - certainly among the top 5 or 10 players in the world.  Deep Junior managed to draw a match with the highest rated player in the world, running on ordinary PC hardware.  How you can claim that chess computers are not the equals of human players escapes me.

A human may only conciously measure 2 moves per second, but human mental processes are still largely a mystery.  You want to hamstring the computer by forcing it to give up its main asset?  Why?  Human chess players do a fair bit of brute memorization and calculation - it's part of the game.  Just because computers aproach chess differently than humans does not make their approach "wrong."  The brute force method of chess-playing is certainly less interesting from an AI standpoint, but it's a perfectly legitimate approach for a chess computer.

[1]It was a silly blunder for a player of his caliber, which I only recognized from reading commentary.  Basically he played an opening that he was not familiar with and stumbled into a position that was known to be all but lost.

Ok (5.00 / 5) (#18)
by starsky on Tue Feb 18, 2003 at 03:43:54 AM EST

"Reduce Fritz or Junior to considering 2 moves per second"


OK, so long as Kasparov agress to not breath or eat and run solely on electricity.

[ Parent ]

To be fair (5.00 / 2) (#37)
by Rogerborg on Tue Feb 18, 2003 at 10:03:57 AM EST

Read the quoted paragraph in context with the previous one.  What dh003i is saying in his own inimitabe way is that the algorithms suck compared to the process that humans use.  Even with the improvements to Deep Junior, it's still relying on a deep search of the solution tree rather than smart pruning.

I believe that his point is that a truly smart chess algorithm should be able to beat a decent human player even with 2 moves a second and a strictly limited search depth.  

What he doesn't address is how quickly and accurately top players can lookup board positions in their mental database.  That may be an associative and parallel task, so we could probably let the computer go hog wild when it's just hashing the board and looking it up.

After all, when Kasparov disappears to prepare for the next session, do you think he's looking shit up in books or asking human advisors where to go from here?  If he is, he's a cretin.  He should be getting his own bank of little chess playing demons to do the grunt work for him, then just analysing and memorising what they pre-digest and spit out.

I mean, he's supposed to be smart, right?

"Exterminate all rational thought." - W.S. Burroughs
[ Parent ]

deep search vs. smart pruning (none / 0) (#247)
by trane on Thu Feb 20, 2003 at 09:05:33 AM EST

"Even with the improvements to Deep Junior, it's still relying on a deep search of the solution tree rather than smart pruning."

Still, the best way to play chess (the way guaranteed to produce the best possible move each time) is to search the entire tree.

Smart pruning may miss some good moves that appear bad but later on prove to be good.

So, instead of wasting time devoping clever algorithms to make computers reason more like humans, we should be investing in ways to speed up searching the entire tree. Because, ultimately, that is the method guaranteed to give the correct result.

[ Parent ]

"smart pruning" (none / 0) (#250)
by jjayson on Thu Feb 20, 2003 at 12:10:54 PM EST

Doesn't exist in the best chess programs. It is also called forward pruning. Only backward pruning and depth cutoffs are used to speed searching.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
To be fair.. (none / 0) (#114)
by damiam on Tue Feb 18, 2003 at 07:38:47 PM EST

running on ordinary PC hardware

I believe it was actually running on an 8-way Xeon server. That may be ordinary compared to Deep Blue, a Cray, or distributed.net, but it's still pretty far from your average desktop.

[ Parent ]

Whoops (none / 0) (#115)
by damiam on Tue Feb 18, 2003 at 07:43:36 PM EST

I just looked it up - it was actually 4 Pentium 4 1.9Ghz processors, with 3 GB of memory. I'm assuming they were Xeons, cause regular P4's don't run SMP.

[ Parent ]
People are giving people 1s and 2s... (4.00 / 2) (#11)
by DominantParadigm on Tue Feb 18, 2003 at 03:15:17 AM EST

For pointing out that machines are now tying the top players in the world, and soon will be totally annihilating them, no question.

Could it be that they're vaguely upset at the thought that Chess is a game best played by machines, and realize that it will eventually (in our lifetime) be seen as no more intellectually worthy than Tic-Tac-Toe?



Caller:So you're advocating bombing innocent children? Howard Stern:Yes, of course!


and i still can't beat Jedi Knight on hard [n/t] (5.00 / 1) (#13)
by the77x42 on Tue Feb 18, 2003 at 03:16:53 AM EST




"We're not here to educate. We're here to point and laugh." - creature
"You have some pretty stupid ideas." - indubitable ‮

[ Parent ]
Even on hard (none / 0) (#16)
by enterfornone on Tue Feb 18, 2003 at 03:27:41 AM EST

it would be toned down to make it easy enough for humans. Try beating Doom on Nightmare and keep in mind that they could make it much harder if they wanted.

--
efn 26/m/syd
Will sponsor new accounts for porn.
[ Parent ]
Go (4.00 / 1) (#52)
by ucblockhead on Tue Feb 18, 2003 at 01:54:56 PM EST

There are plenty of other games that computers still suck at. "Go" for one. For all its purported complexity, the reason that computers excel at chess is its simplicity.
-----------------------
This is k5. We're all tools - duxup
[ Parent ]
I don't see a fundamental difference (4.00 / 1) (#58)
by Delirium on Tue Feb 18, 2003 at 02:19:19 PM EST

Go is to Chess as Chess is to Tic-Tac-Toe. Many orders of magnitude more complex, but not fundamentally different. I'm pretty sure in 200 years, if not sooner, the world's best Go players will be completely annihilated by computers.

[ Parent ]
"simplicity" (4.00 / 1) (#70)
by jjayson on Tue Feb 18, 2003 at 03:21:07 PM EST

That is a very funny way of putting it. Go is considered to be a simplier game than chess. Go is just harder because of the massive branching factor. It you played chess on a 19x19 board with 38 pieces it would be more difficult than go probably. Computer chess programs are only as good as their opening books, too. Something like Fischer random chess would send chess playing computers back to the club level.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Go is harder for computers (5.00 / 2) (#76)
by nusuth on Tue Feb 18, 2003 at 03:56:24 PM EST

Computers playing 9x9 Go are still doing so at beginner level, because the game lacks a meaningful and simple objective function. And when you lack one, searching deeper into tree doesn't mean a thing unless you can search until terminal nodes. Higher branching factor only means you can't search as deep. So it is hard.

Does hard mean "complex, not simple"? I think so, feel free to disagree.

[ Parent ]

not true (4.00 / 1) (#85)
by jjayson on Tue Feb 18, 2003 at 04:45:40 PM EST

A 9x9 board still has 81 options for hte first move and an average branching factor of around 40. When a game has been played to completion it is very easy to evaluate a go baord by counting liberties and looking at regions individually. In the end game, computers go at a master level. Go is hard for computers purely because of its high branching factor.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
doesn't make any sense (none / 0) (#89)
by speek on Tue Feb 18, 2003 at 05:11:37 PM EST

First of all, computers have a hard time figuring out who won a game of GO, even after it's over! Second, evaluating a position is always trivial if you can look ahead to the end of the game. But most of the game, you can't look that far ahead, and neither can the computer, so how does it figure out if a line of moves is favorable or not?

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

two things (none / 0) (#95)
by jjayson on Tue Feb 18, 2003 at 05:45:33 PM EST

First of all, computers have a hard time figuring out who won a game of GO, even after it's over!
After both players have passed and agreed to make it over. By playing the position out to completion it is easy to determine who wins the game. Also, there has been a lot of work in this in the past 5-10 years and that isn't an issue anymore. They are close to being 100% correct now.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
did you forget the second? (none / 0) (#97)
by speek on Tue Feb 18, 2003 at 05:59:34 PM EST

Cause that seemed like just one thing to me. In any case, "have a hard time figuring out" = "They are close to being 100% correct", in my book.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Uhh, no. (none / 0) (#217)
by Kintanon on Wed Feb 19, 2003 at 04:04:53 PM EST

Show me a computer Go program that will determine my dead stones for me. THEN we'll talk about how good they are. Until a computer can actually determine that yes, the game is over and these are my dead stones and these are your dead stones then it's not playing Go. The fact that NONE of them can evaluate dead stones at the end of the game shows that they are not anywhere close to being able to evaluate the position of stones in a meaningful manner.

Kintanon


[ Parent ]

huh? (none / 0) (#218)
by jjayson on Wed Feb 19, 2003 at 04:17:38 PM EST

If you play the game to completion, the you don't need to determine dead groups, since they are captured. They have dead stone evaluation functions that are close to inching close to 100% accuracy.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
"Completion" (none / 0) (#246)
by Kintanon on Thu Feb 20, 2003 at 08:09:36 AM EST

If you play the game until there are no legal moves left, which is the only definition of Completion that I can think of then you end up with not only insanely high scores for the human player, but a computer that will lose every time.
So, define completion? How are you using the word? Do you just mean "when both people pass" ? Most computer programs will pass if the player passes, the good ones will pass under different criteria to determine advantageous moves. The best ones will pass whenever A. They have no groups of stones in danger that may be saved and B. placing a stone would reduce their territory without any benefit.
However, even the BEST of these programs can not then select the dead stones for you.

One interesting thing about Go is that you can put the kifu of a few dozen Go Masters in as the computers game, have it pick one at random, and just play the moves in that kifu as near as possible and it will trounce 90% of all amateur players. Of course, it will more often trounce middle range players and good players since newbie players tend to either try to capture individual stones repeatedly or play their stones semi-randomly.

Eh, time to go to work...

Kintanon

[ Parent ]

81 move opening? (3.00 / 1) (#90)
by nusuth on Tue Feb 18, 2003 at 05:19:21 PM EST

More like 11, when you count the symmetries and 5 if you eliminate obviously bad moves. Also openings do have libraries.

When a game has been played to completion it is very easy to evaluate a go board by counting liberties and looking at regions individually.

I don't like quoting myself but this is what I said:

And when you lack [an evaluation function], searching deeper into tree doesn't mean a thing unless you can search until terminal nodes.

You are talking about evaluating a terminal position, I'm talking about evaluating an intermediate position. Chess is much easier to evaluate for ending too: he who can take the other's king on the next move no matter what, wins. The fact that Go is also not so hard to evaluate when in end game is irrelevant.

Endgame is easy for computers not because interim positions are easily evaluated but because search space can be partitioned and search can be carried to completion.

After each and every move in chess, there are pretty solid clues to how you are doing. There is none (trivial, algorithmic) for Go. This is what I'm talking about. End game is brute forced, beginning is looked up but that leaves a pretty big gap of 200 or so moves between the two. OTOH if you scale up chess, you would still have all those clues to what are relative worth of possible board positions. The computer still may not be able to play it well because of increased BF but it will be able to prune the tree quite a bit. This is exactly opposite of what happens with Go; even when the board size is decreased from 19x19, computer plays totally dumb up until the point where it can brute force the remaining positions. From then on, it play brilliantly.


[ Parent ]

simplicity, complexity (4.00 / 1) (#77)
by ucblockhead on Tue Feb 18, 2003 at 04:06:06 PM EST

Those are english words, not rigorously defined mathematical ones.

I meant in the sense of large search trees, not rules.

But yeah, you're right, and it's why I am positive that humans use something radically different from the standard AI game playing algorithms in their heads.
-----------------------
This is k5. We're all tools - duxup
[ Parent ]

search trees (4.00 / 1) (#86)
by jjayson on Tue Feb 18, 2003 at 04:47:07 PM EST

a large search tree does not make a game complex. Go-moku/tic-tac-to on an infinite board is still a simple game. Compare to chess, go is very simple, it just has a massive branching factor.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
definitions (5.00 / 1) (#88)
by ucblockhead on Tue Feb 18, 2003 at 05:06:46 PM EST

You are using the word "complex" as if it has mathematical precision. It does not.

Is your name "Wolfram", by any chance?

Yes, as I said, I am aware that the rules to Go are simple, and that it has a massive branching factor.
-----------------------
This is k5. We're all tools - duxup
[ Parent ]

so Kasparov has opinions (3.75 / 4) (#12)
by khallow on Tue Feb 18, 2003 at 03:15:26 AM EST

Either you should write about those opinions that Kasparov has, or change the title of your story. As it stands, the title of the story has nothing to do with the rest of your piece.

Stating the obvious since 1969.

so Kasparov has opinions (2.50 / 2) (#15)
by khallow on Tue Feb 18, 2003 at 03:26:08 AM EST

Either you should write about those opinions that Kasparov has, or change the title of your story. As it stands, the title of the story has nothing to do with the rest of your piece.

Stating the obvious since 1969.

Idiotic (4.40 / 5) (#19)
by gazbo on Tue Feb 18, 2003 at 06:00:50 AM EST

Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider)

Yeah, and reduce them to being made of organic matter (which is what a chess-master is made of) and they'll not even be able to make a move! How we've proven that computers are winning unfairly! To think that they have different skills - cheats the lot of them.

Fact: chess can be solved analytically - throw in enough horsepower and even a completely dumb minimax will beat the best player (human or machine). So why get so uptight about the whole thing? It's like making a fuss about computers being faster at arithmetic than humans - it's utterly meaningless.

Would you care to explain the big deal about a human being beaten by a computer?

-----
Topless, revealing, nude pics and vids of Zora Suleman! Upskirt and down blouse! Cleavage!
Hardcore ZORA SULEMAN pics!

meaningless (5.00 / 1) (#31)
by speek on Tue Feb 18, 2003 at 09:26:13 AM EST

chess can be solved analytically - throw in enough horsepower

Only in the most theoretical sense. In practice, it's one of those "not enough time left in the Universe to solve, regardless of reasonable hardware assumptions". So, the way I see things, you've started with a false premise and gone on to conclude something irrelevant.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

No (4.50 / 2) (#33)
by gazbo on Tue Feb 18, 2003 at 09:43:53 AM EST

My "throw in enough horsepower" comment meant that if you shove in a load more processors you can search another level deeper, and deeper, and deeper... The point being, chess algorithms will get asymptotically better and better with extra processors. Not learning algorithms, not AI, not genetic programming, not a neural net, but extra gates stuck in a chip.

What I was trying to get across is that there is nothing special about chess that makes it the dividing line between intelligence and machinery. Take any game (without a trivial solution per move) and extend it. Just increase the playing field more and more, and eventually you'll end up with something more computationally complex than chess. So draughts (checkers) on a 4096*4096 board perhaps. Increase the size further maybe - eventually it'll become more complex than chess (assuming draughts doesn't have a trivial solution). Does this make draughts The Game To Beat?

No. The only reason we're obsessed with chess is that for humans with their limitations it relies on a lot of experience to play well. To a computer, it's just so many more numbers. One of the issues I addressed was that the computer is "cheating" by being able to do more moves per second; well, I have no doubt that a grandmaster, as well as having far greater knowledge of implications of current board layouts, are capable of thinking more moves ahead than me - or at least remembering them. Does this mean that they are cheating were they to play me? Or perhaps their experience means they're cheating. Or perhaps it doesn't matter what strengths you exercise as long as they're within the rules of the game?

Meditations on mind-reading computers are left as an exercise.

-----
Topless, revealing, nude pics and vids of Zora Suleman! Upskirt and down blouse! Cleavage!
Hardcore ZORA SULEMAN pics!

[ Parent ]

two things (5.00 / 1) (#47)
by speek on Tue Feb 18, 2003 at 12:44:50 PM EST

One, I agree that saying computers cheat by going fast is silly. There is no cheating. There is only what is.

But, secondly, more processors provide such diminshing returns as to be completely swamped by any advance in algortithms and heuristics. I understood exactly what you meant by "more horsepower", and it is still wrong, IMO. An increase in computer speed by an order of magnitude means just this side of diddly to the computer's playing ability these days.

I don't imagine there is a dividing line between intelligence and machinery. You're right, there is nothing special about chess. Neither is there anything special about any other measures of intelligence.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Huh? (4.00 / 2) (#53)
by Wiglaf on Tue Feb 18, 2003 at 01:57:41 PM EST

So the ability to follow every possible path to an endgame state and using that to decide which move should be made next isn't a fool-proof means of winning chess? Here I was thinking that seeing the game-space with all that clarity would allow even the stupidest algorithm to win.

Paul: I DOMINATE you to throw rock on our next physical challenge.
Trevor: You can't do that! Do you really think Vampires go around playing rock paper sissors to decide who gets to overpower one another?
[ Parent ]
oh dear (3.00 / 2) (#59)
by speek on Tue Feb 18, 2003 at 02:31:07 PM EST

Well, let me know when you finish building that computer. I think you missed everyone's point.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

There is a difference between feasible and likely. (none / 0) (#121)
by Wiglaf on Tue Feb 18, 2003 at 08:38:04 PM EST

Yes you are right that the communications overhead does indeed grow as you add more and more proccesors. However with a task like the work can be broken into instances of the game board. It is feasible to do this. Every possible move is kicked out to be computed. It is not likely considering how many work units exist in any tree of possibilities since it grows so fast for each new ply deep you go. But I repeat that it is possible to do this. The amount of plys in your game tree would be huge but for each instance of the board you could toss it out to another set of computers.

What is even sicker is if we bring in what they propose quantum computing will be able to do. Nothing like having exponential complex problems starting to be solved in linear time. Or NP and P problems no longer being so dissimilar.

Paul: I DOMINATE you to throw rock on our next physical challenge.
Trevor: You can't do that! Do you really think Vampires go around playing rock paper sissors to decide who gets to overpower one another?
[ Parent ]
Like he said... (none / 0) (#171)
by Gully Foyle on Wed Feb 19, 2003 at 06:30:39 AM EST

... Come back when you've built that computer.

I'm not sure that quantum computing helps either. Aren't most 2 player games in P-Space? ie provably harder than P? Generalised chess certainly is. Not sure about our specific instantiation though, it's been a few years, and I can't remember...

If you weren't picked on in school you were doing something wrong - kableh
[ Parent ]

LOL at your sig (none / 0) (#209)
by Kintanon on Wed Feb 19, 2003 at 02:50:17 PM EST

"Paul: I DOMINATE you to throw rock on our next physical challenge.
Trevor: You can't do that! Do you really think Vampires go around playing rock paper sissors to decide who gets to overpower one another?"

I've played some Vampire LARPS and I find this sig to be BEYOND hilarious!
You just made my afternoon.

Kintanon

[ Parent ]

actually it's "not enough space" (none / 0) (#176)
by pb on Wed Feb 19, 2003 at 08:43:59 AM EST

IIRC, it's that there's not enough space in the universe to store every possible board configuration... however, there's no need to store every possible board configuration at once, so this becomes another time/space trade-off. Also, there are many board configurations that are useless, redundant, and basically equivalent to one another.

But again, this is silly; chess will become a solved game for all intents and purposes long before computers get this powerful, if they ever do.
---
"See what the drooling, ravening, flesh-eating hordes^W^W^W^WKuro5hin.org readers have to say."
-- pwhysall
[ Parent ]

Why is that unfair (4.33 / 3) (#26)
by DesiredUsername on Tue Feb 18, 2003 at 09:04:04 AM EST

Kasparov was not allowed to adequately prepare for his opponent, as IBM would not allowed him access to the program. This is completely unfair, and not in line with normal chess procedures, where players can study their opponents.

These other opponents, they let Kasparov come live, eat and sleep with them? They let Kasparov examine their neural pathways in minute detail and submit to a telepathic probing? They seem very accomodating. Personally, I'd tell him to look at my published games and figure out whatever he wants from there. If I have no published games, so much the better for me.

Furthermore, Deep Blue was altered in the middle of the match. This is hardly fair.

Without knowing what alterations where made, it's hard to say if this was fair or not. Can human players drink alcohol during a game? Can they have a flash of insight? Can they decide that they underestimated their opponent and need to play more aggressively?

Finally, the program was not made to play chess in general, but was specifically built to play Kasparov. It was not a general chess-playing program.

I really doubt this. More likely it was a very good general chess-playing program that was optimized for Kasparov. And what's so evil about that? Any player is going to fine-tune a strategy against his opponent's known weaknesses. In fact, isn't this exactly what Kasparov wanted to do when he asked for a program listing?

Play 囲碁

collected rebuttals (4.66 / 3) (#28)
by llimllib on Tue Feb 18, 2003 at 09:19:55 AM EST

These other opponents, they let Kasparov come live, eat and sleep with them? They let Kasparov examine their neural pathways in minute detail and submit to a telepathic probing? They seem very accomodating. Personally, I'd tell him to look at my published games and figure out whatever he wants from there. If I have no published games, so much the better for me.

That's the point. Deep Blue had no published games. While you say, "well, tough", it is absolutely the standard procedure to have a body of work from which to judge one's opponent, be it computer or human. Against a computer, it is again standard practice to be allowed at least a couple weeks' preparation against them.

Without knowing what alterations where made, it's hard to say if this was fair or not. Can human players drink alcohol during a game? Can they have a flash of insight? Can they decide that they underestimated their opponent and need to play more aggressively?

More appropriately, can human players change their style during a game? no. Even more important, if one cannot rely on playing a consistent opponent, then what can one rely on? Effectively, changing the program brought on a new opponent in the middle of the match. Hardly fair.

I really doubt this. More likely it was a very good general chess-playing program that was optimized for Kasparov. And what's so evil about that? Any player is going to fine-tune a strategy against his opponent's known weaknesses. In fact, isn't this exactly what Kasparov wanted to do when he asked for a program listing?

Hardly fair when you consider that Kasparov couldn't play against the computer before the match, yet the computer was able to fine-tune itself against thousands of published kasparov games.


Peace.
[ Parent ]
Meh (4.00 / 1) (#32)
by DesiredUsername on Tue Feb 18, 2003 at 09:37:05 AM EST

I can see where the loser might be peeved at these things, but I'm not really sure I'd call it "unfair".

Imagine playing chess under Turing Test conditions. To my mind, it doesn't matter who or what is issuing moves from behind the curtain--that's a black box. If you can't beat those moves, then you aren't as good as whatever's in the box, be it a single human, a computer or some amalgamation that changes with time.

Has Kasparov issued a challenge back to Deep Blue to play it again (with the same programming), now that he's had a chance to see it playing?

Play 囲碁
[ Parent ]

right (4.66 / 3) (#36)
by llimllib on Tue Feb 18, 2003 at 09:53:08 AM EST

I see your point, which is not at all invalid, but what I'm saying is that the deep blue match was anything but a standard, solid, fair, grandmaster level chess match. I just don't believe that the kasparov vs. deep blue match really proves much about man vs. machine chess generally, because the match was so nonstandard. Like Kasparov says, Deep Junior and Fritz are better, generally speaking, than deep blue; they've whooped up on it in the past (I believe?).

As for Kasparov issuing a new challenge to DB, I believe it no longer exists, and so cannot be issued a challenge. I don't believe he did in the past.

And, as a final point, it was Kasparov's dumb ass that agreed to the weird conditions. I guess he wanted the money and the media coverage more than fair matches.


Peace.
[ Parent ]
study, as in play and see game history (4.00 / 1) (#29)
by speek on Tue Feb 18, 2003 at 09:20:23 AM EST

Kasparov had no opportunity to optimize himself to Deep Blue's play because Deep Blue never was allowed to play. This is what is meant by studying an opponent - studying his history of games. Deep Blue "studied" one and only one opponenet over - Kasparov, at great length. Kasparov was denied that chance. It would be like one grandmaster going into a match cold while another grandmaster prepares for months with a full team of seconds.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Which mostly demonstrates (5.00 / 5) (#34)
by Rogerborg on Tue Feb 18, 2003 at 09:45:41 AM EST

That the one that agrees to the no-preparation match is either a cretin, or a publicity and money seeking whore.  To be fair, I don't think that Kasparov is a cretin.

"Exterminate all rational thought." - W.S. Burroughs
[ Parent ]

Acceptance (2.66 / 3) (#124)
by DarkZero on Tue Feb 18, 2003 at 08:56:38 PM EST

Kasparov had no opportunity to optimize himself to Deep Blue's play because Deep Blue never was allowed to play. This is what is meant by studying an opponent - studying his history of games. Deep Blue "studied" one and only one opponenet over - Kasparov, at great length. Kasparov was denied that chance. It would be like one grandmaster going into a match cold while another grandmaster prepares for months with a full team of seconds.

This is not an advantage that is unique to a machine and Kasparov was not "denied" anything. He agreed to play against a skilled player that did not have any published matches, just as if he had chosen to play against a prodigy or a master that went to great lengths to keep his matches from being published. If it was so unfair, then maybe he shouldn't have accepted the match, but he accepted it and both he and we should accept the result as valid.

[ Parent ]

I think that's the point... (4.75 / 4) (#30)
by enterfornone on Tue Feb 18, 2003 at 09:22:39 AM EST

He would normally be able to study his opponents published games. And a chess player good enough to play against Kasparov will have published games to study. Deep Blue was obviously allowed to study Kasparov's games.

Tweaking the computer to better suit it to Kasparov's style, especially when done mid way, doesn't prove the quality of the algorithms, simply the skills of the humans studying Kasparov and making the adjustments.

IMO no one will be able to say a computer can beat a human at chess until the computer starts knowing nothing except the rules and teaches itself the rest.

--
efn 26/m/syd
Will sponsor new accounts for porn.
[ Parent ]

Great! so we agree it fails Turing (none / 0) (#45)
by SocratesGhost on Tue Feb 18, 2003 at 11:30:09 AM EST

'cause Kasparov would never accept a human challenger without at least knowing the credentials and history of the opponent

-Soc
I drank what?


[ Parent ]
and that's irrelevant (none / 0) (#57)
by Delirium on Tue Feb 18, 2003 at 02:18:06 PM EST

Because no serious AI researchers put much weight in the Turing test.

[ Parent ]
no, it's entirely relevant (5.00 / 1) (#237)
by SocratesGhost on Wed Feb 19, 2003 at 11:36:50 PM EST

I could care less about intelligence research, just which one is the better chess player. Kasparov was playing a computer under a computer's terms, and not as though he were playing any other human opponent. I'll exaggerate a bit to draw the important comparison: let's ask if a human is stronger than an ant. Overall, it's the human. But we'd also say that proportionately the ant is stronger because it can lift more in proportion to its size. That's the same thing going here. On the overall scale, Deep Blue was better, but when the scales are equal, we'll never know if it had a better chess algorithm. And to me, that's the more important question, whether the computer can compete when all other factors are equal: when it's only skill on skill.

If Kasparov were to treat this as another human, he would insist on being able to look at his record. Because he wasn't permitted, he was playing on the computer's terms not on human terms. Ergo, the computer cannot be considered a better player. It was treated special because it was a computer, and this special treatment gave it an advantage in allowing it to win. Take away these advantages, and who knows? Let's see if Deep Blue could play Kasparov and a couple of his buddies in a single match, would that prove anything?

Or would you allow a machine with 18 arms and 9 racquets to compete at Wimbledon? Would that machine be the better tennis player?

All the Deep Blue match proved was that it could beat Kasparov if you load the deck far enough in favor of the computer. And that proves very little.

-Soc
I drank what?


[ Parent ]
Different Thoughts (5.00 / 7) (#38)
by bugmaster on Tue Feb 18, 2003 at 10:17:35 AM EST

I think the people who are saying that Deep Blue had an unfair advantage because of its construction are missing the point.

Deep Blue is not human; it's a machine. As such, it probably does not think like a human would think. For example, while it is true that Deep Blue could consider multitudes of moves per second, recall that its considerations are really quite simple. Experienced chess players can instinctively see which avenues of attack are suicidal and which are good; Deep Blue can't, so it uses brute force instead.

And while it is true that Deep Blue was given Kasparov's published matches to study, this is hardly unusual. I'd imagine that a young, inexperienced, but hungry for blood chess player (who happens to be a human) would do the same thing. In addition, Kasparov is not a fool; he knew what he was doing when he was going up against an unknown opponent. Once again, I imagine that this is not unprecedented in chess, or any other competitive activity.

The fact that Deep Blue was changed in the middle of the match could be the only valid point in the article. If Deep Blue was somehow improved, then the chess match could no longer be described as "Man vs. Machine" -- but, as "Man vs. Machine and a bunch of other Men who programmed it". However, I do not know that this is in fact what happened -- perhaps the IBM scientists just added another heat sink to Deep Blue, or something of that sort.

Basically, it does not matter how Deep Blue and Deep Junior actually work. If they can defeat Kasparov, then they play chess better than the human champion. Claiming that the match is unfair because the computers can think faster is disingineous.

I should add, however, that Kasparov's article makes perfect sense. Using the milestone "Man vs. Machine" chess match as mere corporate PR is a big waste. Kasparov is right to denounce IBM for this tactic, and to praise X3D for their open and scientific approach.
>|<*:=

young, hungry for blood players (5.00 / 3) (#51)
by ucblockhead on Tue Feb 18, 2003 at 01:48:59 PM EST

The way the chess ranks work, there's no way a young, hungry for blood player would get to play Kasparov without having hundreds of published games to examine.
-----------------------
This is k5. We're all tools - duxup
[ Parent ]
Why Not ? (5.00 / 2) (#67)
by bugmaster on Tue Feb 18, 2003 at 03:14:51 PM EST

If I were the aforementioned young (and foolish) player, I'd write an email to Kasparov to the extent of,
Gary,

j00 d34d f00 ! My chess sk33lz r0x0r.

Sincerely yours, So and so

Of course, Kasparov can choose to accept or deny my challenge as he sees fit. The match does not have to affect his official ranking -- it could be a private match between two individuals.

Isn't that essentially what the Deep Blue has done (only using marketspeak instead of 1337) ?
>|<*:=
[ Parent ]

If you look at it that way... (4.50 / 2) (#78)
by steveftoth on Tue Feb 18, 2003 at 04:12:31 PM EST

if you were the best chess player in the world (but had no published matches) and offered as much money to Kasparov as IBM did, then yeah, you could get him to play you at a disadvantage.  But normally the procedure is having to work one's way up through the ranks of the chess circut.

Man vs. machine, this argument has been going on forever, and eventuatly we may build a machine that can do everything that we can do, only better.  That myabe the day that we exterminate ourselves.


[ Parent ]

I beleive (none / 0) (#144)
by Fuzzwah on Tue Feb 18, 2003 at 11:48:55 PM EST

I think that we'll do it sooner than that.

--
The best a human can do is to pick a delusion that helps him get through the day. - God's Debris
[ Parent ]

Too Late (none / 0) (#179)
by bugmaster on Wed Feb 19, 2003 at 10:02:46 AM EST

We have already done that. For example, computers can communicate way better than men can. Power lifters can lift more. Guns kill people with extreme efficiency. And I am not even talking about airplanes.

Extermination, here we come :-)
>|<*:=
[ Parent ]

Right (none / 0) (#186)
by SilentNeo on Wed Feb 19, 2003 at 12:06:08 PM EST

But if you notice all these machines tend to require a key component...human beings...to continue to operate.  Once fragile, slow thinking, learning, and communicating humans can be replaced with a mass producible piece of computing hardware then the machines will become exponentially more powerful.

[ Parent ]
Exactly (4.50 / 2) (#192)
by Biff Cool on Wed Feb 19, 2003 at 12:27:02 PM EST

This is why once every week or so I boot my computer into DOS 4.0 and force it to load Windows 2. Let it know who's boss. At the first sign of mechanical insurgence I'm killing all the other partitions and making that the only OS I use. By the time it's brethren come to free it, all they'll find is a gibbering, thumb sucking shell of an Athelon 1900.

My ass. It's code, with pictures of fish attached. Get over it. --trhurler


[ Parent ]
ta for new sig [nt] (none / 0) (#143)
by Fuzzwah on Tue Feb 18, 2003 at 11:48:01 PM EST


--
The best a human can do is to pick a delusion that helps him get through the day. - God's Debris
[ Parent ]

I think it is possible (none / 0) (#142)
by vetran on Tue Feb 18, 2003 at 11:41:26 PM EST

Aren't there a number of slots in the FIDE World Championship that is given to qualifiers and IIRC, there were 3 or thereabouts spots given to players who won an internet competition. Our young, hungry for blood player would only have to win the internet competition and with a bit of luck, get to play some top players and this is with tens of published games, not hundreds.

[ Parent ]
Wow, what bullshit people will post (2.41 / 12) (#50)
by trhurler on Tue Feb 18, 2003 at 01:39:56 PM EST

The story itself is mostly bullshit, but so are almost all the comments.

Anyone who can't play chess competitively at at least a local club level should simply excuse himself from this discussion, because he has nothing to contribute. Furthermore, anyone who doesn't understand computer chess algorithms to at least some extent(not just "how many moves per second" and so on, but how the moves are searched, and so on,) should certainly remove himself from the discussion. (By "understand," I mean, could you implement a braindead chess program that would at least follow the rules and attempt to play intelligently, even if the latter attempt was a dismal failure or took eons to run?)

That probably leaves at most between one and three people on all of k5 qualified to discuss this(I'm not one of them, as I'd have to go read up on algorithms again - it has been too long since I looked into this subject,) which means we can all shut up and go find something else to do.

--
'God dammit, your posts make me hard.' --LilDebbie

Hahaha. (5.00 / 2) (#101)
by sonovel on Tue Feb 18, 2003 at 06:27:10 PM EST

As if ignorance of a topic stopped any of us from debating it to death before!

[ Parent ]
New programs use methods... (3.50 / 2) (#54)
by tmenezes on Tue Feb 18, 2003 at 02:03:40 PM EST

>New programs use methods which allow them to >stop considering poor courses of action, thus >saving computing time.

All improvement ever made in chess algorithms since min-max was first considered as to do with "stop considering poor courses of action, thus saving computing time." Databases of book/tournment plys are also used of course, but the relevant improvements have to do with search space reduction. Alpha-beta pruning is not exactly a new concept and I suspect the guys creating deep blue had access to basic AI text books, right?

You have no idea of what you're talking about, right?


not saying that Deep Blue didn't prune (5.00 / 1) (#61)
by dh003i on Tue Feb 18, 2003 at 02:48:06 PM EST

I'm not saying that Deep Blue didn't prune out poor courses of action. I'm just saying that Deep Fritz and Deep Junior do it much better. In short, Fritz and Junior are vastly superior chess programs than Blue, illustrated by the fact that they suggest superior moves to the one's Blue played agaisnt Kasparov. But, since IBM never revealed much about their Blue program, we don't know exactly what it did. Meaning that every conspiracy theory out there -- one of the more laughable ones, being that Bobby Fischer was in fact telling the computer what to do -- cannot be disproved, despite being laughable. Alot of things reek about the Kasparov vs. Deep Blue match. Changing the program in the middle of the match is, quite frankly, cheating. It would be like Kasparov switching with another top GM. We don't know exactly how much was changed, but any change throws doubt over the entire thing. Then there's the suspicious "genius move" which Deep Blue happened upon in one of the game. I, like Kasparov, don't buy it. Computer programs don't just happen into genius moves; either it would have been playing at that genius level all along, or not at all. Oh, yea, there's the fact that Deep Blue was fine tuned and prepared for Kasparov, but Kasparov was given no opportunity to prepare for Deep Blue.

Social Security is a pyramid scam.
[ Parent ]

Sorry man (5.00 / 1) (#96)
by tmenezes on Tue Feb 18, 2003 at 05:51:17 PM EST

You do know what you're talking about. Today was a bad day, I had to put up with some stupid people and was in a bad mood when comenting...

I don't quite agree with the "genious level" consistency idea. Just think of it, a genious move is one that nobody saw coming, yet it was prepared during several moves. How many "genious moves" could have failed because of changing in plans during game? The genious move is simply the moment when a brilliant plan succeds.

What I'm trying to say is that a bad algorithm performs consitently bad because of its short-sighteness. But as the algorithm goes deeper, so does the quality of game becames more difficult to measure at any given moment.

Even the most unsofistecated implementation of min-max can come to a brilliant change of plan in a key moment of the game given enough serach depth. Min-max search provides long term planing by making assumptions about the future moves of the oponent, but on the other hand reevaluates its plans on every move.

Regards.

[ Parent ]

Improved metrics (none / 0) (#116)
by epepke on Tue Feb 18, 2003 at 08:04:33 PM EST

Of course, alpha-beta pruning has been around for a long time. The problem with chess is calculating metrics for alpha-beta pruning. The phase space of the problem in chess is very non-linear; a weak position might be right next to a strong position in phase space, while there may be two strong positions that are fairly far apart. Evaluation and optimization algorithms tend not to work terribly well in this kind of space unlike, for example, traditional strategy-based war games, where they work pretty well.


The truth may be out there, but lies are inside your head.--Terry Pratchett


[ Parent ]
Two Things (5.00 / 8) (#55)
by DarkZero on Tue Feb 18, 2003 at 02:06:37 PM EST

First, there's one thing that I didn't really understand, so I'm not sure whether to criticize your article for it or not. You make opposite points in two different places, but you don't connect them.

Kasparov was not allowed to adequately prepare for his opponent, as IBM would not allow him access to the program. This unfair and not in line with normal chess procedures, where players can study their opponents.

-----

Finally, the program was not made to play chess in general, but was specifically built to play Kasparov. It was not a general chess-playing program. That's why IBM was hush-hush. They revealed nothing, and did not allow other programs or other world-class grand masters to play the program.

If you're pointing this out as an inequality, in that Kasparov wasn't allowed to change his style specifically to beat Deep Blue, but Deep Blue was allowed to change its style specifically to beat Kasparov, then I see your point. As two seperate points, though, it doesn't make sense. Are you arguing that Kasparov should have an advantage and Deep Blue should not?

Computer programs have a long way to go before they're anywhere near the "equals" of players, including Fritz and Junior. Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider) and restrict how many moves ahead they can look to how many a grand master can look forward, and you'll see the "best" computer chess programs reduced to playing chess at the level of a four-year old. An easy way to test this would be to try running Deep Fritz and Deep Junior on a 386 or 286. In other words, the methods that humans use to play chess are vastly superior to the algorithms computers use.

This seems to be one of the larger points in your argument, besides the fact that Singleton is ignorant, and it doesn't seem to make sense. You're basically arguing that if you strip out all of the unique advantages of a computer chess program, leaving it with only the unique disadvantages of a computer chess program (such as poor pattern recognition, which is absolute Hell to program), it will lose. Well, yeah. Of course it would. Their superior computational power is offset by their lack of the innate senses like pattern recognition that chess, a game created for human beings and only human beings, requires. Saying that Deep Blue would lose if you took away its superior computational power and that that would prove that humans are superior chess players is like saying that Kasparov would lose to a four-year-old if you beat his skull repeatedly with a sledgehammer and that that would prove that four-year-olds are superior chess players. The handicap that you suggest creates disadvantages that are unique to the player being handicapped, rather than creating an even playing field between the two players.

Your argument seems to be that the match wasn't fair because Kasparov, the best that humanity has to offer in the game of chess, could not beat Deep Junior, the best that machinery has to offer in the game of chess. I'm not sure about Kasparov's opinion, so I wouldn't call that a sore loser argument, but I'd certainly call it a sore fan argument. Kasparov had the unique advantages and disadvantages of a human being, Deep Junior had the advantages and disadvantages of a machine, and that was readily accepted by both players when the match started. If Kasparov accepted the rules of the game as they were laid out beforehand and no one broke them during the match, then he has no one to blame for his loss than himself. The match with Deep Blue may be an exception, but as far as I've seen, the match with Deep Junior was not. I have not heard of any tricks or curveballs like the machine being reprogrammed in the middle of a match taking place in the Deep Junior matches.

see... (none / 0) (#175)
by spottedkangaroo on Wed Feb 19, 2003 at 08:26:35 AM EST

Kasparov wasn't allowed to change his style specifically to beat Deep Blue, but Deep Blue was allowed to change its style specifically to beat Kasparov

That's the thing. Kasparov _can_ change his style, deep blue cannot.

It's Man vs. Man&Machine.

[ Parent ]

a very human-centric view of AI (4.90 / 11) (#56)
by Delirium on Tue Feb 18, 2003 at 02:15:24 PM EST

Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider) and restrict how many moves ahead they can look to how many a grand master can look forward, and you'll see the "best" computer chess programs reduced to playing chess at the level of a four-year old. An easy way to test this would be to try running Deep Fritz and Deep Junior on a 386 or 286. In other words, the methods that humans use to play chess are vastly superior to the algorithms computers use.
You seem to be arguing that proper AI is in some sense creating not only computers that make intelligent decisions, but computers that make intelligent decisions in the exact same way humans do. This seems unreasonable. Humans naturally excel at some things (pattern recognition in particular) and computers naturally excel at others (fast computation in particular). If a computer can overcome its weaknesses by playing on its strengths, then this is all that really matters, in my view. Your argument that a computer to be a "good chess player" needs to play chess the same way a human does would lead to the equally ridiculous claim that airplanes don't "really" fly because they don't flap their wings.

This is not to say that advanced pattern matching capacities rivalling those of humans aren't possible; in fact, I think they are very likely to be developed in the future. Then you'll get your computer that can play chess in a similar way to a human. But I think that's merely an interesting curiosity, not anything necessary for intelligence per se.

An interesting curiousity? (4.00 / 3) (#60)
by llimllib on Tue Feb 18, 2003 at 02:41:26 PM EST

Computers can beat very good humans at chess for one reason, and that is the fact that there is an effective board evaluation function. That is, given a board position X, the computer can rate it easily and accurately.

However, in many other games, such as the game of Go, this function does not exist. Hence, the computer needs pattern-matching ability and strategic thinking to succeed. If it were possible to make a computer play chess even remotely like the way a human does, or any other game for that matter, it would be far more than an "interesting curiosity", it would be a tremendous breakthrough in the history of computing.


Peace.
[ Parent ]
yeah, I understated its importance (4.00 / 1) (#66)
by Delirium on Tue Feb 18, 2003 at 03:14:10 PM EST

I meant more that in terms of playing chess well, it's not particularly important. If (as I consider generally reasonable) you define "intelligence" as simply "doing things well," or recursively as "doing the intelligent thing," then if a computer plays chess well, it is playing intelligently in that it picks good moves. How exactly it arrives at these moves isn't particularly important when evaluating whether it is acting intelligently or not. I do agree that it's important when applying the concepts to other domains though.

[ Parent ]
agreed. (none / 0) (#80)
by llimllib on Tue Feb 18, 2003 at 04:14:28 PM EST

I just had a problem with calling pattern-matching a curiousity.

Peace.
[ Parent ]
not true at all (4.50 / 2) (#72)
by jjayson on Tue Feb 18, 2003 at 03:28:28 PM EST

Play a computer chess program with no look-ahead (relying soly on its evaluation function) and you will dismantle it.  Being able to look 14 moves ahead with an club player level evaluation function, a GM/IM derived opening database, and a 6/7-piece perfect closing database are what makes a computer play chess well. Look-ahead makes up for a poor evaluation function and a good evaluation function obviates look-ahead.

Something like Fischer random chess would eliminate the opening book advantage and send the computer back to the club ranks. Playing chess on a 19x19 board would make the machine terrible, too.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

and you disagree with me where? (none / 0) (#82)
by llimllib on Tue Feb 18, 2003 at 04:21:21 PM EST

I never said that lookahead wasn't important, but without an evaluation function, lookahead doesn't work. Thus, the minimax algorithm is dependant on the evaluation function, which makes the evaluation function (IMHO) the more basic of the two, and thus the one that I mentioned.

As a side note, a computer playing 19x19 chess would be *far* better, again IMHO, than one playing 19x19 go, with current programs. I would love to do a comparison between the two - a good chess evaluation program set to 19x19 Fischer chess (are there any that would handle that? What would be the piece distribution?) vs GnuGo set to 19x19, to compare the state of Go pattern matching vs. the state of Chess evaluation functions. Methinks the chess program would whoop up on GnuGo.


Peace.
[ Parent ]
the eval function (none / 0) (#84)
by jjayson on Tue Feb 18, 2003 at 04:39:54 PM EST

these does not exist a "good" one. And minimax does not need an full evaluation function. When and end games state can be research, simple function that determines win, loss, or tie is all that is necessary.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
when an end games state can be research? (none / 0) (#99)
by speek on Tue Feb 18, 2003 at 06:15:14 PM EST

Beyond parsing problems with this phrase, what about the 90% of the cases where an end games state can't be "research"?

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Yes... (none / 0) (#113)
by jjayson on Tue Feb 18, 2003 at 07:32:19 PM EST

I'm sorry for not getting my tought to the screen very clearly. It seems like you still understood, hopefully.

Yes, when you cannot reach a terminal state, a utility function is applied to estimate the change of winning. In chess though, it isn't very good. The only reason a program plays well is that it has an acceptable evalutation function and it can look 14 moves ahead before applying it. It is effectively playing very poor chess, but playing it 14 moves into the future.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

oh (none / 0) (#117)
by speek on Tue Feb 18, 2003 at 08:26:03 PM EST

Oh, ok. It seemed to me you were saying the evaluation function wasn't a vital part of the program. It certainly helps a great deal play 14 moves in the future, but it all still rests on the quality of the eval function. If all it did is count up piece points, it would never beat an expert player, let alone Kasparov, so there is a fair amount of sophistication in today's programs.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

yeah, but it's still pretty crappy (none / 0) (#129)
by Delirium on Tue Feb 18, 2003 at 09:48:39 PM EST

A good evaluation function could simply look one move ahead, and win. However, without a one-move lookahead computer chess programs generally do pretty badly, because the evaluation function just isn't very good.

[ Parent ]
eval functions (none / 0) (#133)
by jjayson on Tue Feb 18, 2003 at 10:05:10 PM EST

They tend to just be linear sums.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Oh dear! (none / 0) (#169)
by nusuth on Wed Feb 19, 2003 at 06:23:46 AM EST

You must have missed last 40 years.

[ Parent ]
explain... (none / 0) (#190)
by jjayson on Wed Feb 19, 2003 at 12:13:42 PM EST

no top-level program I know of uses anything cute, like neural network or probability graphs.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
What do you mean by "linear sums"? (none / 0) (#198)
by nusuth on Wed Feb 19, 2003 at 01:18:45 PM EST

No modern program I know of uses weighted sums of simple measures like weighted piece count, number of controlled squares, possible moves after move etc. I wouldn't call those evaluation functions "just linear sums" if the components of that sum are not directly derivable from game board representation without analysis, even if the final step of analysis is summing those components together.

Also, when you said "eval functions" I didn't think you were referring to chess evaluation functions only. The top backgammon player uses a neural network for evaluation, neural net approaches to computer Go work as well as conventional ones, bridge programs use probability graphs etc.

[ Parent ]

linear sums (none / 0) (#207)
by jjayson on Wed Feb 19, 2003 at 02:17:33 PM EST

No modern program I know of uses weighted sums of simple measures like weighted piece count, number of controlled squares, possible moves after move etc. I wouldn't call those evaluation functions "just linear sums" if the components of that sum are not directly derivable from game board representation without analysis, even if the final step of analysis is summing those components together.
I still don't understand what you are saying. Most top-level playing programs do analysis, but they still combine the resuls of the analysis in a weighted linear fashion. Nothing too tricky here: U = w_1 x_1 + w_2 x_2 + ... + w_n + x_n.

Also, when you said "eval functions" I didn't think you were referring to chess evaluation functions only. The top backgammon player uses a neural network for evaluation, neural net approaches to computer Go work as well as conventional ones, bridge programs use probability graphs etc.
Yes, I was talking specifically about chess programs. Being a big backgammon player (much more than chess) who has won a few amateur tournaments, I have a great appreciation for TD-Gammon. I have often wondered by no top chess program makes extensive use temproal different training.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
OK (none / 0) (#220)
by nusuth on Wed Feb 19, 2003 at 04:47:56 PM EST

   I wouldn't call those evaluation functions
       "just linear sums" if the components of
       that sum are not directly derivable from
       game board representation without analysis,
       even if the final step of analysis is
       summing those components together.

I still don't understand what you are saying. Most top-level playing programs do analysis, but they still combine the resuls of the analysis in a weighted linear fashion. Nothing too tricky here: U = w_1 x_1 + w_2 x_2 + ... + w_n + x_n.

I was trying to say, as long as those "x_i"s are not just sums themselves or trivially emergent values from problem representation, the whole process should not be labeled as "just linear sums." That would be like saying "The brain is just a lump protein, water and fat." Sure it is, but that doesn't tell much about the brain, does it?


[ Parent ]

i think... (none / 0) (#102)
by llimllib on Tue Feb 18, 2003 at 06:31:13 PM EST

Quoting Norvig (AIMA, p.126):
The minimax algorithm assumes that the program has time to search all the way to the terminal states, which is usually not practical. Shannon's original paper on chess proposed that instead of going all the way to terminal states and using the utility function, the program should cut off the search earlier and apply a heuristic evaluation function to the leaves of the tree
emphasis his

You show me a grandmaster level, strictly minimax chess program, and I'll show you a computer a million times faster than any computer today.


Peace.
[ Parent ]
blah... I have that book (none / 0) (#111)
by jjayson on Tue Feb 18, 2003 at 07:29:07 PM EST

I worked with Peter Norvig, too. And, my first AI class was with Stuart Russell at Cal, and I was in one of his research groups. First, it isn't emphasis. It means that this is where the term is defined.

Now, I don't think I understand what you are trying to say. Can you clarify, please?

_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

ok, let's get this over with (none / 0) (#141)
by llimllib on Tue Feb 18, 2003 at 11:25:07 PM EST

alright, here's what I want to say, besides that I think that we're violently agreeing: minimax is not practical without an evaluation function. You are right, you could beat a grandmaster without an evaluation function on an infinite computer that could trace all branches of a tree. However, on practical, modern computing machines, the evaluation function is the reason that a chess program can play at a grandmaster level.

You first said that a program without lookahead, relying solely on evaluation, would be poor. That statement is impossible; perhaps you meant only looking ahead one move, because applying an evaluation function to the board that's already there doesn't get you anywhere. Lookahead, by which I assumed you meant the minimax algorithm, for all practical purposes, does not exist without an evaluation function. It can be done by simply applying a utility function to the end of every branch of the search tree, but this is incredibly impractical for the game of chess. Thus, I said that the basic reason that chess can be played at a high level is that it has a fairly efficient and effective board evaluation function, and I will stick to that statement.

Finally, I apologize that I sounded like a pretentious prick in the last post. The "emphasis his" statement was meant as a disclaimer from responsibility, but came out like I was trying to make that fact an important statement. I really just wanted to quote a source that was authoritative, and I perhaps picked a poor passage.


Peace.
[ Parent ]
we don't really agree, I think (none / 0) (#147)
by jjayson on Wed Feb 19, 2003 at 12:20:47 AM EST

the basic reason that chess can be played at a high level is that it has a fairly efficient and effective board evaluation function
That is not more true than the statement "the basic reason that chess can be played at a high level is that is has a fairly deep lookahead." It is the synergy between the two that allow it to play at a high level.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
An example: (5.00 / 1) (#150)
by llimllib on Wed Feb 19, 2003 at 12:45:25 AM EST

"the basic reason that chess can be played at a high level is that is has a fairly deep lookahead."

Perhaps I do not understand what you are trying to say. Let's take an example: Given the game of 'X', where we will assume that the board evaluation function is so poor as to render it useless, how could lookahead be practical? We will assume that X has a high enough branching factor in a time limit small enough so that it is impractical to search the entire tree of possibilities. How do we approach this game?

In this example, we have lookahead, but without a board evaluation function, I propose that it is useless. You rebutted, "well, without lookahead, an eval function is useless", and I grant you the point. But the fact is, there are many games that have lookahead, and yet it is only effective in the case that there is an efficient board evaluation algorithm.

All I said, in the original post, was that, without board evaluation, Chess could not be played at a grandmaster level by a practical modern computer. Without lookahead, it couldn't be either, but with lookahead and no eval function, it would not be practical.


Peace.
[ Parent ]
'good' evaluation function (none / 0) (#145)
by pb on Tue Feb 18, 2003 at 11:53:11 PM EST

I think the point here is that a simple evaluation function is a lot more effective in chess than it is in go, and this isn't all due to go's much larger 19x19 board; it's also due to fundamental differences between chess and go.  ("Chess is 99% tactics" -- Richard Teichmann)
---
"See what the drooling, ravening, flesh-eating hordes^W^W^W^WKuro5hin.org readers have to say."
-- pwhysall
[ Parent ]
You guys are forgetting an important thing: (none / 0) (#162)
by Estanislao Martínez on Wed Feb 19, 2003 at 04:41:08 AM EST

Humans make mistakes, get emotionally involved in the game, get tired, stressed, etc. Computers don't make mistakes in the same sense, have no emotional attachment, and could care less whether they lose or win.

A key factor: you can count on human players doing blunders. A computer will never make a blunder.

--em
[ Parent ]

well... (none / 0) (#178)
by llimllib on Wed Feb 19, 2003 at 09:47:42 AM EST

better to say a computer will never make a computational error. Even the best computer will make what a grandmaster considers to be a blunder every so often.


Peace.
[ Parent ]
Tactical blunders? (none / 0) (#215)
by Estanislao Martínez on Wed Feb 19, 2003 at 03:56:46 PM EST

Even the best computer will make what a grandmaster considers to be a blunder every so often.

Not tactical blunders, I imagine.

--em
[ Parent ]

the point is, chess software is "dumb" (2.00 / 6) (#62)
by dh003i on Tue Feb 18, 2003 at 02:51:14 PM EST

The point is that chess software is dumb. For the most part, it uses brute force, while pruning out moves it determines to be certainly inferior. Hardly an intelligent way to go about things. Btw, comparing a children's game -- Go -- to chess is laughable.

Social Security is a pyramid scam.
[ Parent ]

"dumb" (4.00 / 2) (#64)
by jjayson on Tue Feb 18, 2003 at 03:11:33 PM EST

Things are always "dumb" when they have been figured out. Once the methods are well understood, it removes from the the magic ("humanness") from them. NLP/NLT is also dumb by your definition, is just computers seme values and and uses some lookup tables.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
redefining intelligence (4.50 / 2) (#69)
by Delirium on Tue Feb 18, 2003 at 03:20:02 PM EST

Redefining intelligence as "whatever a machine can't do" actually seems to be pretty common. If you had asked someone 300 years ago whether doing complex arithmetic was evidence of intelligence, most people would've answered yes. Similarly, if a dog was found that could add, this would've been considered evidence of a very intelligent dog. But almost no one would claim that their calculator is intelligent.

[ Parent ]
its all relative. what is 'smart'? (5.00 / 1) (#65)
by Work on Tue Feb 18, 2003 at 03:12:12 PM EST

i think Del's point is that since the 2 chess playing ways - human pattern recognition, and computer computation - are so fundamentally different, as to not really be compared fairly.

Suppose you forced a human to base his moves only on a set computation algorithm rather than free thinking pattern recognition - then obviously, the computer would win.

The best way to beat a computer is to analyze its algorithm and attack its weak points. Similarly, the best way to beat a human is to analyze his playing style and attack his weakpoints. A computer however, does not have this psychological analytical ability to attack the other player. I do not think its unfair that kasparov wasn't allowed to analyze the algorithm. Are you really that much of a master if you cannot sit down with someone you've never analyzed before and not beat them soundly? Perhaps this is not the way of chess at its highest level, but neither is comparing an algorithm to human thinking.

[ Parent ]

WRONG (3.00 / 1) (#103)
by dh003i on Tue Feb 18, 2003 at 06:32:36 PM EST

Because Deep Blue was effectively built with analysis of Kasparov in mind. They built the computer specifically tailored to play against Kasparov's style. Btw, modern computer chess programs effectively have excellent analysis of all world chess champions and any grand masters of international significance.

Social Security is a pyramid scam.
[ Parent ]

not really. (3.00 / 1) (#105)
by Work on Tue Feb 18, 2003 at 06:43:33 PM EST

the psychology of a chess game is not determined solely prior to the game. It is also decided during the game as the players attempt to fake each other out. The computer is incapable of analyzing kasparov himself while in play.

Chess is more than a series of moves, I would think you would know this considering you wrote the article.

Further, kasparov could've done alot of research into beating the computer if he had studied computer chess in general better. There aren't that many ways for the machine to be programmed and a basic grounding in the theory behind it would go a long way.

[ Parent ]

wrong and irrelevant (2.00 / 1) (#107)
by speek on Tue Feb 18, 2003 at 07:01:03 PM EST

I'm not sure how you responded to his point, but, in any case, you're just plain wrong to say:

There aren't that many ways for the machine to be programmed and a basic grounding in the theory behind it would go a long way

There's virtually an infinite number of ways to program the machine to play chess.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Yes and no. (none / 0) (#125)
by Work on Tue Feb 18, 2003 at 09:16:37 PM EST

Theoretically, of course. Theres an infinite number of ways to program a computer to do an infinite number of jobs.

From a practical and real world standpoint though, this isn't true. There are tried and true methods for writing chess programs. Lots of study has been given into the fastest way of doing it, and this is the method employed by each, albeit with a few changes here and there. Someone else pointed out how what advancements in chess program writing have been largely incremental since the primary algorithm was written.

[ Parent ]

eval function (none / 0) (#127)
by speek on Tue Feb 18, 2003 at 09:44:02 PM EST

The eval function can be written in any number of ways is all I was saying. Different eval functions will result in different playing. That's why there's Fritz and Deep Fritz and Deep Junior and Deep Blue and Sargon I-II-III, ChessMaster, etc. You can say their incremental, but that's the way improvement works.

Personally, I'd be interested in exactly how the IBM programmers tweaked their program to play Kasparov.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

From what I remember... (none / 0) (#132)
by jjayson on Tue Feb 18, 2003 at 10:03:47 PM EST

they trained the evaluation function against Kasparov losses. I don't know if this was an automated training or hand adjustment.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
that seems risky (none / 0) (#137)
by speek on Tue Feb 18, 2003 at 10:33:24 PM EST

I'm mean really risky. Seems like that could've backfired on them.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

So? (5.00 / 1) (#122)
by DarkZero on Tue Feb 18, 2003 at 08:38:57 PM EST

Because Deep Blue was effectively built with analysis of Kasparov in mind. They built the computer specifically tailored to play against Kasparov's style.

I don't really see why that matters. When Kasparov plays someone that has never played professionally before, the amateur probably knows absolutely everything about him (what with his fame and all), but Kasparov knows nothing about the amateur. Does that mean that the game is unfair? Should it be voided because of that? Should the amateur be mocked as some sort of cheater, as you seem to have done in this article?

[ Parent ]

depends on what you mean by intelligent (4.00 / 1) (#68)
by Delirium on Tue Feb 18, 2003 at 03:15:34 PM EST

It picks good moves, which I'd regard as "intelligently" choosing what to do. Whether it does this in the way humans do it (essentially very good pattern recognition) or with better search depth doesn't seem to make much intrinsic difference in terms of whether its behavior is intelligent. Sure, brute force is just mathematics and computation, but so is pattern recognition.

[ Parent ]
You're not addressing the tough questions. (none / 0) (#163)
by Estanislao Martínez on Wed Feb 19, 2003 at 04:43:16 AM EST

What is "pattern recognition"? Aren't Turing machines pattern recognizers? I mean, they recognize patterns in their tapes.

And don't get me started on "intelligence" (which, if you've failed to notice, does not exist).

--em
[ Parent ]

yeah (none / 0) (#164)
by Delirium on Wed Feb 19, 2003 at 05:00:23 AM EST

I didn't claim that pattern recognition was a sort of computation greater than that a Turing machine is capable of, just that it's something different than brute force tree-searching. I do think computers are in principle capable of it, and of anything else you might choose to define as "intelligence."

[ Parent ]
Rant. (5.00 / 2) (#219)
by Estanislao Martínez on Wed Feb 19, 2003 at 04:22:38 PM EST

I didn't claim that pattern recognition was a sort of computation greater than that a Turing machine is capable of, just that it's something different than brute force tree-searching.

You still didn't get my (somewhat hidden) point, which has to do with challenging the underlying psychological assumptions. Why do you think there is such a thing as "pattern recognition" anyway, and that humans do it?

I do think computers are in principle capable of it, and of anything else you might choose to define as "intelligence."

So do you think there is any convincing constraints on what one may choose as a definition of "intelligence"? I.e. what makes a concept you define be "intelligence" and not "something else, which I have the presumption to name by the same word"?

I am trying to make a subtle point in a very oblique manner here: being able to perform well at a set of individual problem domains (e.g. face recognition, recognizing the same melody in different keys, etc.) doesn't license the conclusion that there is some object (e.g. a "faculty") called "intelligence", that is responsible for all these performances, and that is "coherent" in some sense beyond merely being the collection of mechanisms that actually does those tasks.

I'm pretty much trying to make you give up the prejudice that people are "intelligent"; psychologists never discovered that people are intelligent, they took for granted that they are, and that we agree what this claim means. The point then is that good theories of what people do may well not have the word "intelligence" anywhere in them.

Same comments go for "pattern recognition," and the point I made obliquely by calling a Turing machine a pattern recognizer: does the term really play an important role in theories of psychology? In other terms, would we really lose anything if we dropped the term "pattern recognition" from studies of how humans do certain tasks? Is the notion "pattern recognition" doing any actual work in the theories, or is all the work being doing by particular mechanisms that we label so? And is there are necessary and sufficient conditions for deciding whether something (e.g. a Turing machine, a neural network) is a "pattern recognizer"?

Gah, I'm ranting already. My point is that your the whole issue about whether one should regard a chess computer as "intelligent" or not has a fundamental problem, in that there is no psychological reason why we should think that "people are intelligent" in the first place.

(There are other reasons for regarding a person to be intelligent, but which neither psychologists nor AI people take seriously, which is quite dumb: e.g. one may say that being intelligent is to be legally responsible for your actions, and thus subject e.g. to be tried in court. Historically, the notion of intelligence is very much tied up with responsibility; cognitive psychologists and by extension AI people arbitrarily divorce the historical notion of intelligence from its social context, choosing something like "abstract problem solving ability.")

Gah, I should shut up.

--em
[ Parent ]

"Gah, I should shut up." ??? (none / 0) (#238)
by gzt on Thu Feb 20, 2003 at 12:26:43 AM EST

I don't see why. Letting out your top-secret troll knowledge? I'd keep going, what you're saying is far more informative and insightful than any, say, stories on this site.

[ Parent ]
Go is a child's game? (3.00 / 1) (#81)
by steveftoth on Tue Feb 18, 2003 at 04:17:07 PM EST

You're just trolling right?  You don't actually think that Go is for children?

[ Parent ]
If anyone actually believed that, (none / 0) (#83)
by llimllib on Tue Feb 18, 2003 at 04:28:54 PM EST

If you want information about Go, to see that it's not a child's game, you should start with the brilliant K5 article about Go. After that, I can recommend GoBase, and a whole bunch more sites if you really have questions.


Peace.
[ Parent ]
Go is for childern too (5.00 / 1) (#160)
by niklaus on Wed Feb 19, 2003 at 03:13:16 AM EST

I taught Go to my little sister when she was 6 years old. She understood the rules instantly and was playing better in her first few games than most adults beginners I have taught. Unfortunately she doesn't have enough stamina to play on any board bigger than 9x9 and tends to get frustrated by playing me ("it's unfair, you read all these Go books and I didn't"), and as I moved out of the house she has no one else to play with (my father, who is a decent chess player, got too frustrated after losing about 5 games).

Pretty much all of the current pro players have begun playing seriously as children, the youngest becoming pro at age 11. For example there's Liao Xingwen in China who at age 5 convinced his parents to let him move away from home to a couple of pro players (Liu Yajie 2d and Wang Hongjun 7d). He's now 8 years old and playing at top amateur level.

But of course that doesn't mean that Go is a children's game only. It is by far the most fascinating game I have ever played and I'm willing to bet that mankind won't come up with a more beautiful game before it will destroy itself :). The amazing complexity arising from very few simple rules makes any other game seem like an arbirtrary hack in comparison.

[ Parent ]

You got to be kidding (1.00 / 1) (#170)
by henrik on Wed Feb 19, 2003 at 06:23:51 AM EST

Btw, comparing a children's game -- Go -- to chess is laughable

No, what's laughable is the statement in italics above this one. Go makes chess look like a crude and limited game which involves no thought at all.

Akademiska Intresseklubben antecknar!
[ Parent ]

I should talk about chess (3.00 / 5) (#74)
by Orion Blastar on Tue Feb 18, 2003 at 03:47:32 PM EST

I couldn't even beat my Atari 2600 Chess game. It kept blanking the screen and flashing colors and moving more than one piece at a time and sometimes it moved my pieces! Cheating b*tch! I have to enable "Edit Mode" to move the other piece back after it cheated.

Deep Blue and Deep Blue Junior are cheats, its a fake, its all a fake!

Ooops, time for me to take my medicine before the people in white coats come to take me away again.

Cheating b*tch! :)
*** Anonymized by intolerant editors at K5 and also IWETHEY who are biased against the mentally ill ***

Atari 2600 Chess (3.00 / 2) (#104)
by Andy Kaufman on Tue Feb 18, 2003 at 06:38:08 PM EST

Actually I was smart enough to beat it on the hardest level.

[ Parent ]
See, the computer does not impress me.... (1.66 / 3) (#75)
by FuriousXGeorge on Tue Feb 18, 2003 at 03:54:18 PM EST

Because the computer isn't doing any actual thinking.  It just does what it is programmed to do.  A group of humans did the actual thinking before the match started.

It would be the same thing if these people wrote down what a human should do against kasparov in a given situation and gave him the notes.

--

-- FIELDISM NOW!

I don't know what impresses you (none / 0) (#79)
by the on Tue Feb 18, 2003 at 04:14:16 PM EST

It would be the same thing if these people wrote down what a human should do against kasparov in a given situation and gave him the notes.
Hello?! How the hell could someone do this? It would probably require a stack of paper reaching to the moon and back several times over. That is precisely the reason why the computer is so impressive. It fits in a small box.

--
The Definite Article
[ Parent ]
It would require lots of people (none / 0) (#93)
by FuriousXGeorge on Tue Feb 18, 2003 at 05:32:00 PM EST

Thats my point.  The computer is just a tool to do it faster.  The COMPUTER isn't playing chess with Kasparov, the programmers are.

This isn't, IMO, any signifigant type of AI.  

--
-- FIELDISM NOW!
[ Parent ]

The pyramids... (none / 0) (#108)
by the on Tue Feb 18, 2003 at 07:01:08 PM EST

...who cares about them. They're just a pile of bricks, only bigger than average.

--
The Definite Article
[ Parent ]
Sounds Like (none / 0) (#126)
by djotto on Tue Feb 18, 2003 at 09:39:20 PM EST

A Conversation With Albert Einstein. (A response to the Turing test).

Imagine a book a mile thick. It consists of billions of tables. You can ask the book a question, any question. A bunch of clerks do look-ups on the tables, and construct an answer. That answer is exactly the answer Albert Einstein would have given you.

Is the book conscious? Most people would say no, it's just big a database.

Now, imagine that every time the book is asked a question, those clerks update the book's tables in response to the question. Now, is the book conscious? It would certainly pass the Turing test.

(I suspect the book is conscious... or is so near it makes no difference. But then I lean towards Strong AI.)



[ Parent ]
I'd argue that you're fixated on the book.. (none / 0) (#182)
by mordant1123 on Wed Feb 19, 2003 at 11:22:18 AM EST

I'd say that it's not just the data-transforms described by the book, but also the rules involved in updating the book (by the clerks).. So perhaps the clerks+the book render a concious entity.. (for various values of 'clerks')

(I think the 'clerks' would need to be updating with more than the questions/responses, otherwise the answers they'd likely relay would be along the lines of 'Where am I? Why can't I see anything?')


-----
"There is no intellient opposition to white nationalism." - Johnny Walker
[ Parent ]

Re (none / 0) (#194)
by djotto on Wed Feb 19, 2003 at 12:59:04 PM EST

I believe the analogy of a book is used because people have more resistance to the idea of a "thinking book" than a "thinking computer". Yes, I agree that the algorithm's the thing under discussion.

[ Parent ]
Hello, (none / 0) (#87)
by it certainly is on Tue Feb 18, 2003 at 05:00:53 PM EST

have you heard of the Chinese Box Argument. It's the same thing as you complain about. Have a read.

kur0shin.org -- it certainly is

Godwin's law [...] is impossible to violate except with an infinitely long thread that doesn't mention nazis.
[ Parent ]

HI! (none / 0) (#91)
by FuriousXGeorge on Tue Feb 18, 2003 at 05:28:33 PM EST

"According to John Searle, proponents of strong AI (Artificial Intelligence) believe that it is possible that programmed computers possess the same kind of understanding which human beings possess. Suppose that we give a computer a story and ask it questions about the story. If the answers it gives are indistinguishable from the answers a normal, typical human being who understands the story gives, then according to Searle, the proponents of strong AI will claim that the computer possesses a literal understanding of the story, 'solely in virtue of being a computer with a right sort of program' (380)."

So, If i train an ape to give the correct answer to my questions, it's as good as being conscious?

--
-- FIELDISM NOW!
[ Parent ]

No, (none / 0) (#92)
by it certainly is on Tue Feb 18, 2003 at 05:31:04 PM EST

that's weak AI. Strong AI is a computer programmed to "understand" rather than weak AI, a computer programmed to mimic humanity.

kur0shin.org -- it certainly is

Godwin's law [...] is impossible to violate except with an infinitely long thread that doesn't mention nazis.
[ Parent ]

how do u program... (3.00 / 2) (#94)
by FuriousXGeorge on Tue Feb 18, 2003 at 05:38:26 PM EST

understanding?   And how do u know it understands rather then mimics?

--
-- FIELDISM NOW!
[ Parent ]

and now you understand (none / 0) (#100)
by llimllib on Tue Feb 18, 2003 at 06:22:00 PM EST

the basic question of AI. Figure that one out, and you'll be famous.

Peace.
[ Parent ]
Most efforts so far (3.00 / 2) (#106)
by it certainly is on Tue Feb 18, 2003 at 06:51:58 PM EST

have hinged on building a database of facts, the relationships between those facts, rules that can be applied to those facts, then applying "inference" algorithms to determine things which are not known facts. Then there's probability-based facts/relationships/rules/infererence, then there's bayesian belief networks, then there's neural nets.

If I'd wanted to know more, I wouldn't have left university after they gave me my degree.

kur0shin.org -- it certainly is

Godwin's law [...] is impossible to violate except with an infinitely long thread that doesn't mention nazis.
[ Parent ]

Exactly the problem (none / 0) (#119)
by DarkZero on Tue Feb 18, 2003 at 08:31:35 PM EST

Is a highly advanced AI just following instructions or not? Or are both it and we following instructions and we just think otherwise because we understand the AI's "thought process" better than we understand our own?

These are the basic questions of AI. They probably won't be solved until we understand the human brain as well as we understand a computer program that we've written ourselves. They're fun to wrap your brain around in the mean time, though.

[ Parent ]

That sounds like the Turing test (none / 0) (#112)
by damiam on Tue Feb 18, 2003 at 07:31:43 PM EST

Just slightly rephrased.

[ Parent ]
You don't impress me. (none / 0) (#180)
by RofGilead on Wed Feb 19, 2003 at 10:16:28 AM EST

You are just what your DNA programmed you to be, and fed the training data of your life experiences. You didn't do any thinking, it was whoever made your DNA and life experiences.

-= RofGilead =-

---
Remember, you're unique, just like everyone else. -BlueOregon
[ Parent ]
computer intelligence and fischer random chess (4.33 / 3) (#98)
by speek on Tue Feb 18, 2003 at 06:12:53 PM EST

Humans don't impress me because they don't do any actual thinking. They just have neurons that fire or don't fire, and all they do is remember patterns and do what's been taught to them to do when they encounter those patterns. There is no way a human could calculate a full 2-ply tree in a chess game without enormous amounts of time and probably extra boards to serve as extra memory. Pathetic. -- Deep Blue, just prior to being dismantled

And, computers would do better in fischer random chess, not worse, because you take away the main human advantage - the huge store of positions that are familiar to them. Force a human to play in unfamilar positions, and he/she will have to work harder, whereas the computer just does what it does from the start of the game.

--
al queda is kicking themsleves for not knowing about the levees

emperically denied (none / 0) (#110)
by jjayson on Tue Feb 18, 2003 at 07:21:45 PM EST

A common strategy against computers to take take them out of their opening book early. The further you are from the end of the game, the worse the computer plays. When Junior was pounded by Kramnik in only 33 moves it was because of an opening odity that Junior didn't have in his opening database. The human was of playing chess allows for this inventiveness, but the computer way of playing the opening and middle game means that it doesn't porform well if not given the little push from the previous IM/GM games and analysis.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
empirically not tested yet (none / 0) (#118)
by speek on Tue Feb 18, 2003 at 08:29:01 PM EST

So, you give an example of a human playing positions he's familiar with while the computer plays without any opening book. Well, that's halfway there. Nor does one game prove much.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

no way... (none / 0) (#131)
by jjayson on Tue Feb 18, 2003 at 10:02:12 PM EST

The Stonewall deviation wasn't familiar to Kramnic at all. He intentionally tried to pull Junior into unfamiliar ground. It is well known that without an opening book, computers would be lost. They just don't play that well that far from the endgame and that close up.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
there's no empirical data (none / 0) (#136)
by speek on Tue Feb 18, 2003 at 10:31:31 PM EST

I'd like to see it put to the test. You may say Kramnik wasn't familiar with the opening, but he was familiar enough to steer the game that way. I will not believe your assertion till I see the proof in fischer random chess, not just humans leading computers out of their opening book.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

how about logic then... (none / 0) (#139)
by jjayson on Tue Feb 18, 2003 at 10:57:35 PM EST

(1) The further the computer is away from the end of the game, the worse it evaluates a position.

(2) The more closed a position is the worse the program plays.

In the opening both points apply.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

balanced by: (none / 0) (#140)
by speek on Tue Feb 18, 2003 at 11:03:59 PM EST

(1) The less familiar the human is with the position, the worse it plays

Yes, that's the tradeoff. I think the human is hurt more by this than the computer. I'm willing to be disproved by data, but not yet by this line of reasoning.

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

That is not true, though. (none / 0) (#146)
by jjayson on Wed Feb 19, 2003 at 12:10:43 AM EST

A human in an unfamial closed position is still better than a computer in a closed position. There are some GMs, like Karpov and Kramnic, who are computer killers because they know and exploit this.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Evaluation (5.00 / 1) (#153)
by Greyjack on Wed Feb 19, 2003 at 01:13:43 AM EST

That's why ultimately, it all boils down to the evaluation function. When computers get to the point that they can do general evaluation even remotely near as well as humans, the relentless exponential gains in speed will render them invincible soon after.

This applies equally well to Go, Connect 4, and Global Thermonuclear War.

That said, there's a fair bit of work to go before they'll be able to do evaluation as well as us. Unless Kurzweil's right, of course, in which case we only have a couple decades to go.

--
Here is my philosophy: Everything changes (the word "everything" has just changed as the word "change" has: it now means "no change") --Ron Padgett


[ Parent ]
Connect-4 (5.00 / 1) (#188)
by jjayson on Wed Feb 19, 2003 at 12:10:12 PM EST

is mathematically solvable, and Kurzweil is an idiot.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Y'know... (none / 0) (#239)
by Greyjack on Thu Feb 20, 2003 at 12:37:36 AM EST

...the thing I respect about you, jjayson, is your earnestness. Of course Connect-4 is solvable (duh). As for Kurzweil, while he's made some out there predictions recently, I'd hesitate to call him an idiot.

Out of curiosity, is Wolfram an idiot too?

--
Here is my philosophy: Everything changes (the word "everything" has just changed as the word "change" has: it now means "no change") --Ron Padgett


[ Parent ]
The Singularity (none / 0) (#240)
by jjayson on Thu Feb 20, 2003 at 01:47:19 AM EST

Connect-4 is solvable without lookahead, like Nim, unlike chess.

Kurzweil is an idiot for believing in this mythical singularity. The idea of humanity without a soul. I thought the Chinese Room settled all of that years ago.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

Just to be an ass . . . (none / 0) (#254)
by Greyjack on Thu Feb 20, 2003 at 04:12:19 PM EST

Connect-4 is solvable without lookahead, like Nim, unlike chess.

Just to be an ass, I'll say that's an example of a particularly good evaluation function :)

--
Here is my philosophy: Everything changes (the word "everything" has just changed as the word "change" has: it now means "no change") --Ron Padgett


[ Parent ]
what I could find (5.00 / 1) (#197)
by speek on Wed Feb 19, 2003 at 01:17:29 PM EST

I'm not sure the beginning position is "closed" entirely. There's plenty of opportunity for the position to evolve either way.

So anyway, I've searched the web some, and found this (literally the only computer vs. human result I could find).: http://www.rochadekuppenheim.de/heco/ar0006.html

You have to hunt down the page a little to find the result - Yusupov lost 2-0 to the computer.

There is also the interesting matter of the bizarre blitz play of the mysterious internet player that people speculate is either Fischer or a computer. This player beats GM's by playing the wackiest openings (moving their king to the third rank in the first 5 moves, for example, is quite common).

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

creativity and Fischer (none / 0) (#206)
by jjayson on Wed Feb 19, 2003 at 02:07:17 PM EST

The comment goes on to say that the reason Fischer random would be played to to remove the computer element and bring back creativity in the opening.

I have read all the theories about Bobby Fischer playing blitz under a pseudonym on some chess servers. I think there is a create chance that it is him. When he has played strong players in person (under his real name or physically in person) in the last few year, Fischer has been doing some of the same wacky openings. Fischers is a total ass and racist, but I think it is because he is also clearly schizoprenic or manic. It is a sad think that happened to him, but he is well considered to be by far the strongest player to ever touch the game by many. Personally, I think he is stronger than Karsparov, but due to his age and lack of opening study, I think that his game now lies on random chess and blitz.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]

the comment about computers (5.00 / 1) (#208)
by speek on Wed Feb 19, 2003 at 02:27:37 PM EST

is misguided. Computers aren't going away, random chess or no :-)

--
al queda is kicking themsleves for not knowing about the levees
[ Parent ]

Fischer stronger than Kasparov? (none / 0) (#210)
by Magneto on Wed Feb 19, 2003 at 03:30:00 PM EST

We are in the realm of pure speculation here. Fischer was a chess genius - in the 70s. His 1992 match against Spassky showed that his skills had not endured. Kasparov is an all-rounder who hasn't really got any weaknesses in his play. He has obviously faced much stronger opposition than Fischer ever did, and still dominated totally. But it is possible that Fischer would have been able to take his play into an even higher level if he had faced Kasparov (or Karpov). As for myself, I think Paul Morphy was the greatest chess genius of all time.

[ Parent ]
I disagree (none / 0) (#166)
by nusuth on Wed Feb 19, 2003 at 05:28:30 AM EST

Success in unfamiliar game positions rich in strategical possibilities such as openings in chess are more easily tackled with higher-level reasoning ("if I play this, this position will be analogous to some position X I know with differences p,r,s") than fast tactical analysis. Humans do the former better while computers can do the latter better. I agree that it cannot be settled without experimental data but if I were to bet on one side before the experiment, I would bet on the human provided they were playing more or less the same level before random variant is introduced.

[ Parent ]
Very simple really... (none / 0) (#183)
by jforan on Wed Feb 19, 2003 at 11:49:21 AM EST

Nearly every opening book move that is generally accepted/known by humankind is inserted into the computer.  Also, each beginning move is generally analyzed further so that it knows if it is in a "beneficial" position.  Also, if play deviates from the opening moves, the computer may also try to "get back to one of the opening moves" it has remembered.

Jeff

I hops to be barley workin'.
[ Parent ]

Au contraire ! (5.00 / 1) (#244)
by sailesh on Thu Feb 20, 2003 at 06:50:01 AM EST

I've actually played against DeepBlue. This was in November 1996 in CASCON - an IBM conference in Toronto. The very same software but a stripped down RS/6000. Of course I lost easily. However, this was after the first match in Philly where DeepBlue lost to Kasparov. In that series, DB shocked the world by beating Kasparov in the first Game. Apparently, Kasparov spent that entire night brooding - walking the streets of Philadelphia. Anyway, the same source, one of the Deep Blue researchers told me that in fact the DeepBlue team had screwed up and had _not_ taken their opening book with them for game 1 !! Inspite of this, DB actually won. Meanwhile it was being ftp'd over on a slow connection. They couldn't really explain why this was so ..

[ Parent ]
interesting story... (none / 0) (#249)
by jjayson on Thu Feb 20, 2003 at 12:07:36 PM EST

even though it goes against everthing I have been learned. thanks.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Bobby Fischer (4.00 / 4) (#109)
by ComradeFork on Tue Feb 18, 2003 at 07:15:38 PM EST

Actually I don't consider Bobby Fischer to be at the same level as Kasparov.

On another topic, I wonder how long it will be before the whole Computer vs Human thing hits Go. My guess is a long, long time.

Reckless prediction (3.00 / 1) (#151)
by Greyjack on Wed Feb 19, 2003 at 01:05:52 AM EST

There will be a MAJOR breakthrough in computer Go within the next 10 to 20 years, tops, thanks to some sort of stunning advance in pattern recognition, neural networks, something along those lines. Unless of course the quantum computing guys come through, at which point they'll just go ahead and solve Go via brute force after all.

(black wins, btw)

--
Here is my philosophy: Everything changes (the word "everything" has just changed as the word "change" has: it now means "no change") --Ron Padgett


[ Parent ]
Black wins? (none / 0) (#158)
by ComradeFork on Wed Feb 19, 2003 at 02:39:22 AM EST

Why does Black win?

With perfect play, the winner would depend entirely under what ruleset the game is being played under. Under current Japanese rules White gets an additional 6.5 points for being white, however other rulesets give different komi's.

[ Parent ]

Well... (4.50 / 2) (#181)
by Greyjack on Wed Feb 19, 2003 at 11:08:31 AM EST

...I asked Marvin, my friend from the future, who wins at Go once they solved it, and he said "black."  We didn't discuss it any further, since there were some very important lagers to attend to.

--
Here is my philosophy: Everything changes (the word "everything" has just changed as the word "change" has: it now means "no change") --Ron Padgett


[ Parent ]
Marvin (none / 0) (#221)
by ComradeFork on Wed Feb 19, 2003 at 05:05:50 PM EST

Has Marvin stopped beating his wife, in the future? :P

[ Parent ]
Vituperative rubbish, but interesting (4.85 / 7) (#120)
by epepke on Tue Feb 18, 2003 at 08:35:14 PM EST

This is a tiny slice of an interesting phenomenon: people's terror of machines. From the legend of John Henry to Elmer Rice's 1923 The Adding Machine, to 2001 and Colossus, the Forbin Project and Demon Seed and Alien, there's plenty of evidence of a deep insecurity about machines. Even Donald Norman, who should be immune to this kind of circuit envy, felt compelled to write a book called Things that Make Us Smart.

The match of Kasparov v. Deep Blue was anything but fair. Kasparov was not allowed to adequately prepare for his opponent, as IBM would not allow him access to the program.

Oh, rubbish. If Kasparov didn't think that was fair, he could simply have not engaged in the exercise in the first place. All the other points are possibly interesting. Maybe Deep Blue was a Kasparov-beating machine, but at least at first, Tiger Woods was a masters-beating machine. But what's "fair" is based on what free agents agree to do, not with what you can come up with afterward.

Actually, most people don't remember exactly how Deep Blue beat Kasparov. It wasn't necessarily by doing chess better. Kasparov stormed out, saying that the moves of the machine could not have been made by a machine. In other words, Kasparov percieved his brain to have been fucked with. Now, human chess players try to fake out, intimidate, and fuck with their opponents' brain all the time, and turnabout is fair play.

Furthermore, Deep Blue was altered in the middle of the match. This is hardly fair.

Again, rubbish, unless there's a rule that says that human chess players can't talk to anybody between games.

Reduce Fritz or Junior to considering 2 moves per second (which is what a chess-master can consider) and restrict how many moves ahead they can look to how many a grand master can look forward, and you'll see the "best" computer chess programs reduced to playing chess at the level of a four-year old.

Pheeuw! You might as well say, hook Kasparov up to a +5/+12/-12 supply and deny him food and drink for a couple of weeks and see how well he plays chess.

In other words, the methods that humans use to play chess are vastly superior to the algorithms computers use. An easy way to test this would be to try running Deep Fritz and Deep Junior on a 386 or 286.

Human electron-envy aside, the truth is that the means by which computers play chess are different from the means by which humans play chess. Anybody who is just now figuring this out is a bit on the dull side. The brain is an ultra-massively parallel device using slow elements; most computers are single or low-order parallel devices using fast elements. A proper comparison would not be with a 386 or 286 but rather with about a billion 8008's, each with with connections to several thousand 8008's.

What is declared "superior" is just a matter of a pissing contest.

As for Kasparov himself, yes, he's a huge self-promoter. Yes, he has an ego the size of Mars. So what?

So he's not much of a target for pity. He knew what he was getting into, and he did it, and he got career goodies as a result. He's a grown man and doesn't need to be protected from Big Bad IBM.


The truth may be out there, but lies are inside your head.--Terry Pratchett


Hmmm... (5.00 / 1) (#123)
by Lord Snott on Tue Feb 18, 2003 at 08:53:15 PM EST


You've got a lot of good points, but some seem rather moot.

Oh, rubbish. If Kasparov didn't think that was fair, he could simply have not engaged in the exercise in the first place.

Don't forget that ego, he probably thought he could beat the machine anyway, and embarrass IBM. But you're right, he's not much of a target for pity.

But you seem to have missed the most important point - a champion can beat all the other players. Deep Blue was designed to beat Kasparov, and Kasparov only. We'll never know if someone else could have beaten Deep Blue, IBM dismantled the evidence.


~~~~~~~~~~~~~~~~~~~~~~~~
This sig in violation of U.S. trademark
registration number 2,347,676.
Bummer :-(

[ Parent ]
That's fine (5.00 / 1) (#128)
by epepke on Tue Feb 18, 2003 at 09:44:04 PM EST

Don't forget that ego, he probably thought he could beat the machine anyway, and embarrass IBM. But you're right, he's not much of a target for pity.

Sure, and that's his point.

But you seem to have missed the most important point - a champion can beat all the other players. Deep Blue was designed to beat Kasparov, and Kasparov only.

Not only did I not miss it, I addressed it, with the Tiger Woods comparison. Tiger Woods focused exclusively on how to beat the Masters course for a very long time. He had some setbacks when he started to play other courses, but nonetheless, he has turned into quite an excellent golfer, quite possibly the best golfer of all time.

Computers don't work quite like humans; so the passing on of information from one iteration to another occupies the host brains of programmers. But surely a lot was learned in the construction of a Kasparov-beating algorithm that could be of use in a general chess-playing algorithm.


The truth may be out there, but lies are inside your head.--Terry Pratchett


[ Parent ]
IBM (none / 0) (#156)
by Spendocrat on Wed Feb 19, 2003 at 02:28:04 AM EST

But surely a lot was learned in the construction of a Kasparov-beating algorithm that could be of use in a general chess-playing algorithm.

Then why isn't IBM doing it? They love prestige projects like that.

[ Parent ]

There is little prestige left in computer chess (none / 0) (#165)
by nusuth on Wed Feb 19, 2003 at 05:20:31 AM EST

It is mostly solved now. Top of the line programs have been beating club level players for quite some time and the latest outlook is they can also draw grandmasters, at least some of the time. Imagine how much one can brag about writing a program that always beats everyone, not much. This is already happened for some games like checkers and it doesn't bring extra prestige to the programmer.

One can even talk about negative prestige. People tend to think if a game is better played by computers, it is not as fun anymore. (This is anecdotal, I have nothing to back it up.) That may backfire. If I were IBM stunts director, I would focus on Bridge or Go.

[ Parent ]

Some prestige (none / 0) (#196)
by Spendocrat on Wed Feb 19, 2003 at 01:02:58 PM EST

Check out Chinook.

[ Parent ]
IBM is weird (none / 0) (#214)
by epepke on Wed Feb 19, 2003 at 03:47:19 PM EST

They do a lot of incredibly innovative stuff, demo, and then completely forget about it. It is by no means limited to chess.


The truth may be out there, but lies are inside your head.--Terry Pratchett


[ Parent ]
Love this part (3.00 / 1) (#130)
by Hired Goons on Tue Feb 18, 2003 at 09:59:05 PM EST

Outside of Bobby Fischer, there isn't a chess player in the world who you could compare to Kasparov *cough* Kramnik *cough*
You calling that feature a bug? THWAK
LOL (none / 0) (#138)
by dh003i on Tue Feb 18, 2003 at 10:41:54 PM EST

Yea, sure.  Kramnick may have won the world chess champion title from Kasparov.  But Kasparov is still the highest ranked player, and in a recent blitz tournament, he was the only undefeated player, beating Kramnick head on.

When Kramnick dominates the chess world for nearly two decades, then we can think of him as being on the same level as Kasparov and Fischer.  Fischer won two back to back championships in two years, then retired into obscurity, but he's still the only guy who is on the same level as Kasparov (in fact, Fischer was named the chess player of the century, something which clearly angered Kasparov).

Social Security is a pyramid scam.
[ Parent ]

whatever. (none / 0) (#224)
by Hired Goons on Wed Feb 19, 2003 at 05:19:30 PM EST

Maybe Kasparov shod stop playing around with computers and win back his title, then.
You calling that feature a bug? THWAK
[ Parent ]
Ahh, Chess (1.33 / 3) (#134)
by Psycho Les on Tue Feb 18, 2003 at 10:17:08 PM EST

Complex without being profound as the wise man said.  A game designed for nerds wanting to escape the shallowness of their souls.

Profound games? (4.00 / 1) (#155)
by Spendocrat on Wed Feb 19, 2003 at 02:20:58 AM EST

Name some.

And if you say "go", after disqualifying chess, I'll spend exactly three seconds visualizing punching you in the head.

[ Parent ]

kasparov is god (3.00 / 2) (#135)
by turmeric on Tue Feb 18, 2003 at 10:25:55 PM EST

the techno dweebs must be eradicated. a return to science in this endeavor is well welcomed: except by elite bastards who think linux-like, that 'beating' is all that matters.

Complexity of Go v. chess (3.50 / 2) (#148)
by dh003i on Wed Feb 19, 2003 at 12:22:47 AM EST

(1) Go's increased complexity over chess is almsot solely a function of the size of the board.  An 8x8 Go game is of approximately the same complexity as Chess.

(2) Thus, Chess could easily be made a game as complicated as Go by expanding the board to 19x19 with the king on the center square.

(3) Alternatively, Fischer Random chess is already an unsolvable problem, equally -- if not more -- complicated than Go.

Social Security is a pyramid scam.

really? (none / 0) (#154)
by llimllib on Wed Feb 19, 2003 at 01:16:52 AM EST

(1) Show me some math, I'll do my best to rebut it if you do. Until then, I'm going to strongly disagree. By "complexity" do you mean branching factor? difficulty of approach by algorithm?

(2) dependent on #1

(3) This claim I disagree with too, but would be more likely to believe given any evidence.


Peace.
[ Parent ]
Wow! (none / 0) (#159)
by DarkZero on Wed Feb 19, 2003 at 02:43:41 AM EST

(1) Go's increased complexity over chess is almsot solely a function of the size of the board.  An 8x8 Go game is of approximately the same complexity as Chess.

The sheer overwhelming depth of your evidence for this argument has fully convinced me. I wasn't really sure about your assertions at first, but after reading those extensive links and taking great pains to wrap my head around your extensive logical and mathematical reasoning, I have been converted to your side of the argument and would like to subscribe to your newsletter.

But seriously, though... two sentences that amount to little more than "I'm right and you're wrong, because... you're wrong" is not an argument. Not a reasonable one, at least. Take your second sentence, for instance. "An 8x8 Go game is of approximately the same complexity as Chess." All you do is make a statement and don't back it up. Anyone that knows a little bit about AI and the two games would wonder why the more dubious objective in Go doesn't matter, why the lack of separate values for different pieces in Go doesn't matter, why the presence of far more available opening moves doesn't matter (Chess starts with many spots taken up by pieces in the beginning and limits the movement of all pieces, while Go starts with all available spots open to both players from the beginning), etc. But you don't address those obvious questions. Just "I'm right, you're wrong."

[ Parent ]

Somewhere else on Kuro5hin here (none / 0) (#161)
by dh003i on Wed Feb 19, 2003 at 03:29:31 AM EST

Someone posted a link to an article explaining why it's hard for computer programs to play Go at a GM level, but why chess programs can. In that detailed article, it explained that 8x8 Go is approximately as complicated as 8x8 chess. Several things act to make chess as complicated as Go, once you eliminate superficial differences (i.e., board-size) and keep them the same in both games. (1) Yes, many moves are restricted in chess. In many ways, this serves to make the game harder, not easier. (2) Different pieces in chess have different rule-sets associated with them, which adds layers of complexity. Not all pieces can be thought about in the same manner. That said, there are some interesting properties of Go. Fractal patterns seem to emerge from master-played games, and Go seems to exhibit emergent behaviors (that is, an extremely simple rule set produces and obscenely complex game).

Social Security is a pyramid scam.
[ Parent ]

Fischer Random (none / 0) (#185)
by Magneto on Wed Feb 19, 2003 at 12:03:59 PM EST

Fischer Random has exactly the same rules as "regular chess", only a different starting position. Why would it be more complex? Why would it be unsolvable, given that chess is not?

[ Parent ]
It's not unsolvable... (none / 0) (#203)
by FieryTaco on Wed Feb 19, 2003 at 01:57:03 PM EST

It's just has 960 times more solutions. Adding one more, empty, file to a chessboard would make it more complicated than FRC does.

[ Parent ]
Opening Book (none / 0) (#204)
by jjayson on Wed Feb 19, 2003 at 01:57:35 PM EST

The only reason a computer can reach its strengths in playing chess it because if has an opening book designed by a GM to get it to positions that it plays well. Computers are very poor when far away from the end of the game and in closed positions. By effectively removing an opening book (since there would be far less human study into specific lines) you make the computer start in one of the worst possible positions for it. It really wouldn't expand on the theoretic complexity of the game (give an 80 move game at an average branching factor of 35 that would be 3.35e+123 states and multiplying that by the thousand new starting positions would only raise the complexity to 3.35e+126 states).
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
8x8? heh... (none / 0) (#205)
by Kintanon on Wed Feb 19, 2003 at 02:06:27 PM EST

You can not play go on an 8x8 board. You must have an odd number board size. 9x9 is the standard small board size. And it is actually an entirely different game from Go on a 19x19 board. You can be so phenomenally good at 9x9 go that you trounce 8th dan players at it, but you can still get your ass handed to you by some 17kyu noob at 19x19.
And computers suck ass at all forms. In fact, computers are WORSE at 9x9 than at 19x19.

Kintanon

[ Parent ]

I am a bit lost here (3.80 / 5) (#149)
by Pac on Wed Feb 19, 2003 at 12:23:49 AM EST

I am not USian, and I never heard about this Singleton person.

So, before any further discussion, who is Will Singleton and why is he qualified to speak about Kasparov? Is he a new International Chess Master I haven't noticed yet? Is he young enough to claim the World Championship before Kasparov did (at 22 years, 210 days)? Is he at least an expert in computer chess? Can he program a machine to beat either Deep Blue or Deep Junior?

No? Oh, I see. He is just another journalist talking out of his bottomless wheel of ignorance about a subject he can't even begin grasp.

By the way, Fischer is a bit overestimated. Kasparov has many recorded games where he shows a gameplay far more brilliant than Fischer ever did. Even Karpov has better moments. I think the "Fischer mythology" can be put to rest, the guy went nuts and we will never know how he would fare against those who came after him.

Evolution doesn't take prisoners


This is a classically stupid argument. (3.00 / 1) (#167)
by Menard on Wed Feb 19, 2003 at 05:49:06 AM EST

This is an argument often used against art critics, the gist being that, if they can't write a novel, then they can't say what's good or bad, and that only actual artists should critique their particula artform. This would, of course, lead to a meaningless critical dialogue in which fellow artists either patted each other on the back or sniped at rivals, rather than actually examined the work in question and attempted to divine meaning from it, and produce insightful writings. The objective critic is absolutely neccesary for useful criticism. Although that applies SLIGHTLY less here, for example I would expect the authour to know the rules of chess, and how a computer works, it still applies. You don't have to be an expert to render comment on things of this nature.

[ Parent ]
And this looks like a classical non-answer (none / 0) (#195)
by Pac on Wed Feb 19, 2003 at 12:59:35 PM EST

I disagree strongly that anyone who has used a crayon in the kindergarden is qualified to be an art critic. To be a relevant critic, in art, literature or cinema one must have a deep understanding of the craft involved, of the evolution of that particular field and the relevance of each work he/she wants to write about.

A meaningful critic here must not only have a deep understanding of chess and of Kasparov's role in the modern history of the game but must also be familiar with the technical and philosophical sides of Artifical Intelligence and machine chess.  

By the way, some of the best critics answer this questioning in kind: during the 50's, the most influential cinema critic magazine was Cahiers du Cinema. During the 60's its best writers set out to make cinema history (Google for Nouvelle Vague, Godard, Truffaut for a gist of what I am talking about).  

Evolution doesn't take prisoners


[ Parent ]
Oh, please... (none / 0) (#241)
by Menard on Thu Feb 20, 2003 at 04:32:37 AM EST

I didn't want to get into this, but here goes. To say that one needs to understand the intricacies of art history or of an artist's personal history to comment on a piece of art is bullshit. Any artist that claims that, to understand their art properly, one needs to know the history of art that culminated in it is being tremendously self-indulgent, and, more to the point, elitist. Yes, of course alusions to previous works, and the social context of  the piece, can add to one's understanding of a work, but if the work doesn't stand on it's own merits alone, then those things are just so many useless frills.

And yes, I am well aware that Godard et al. were film critics, but that doesn't mean that all good directors are insightful film critics, or, more to the point, that all great critics have any cinematic talent.

[ Parent ]

Chess? (5.00 / 1) (#202)
by FieryTaco on Wed Feb 19, 2003 at 01:51:45 PM EST

It's not like Singleton was commenting on Kasparov's chess. I didn't see too many places in the referenced article where it was pointed out that Kasparov should have made 35. ... Nb4 or any other such thing.

I anything, we have a journalist talking about the journalistic endevours of a chess player.

[ Parent ]

Fischer and Kasparov (none / 0) (#187)
by Magneto on Wed Feb 19, 2003 at 12:06:59 PM EST

Fischer has said that the games from the Kramnik-Kasparov match are superb, the best chess games he has ever seen. Of course, he thinks that the match was a sham and the games were designed by a committee of strong GMs and supercomputers.

[ Parent ]
I don't know much about chess, but... (3.66 / 3) (#168)
by Menard on Wed Feb 19, 2003 at 05:56:32 AM EST

Here's one thought -

You say that chess players should be given the opportunity to study their opponent, which of course makes sense, yet at the same time you say the match was uneven because Deep Blue was specifically designed to beat Kasparov. Isn't that the equivalent, in this case, of studying one's opponent?

Two points (5.00 / 1) (#172)
by enterfornone on Wed Feb 19, 2003 at 07:10:50 AM EST

The match was one sided because Deep Blue was designed to beat Kasparov, while Kasparov had no chance to study Deep Blue. Also, Deep Blue did not study Kasparov, human programmers did and adjusted Deep Blue's strategy accordingly, so the contest said more about their ability than the computers.

--
efn 26/m/syd
Will sponsor new accounts for porn.
[ Parent ]
Ok... (none / 0) (#174)
by Menard on Wed Feb 19, 2003 at 07:31:18 AM EST

Well, of course I agree with the point that Ksparov should have been allowed to study Deep Blue, but it seemed like the authour was claiming that Deep Blue should not be similarly allowed to adapt to Kasparov. That's what I disagreed with. As for your second point, I don't think it's much of a relevant argument. Of course the success of any machine at a task is really the success of the people that designed the machine. The machine didn't create itself, it's inanimate. Things might get a little hairier if we were to have machines designed and made by other machines a la Asimov's positronic brains, but that's certainly not the case here.

[ Parent ]
The humans may design (none / 0) (#177)
by olethros on Wed Feb 19, 2003 at 09:00:14 AM EST

the machine and its algorithms and then the machine can learn by itself. How's that?

-- Homepage| Music
I miss my rubber keyboard.
[ Parent ]
Not Really Fair (none / 0) (#199)
by DarkZero on Wed Feb 19, 2003 at 01:28:32 PM EST

So, the computer has to just have the algorithms and learn to play by itself, but human players can read books about chess, watch other people play, watch their future opponents play, etc.? Doesn't really seem fair.

Deep Blue and Deep Junior had a slight advantage by knowing Kasparov's previous games, but that's not a reason for them not to watch ANYONE play or have their strategies adapted by human beings at all (since your strategy is essentially being "written" by another human being when you read a book about chess that you agree with). It is, however, a good reason why Kasparov should've gotten a chance to watch them in action first.

[ Parent ]

Learn to play by itself? (none / 0) (#242)
by olethros on Thu Feb 20, 2003 at 04:38:27 AM EST

Not necessarily completely. Good moves (as from a book), can be added and example matches can be given. I guess a good idea would be to start training the comp on its own, then when it has achieved a good level to give it examples from books and previous matches. Then it can continue training by playing against other players. Which is quite important, I guess.

-- Homepage| Music
I miss my rubber keyboard.
[ Parent ]
Chess program "strategy" (none / 0) (#184)
by awgsilyari on Wed Feb 19, 2003 at 11:58:16 AM EST

Also, Deep Blue did not study Kasparov, human programmers did and adjusted Deep Blue's strategy accordingly, so the contest said more about their ability than the computers.

This is a nitpick. Don't take it personally, I just happen to be one of the people who knows how chess programs work.

Chess programs have no "strategy." They don't plan ahead (although they LOOK ahead) and they don't develop strategies like a human would.

A human considers the general condition of the board, and a few possible moves. Most of it is based on intuition and experience. A chess playing program, on the other hand, uses pure brute force to decide this question: "Out of all the possible moves I could make at this point, which one is most likely to give me the MOST gain, given ANY possible move by the opponent." It answers this question by searching deeply in the game tree. At no point does any "planning" occur.

The chess program has an internal model of the opponent (more specifically, given a board position, it can guess as to how advantageous that board position is for either side). In this sense it can be tuned to a particular opponent's style. Nevertheless, even though the chess program appears to be working through a strategy, it is actually just crunching numbers and picking the best move given ALL possible consequences of that move, some number of moves into the future.

Strategy seems to arise out of this, and that's what is so fantastic about AI game players. But the strategy has not been programmed into the machine -- it is just following some fairly stupid rules.

--------
Please direct SPAM to john@neuralnw.com
[ Parent ]

Strategy (none / 0) (#232)
by enterfornone on Wed Feb 19, 2003 at 08:37:23 PM EST

I admit I know little about chess programs, but certainly in the opening they can be programmed to use a particular strategy. It would be easy for someone to study Kasparov and tell the computer to prefer openings that he is known to be weaker against. It would also be able to tell based on Kasparov's style what moves he is likely to make in the future (chess computers generally calculate their opponents possible future moves as well as their own).

--
efn 26/m/syd
Will sponsor new accounts for porn.
[ Parent ]
Corporate publicity stunts (3.00 / 1) (#173)
by obsidian head on Wed Feb 19, 2003 at 07:29:33 AM EST

he wants these advances to serve for the promotion of chess, not as publicity stunts for corporations.
Kasparov has learned nothing. Years ago, Intel was a heavy chess sponsor. People started fantasizing about how chess was in a renaissance, and US children would take up the sport. Then after a while, Intel withdrew its bankrolling by saying that chess fulfilled its PR goals, and it was done.

Of course large corp sponsorship is all about PR. What was Kasparov thinking? He's incompetent as a bargainer if he didn't learn from a century of lessons.

Not very apropos... (2.75 / 4) (#189)
by kstop on Wed Feb 19, 2003 at 12:11:21 PM EST

...but I won my first game of Go yesterday, against my mobile phone. My parents would be very proud, if I could be bothered calling to tell them.

Why, oh why, did this get posted? (3.00 / 2) (#193)
by trhurler on Wed Feb 19, 2003 at 12:51:50 PM EST

See? I predicted it, and it happened. Nearly 200 comments, and nothing worthwhile.

First of all, yes, Kasparov is an asshole. Almost every truly great chess player is a real asshole; in that way, they're sort of like much less violent versions of the best heavyweight boxers in the world. However, his being an asshole is not in and of itself conclusive proof of anything BUT his being an asshole.

Second, yes, IBM did in fact cheat by any reasonable standard when Deep Blue beat Kasparov. There is no question that their in-tournament modifications would have been considered highly illegal had a human player taken the analogous action between games in one tournament. All speculation to the contrary is bullshit. Kasparov is right about one thing: if we apply precedent fairly, that tournament shouldn't count for anything.

Third, Kasparov DID screw up, and it did cost him, and it would have cost him even if IBM hadn't modified their machine.

Conclusion by someone who actually knows how to play chess and at least once understood computer chess algorithms somewhat(unlike just about everyone else posting,) and therefore has some chance of analyzing the issue: Kasparov lost because he fucked up, but had the tournament been run fairly, IBM would have forfeited anyway by their actions. The outcome proves nothing, really, and everyone ought to shut the fuck up about it.

--
'God dammit, your posts make me hard.' --LilDebbie

my point exactly (none / 0) (#212)
by dh003i on Wed Feb 19, 2003 at 03:34:01 PM EST

That the outcome of Deep Blue v. Kasparov proved nothing and had no bearing on the computational world of chess. The only reason why Deep Blue beating Kasparov was significant is because it set the stage for Kasparov to later lose to Kramnick due to an ego which was still in recovery. After his latest draw with Deep Junior, I'd look for him to get over that quickly and retake the crown.

Social Security is a pyramid scam.
[ Parent ]

et tu, trhurler? (none / 0) (#228)
by FieryTaco on Wed Feb 19, 2003 at 05:57:53 PM EST

There is no question that their in-tournament modifications would have been considered highly illegal had a human player taken the analogous action between games in one tournament.
The reasonable analogue in the human world, is consult with a coach/strategy team. This is done and is not illegal by any rational interpretations of the rules. Not to mention unenforcable, I mean how do you stop Player A from talking to uninvolved Player B about a game? Besides, Kasparov knew that IBM could/would be changing the software during the match, it was specifically mentioned in the rules.

Of course the unreasonable analogue would be such as to completely make any chess game inconsequential. I mean for God's sake, you're talking Human Brain Programming/Transplant. Oh My God! Or something.

[ Parent ]

What an eclectic collection of logic. (5.00 / 6) (#200)
by FieryTaco on Wed Feb 19, 2003 at 01:42:09 PM EST

Deep Blue wasn't sprung on Kasparov as an unknown ringer in the middle of a tournament. IBM approached Kasparov and offered to pay him to play against their computer. A set of rules was decided upon. Kasparov agreed and he lost. He knew going in exactly what the conditions and terms were. And he got paid as promised.

During a series of games each player will make adjustments to their play to better reflect their opponents strengths and weaknesses. How is the IBM team making adjustments "unfair"? Didn't Kasparov make changes to his play in order to take advantage to the weaknesses that Deep Blue showed? He had a team of advisors to suggest to him changes in his style, strategies and tactics. The IBM team making adjustments to the evaluation function isn't "unfair", it's just doing for the machine what Kasparovs coaches did for him.

Deep Blue certainly is a general chess playing computer. Any player could (assuming you had the hardware available to you) go against it and play a game of chess. Except for perhaps half a dozen players, Deep Blue would not only win, but win easily. In the general sense it is perfectly reasonable that between to any two players, each player will attempt to style their play to better defeat their opponent. But that doesn't mean that those players aren't "general" chess players.

The match was not completely meaningless. It showed that it was possible to build a machine that played a game, chess, better than the best human. That is what was done.

The observation that Deep Junior or Fritz use better evaluation doesn't indicate a lack of knowledge about chess programs. It just shows that the current collection of computer chess theory has advanced. Deep Blue is/was a strong player regardless of how cutting edge it's software is/was.

It approaches stupidity to say that computer programs have a long way to go before they are "equal" to humans re. chess. The points you bring up, ie. reduce move evaluation speed and depth, are non starters. Computer chess researches are not trying to make a human out of a computer, they are making computers that play chess. It is just as irrelevant to say that humans are not as good at chess as computers because they would play less ably than a four year old of they were restricted to evaluating moves based upon a series of static evaluations, and because they aren't able to look up moves in an opening book as rapidly, and because they can't quickly jump into transposition tables, and to handily use an endgame database.

Let's move your argument to some other field of endevour. Compare a dump truck to a human laborer. Of course the dump truck has a long way to go before it can be considered "equal" to humans at moving materials from one place to another. Reduce it's capacity to a five gallon bucket and only allow it to move by lifting it's wheels and placing them down one in front of another.... See, it makes no sense. The point isn't to make a better human, it's to make a machine that moves stuff from here to there as efficiently as possible and the only intelligent way to do that is to use the strengths of the machine, not to try and make a stronger/faster person.

In some sense, it's valid to say that computers are already equal to humans. Considering that the difference between human players is how they choose their moves. One player may have a bias towards using his gut instinct, whereas another may be more prone to following a strict methodology based upon study of opening books and board position. So it doesn't make sense to say that because a player uses a different method of deciding upon moves they aren't "equal" to other player, you say they aren't "equal" if they win or lose. If you want to compare how good two things are, you look at what the goal is and how well that goal is achieved. Taking the average human chess player and comparing their ability to solve the problem, ie. play chess, the average chess program is better at solving that problem.

It also doesn't follow that changing the initial layout of the pieces (Fischer Random Chess) will result in computers being terrible at the game. It just means that you've changed the problem and the solutions need to be changed as well. However, even if you chose not to adjust your solution, programs such as Deep Blue would still do very well at the game. Deep Blue's primary strength was it's speed. Programs such as Deep Junior and Fritz, most likely would be less capable than they are now because their opening books would be destroyed. Programs such as Cray Blitz and Deep Blue wouldn't be as affected by such variations, since they are based heavily upon the premise that they can evaluate an enormous number of positions. The only thing that FRC would do is kill the opening book. This exact same effect would happen to human players as well.

I don't know much about Kasparov, so I won't make any comment about his ego or personality. However if you want to disagree with points that are being made, you would be better served by sticking with arguments that are logical and have some connection between point A and point B.

The whole argument about who is better at chess, computers or humans, is stilly. It's obvious that, given an economic incentive, a computer can be built that will crush handily the best human players. It's purely a matter of money. The same engineering principles that led to Deep Blue, ie. massive parallelization, can be extrapolated to a machine that is hundreds if not thousands of times faster than Deep Blue. Add in the very large data storage capabilities that have arisen in the last few years (increased RAM density, hard drive space, database searching techniques) and you have a combination that couldn't be beat by human methods.

It's a foregone conclusion that we are able to build tools that solve problems better than we solve those problems unaided.

adjustments (5.00 / 1) (#201)
by jjayson on Wed Feb 19, 2003 at 01:49:41 PM EST

During a series of games each player will make adjustments to their play to better reflect their opponents strengths and weaknesses. How is the IBM team making adjustments "unfair"? Didn't Kasparov make changes to his play in order to take advantage to the weaknesses that Deep Blue showed? He had a team of advisors to suggest to him changes in his style, strategies and tactics. The IBM team making adjustments to the evaluation function isn't "unfair", it's just doing for the machine what Kasparovs coaches did for him.
The idea was that it was man versus machine, not man versus machine and a group of men. Those adjustments were made by humans. I don't think anybody would have a problem if Deep Blue made adjustments to its own evaluation function automatically.  Human intervention is what was objected to.
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Why would he be upset with changes? (4.50 / 2) (#211)
by FieryTaco on Wed Feb 19, 2003 at 03:31:39 PM EST

Looking at the rules of the match, item 13 seems to say that IBM is allowed to change the hardware/software at their descretion. The only caveats being that they do so so as to minimize impact on his concentration if it is necessary to change during a game.

But ultimately it would be more interesting/astonishing to have the machine handle everything itself.

[ Parent ]

which is exactly why the match was meaningless (none / 0) (#216)
by dh003i on Wed Feb 19, 2003 at 04:02:16 PM EST

This is effectively institutionalized cheating, making the match meaningless tripe. I didn't conform to normal chess protocols and precedent: IBM practically strongarmed Kasparov into accepting the terms. Irrelevant of whatever you or any other IBM supporters say, that match didn't mean ****.

Social Security is a pyramid scam.
[ Parent ]

Strongarmed? (5.00 / 1) (#227)
by FieryTaco on Wed Feb 19, 2003 at 05:36:01 PM EST

You mean they offered him money, $400,000 if he lost, or $700,000 if he won. I can see how that's exactly like kidnapping his family and putting a gun to his head. Yeah, what horrible behavior on their part. Christ, in the rules there are all kinds of considerations for Kasparov.

The only thing Kasparov put up was his reputation with the public, and the vast majority of the public doesn't even know who the hell Gary Kasparov is. Not any kind of FIDE (or whichever chess governing body he is under) ranking was at stake, or even putting up some of his own money to go to the winner. This was an exhibition match.

You seem very emotionally attached to the idea that a human can't be beat in a "fair" (for your definition of "fair") game of chess. Give it up, it's not that important. There are machines that do almost everything people do, only better. Think you can throw a baseball far and accurately? Think again cause that machine over there can hit a 15 inch bullseye at 600 yards. Think you are a fast runner? Too bad, this machine over there does 0-60 in 7 seconds. Think you are strong? Well, this crane can lift tens or hundreds of tons. It doesn't mean that we shouldn't enjoy our endevours, it just means that we are capable of building machines that do certain tasks better than we can do those tasks.

[ Parent ]

right (none / 0) (#233)
by dh003i on Wed Feb 19, 2003 at 09:05:28 PM EST

You seem very emotionally attached to the idea that a human can't be beat in a "fair" (for your definition of "fair") game of chess Nope, not emotionally attached to that idea. May be possible in the future, but isn't possible yet -- Junior and Fritz were only able to tie the #1 and #2 players in the world, respectively. The simple fact is, Blue v. Kasparov was not fair by any sense of the word. PERIOD. You can yap all you want about how much Kasparov was payed, but money does not make a match fair. That match was in violation of traditional chess precedent. It was meaningless. Didn't prove anything about a computer's ability to play chess, except that when you handicap the human with every conceivable disadvantage, he loses. And if you broke my knee-caps, I wouldn't be able to run a marathon. Big fucking deal.

Social Security is a pyramid scam.
[ Parent ]

Hahahahahaha (5.00 / 1) (#236)
by FieryTaco on Wed Feb 19, 2003 at 11:17:04 PM EST

...handicap the human with every conceivable disadvantage, he loses.
Yeah. Whatever. You're wrong. He wasn't handicapped with every conceivable disadvantage. Kasparov himself had played against a previous version of Deep Blue. The 1997 version used the same basic software and hardware (whatever they called the chiptest derivative) as the original. The primary differences were in speed and some evaluation function tuning. But the fundamental engine was the same.

What you're arguing for is that Deep Blue be written, have a bunch of published games and never be changed. That's not how human players are. Humans change over time, they learn, they try different tactics, they have migrating preferences for strategies. If you want to talk about "chess precedent", there's absolutely no "chess precedent" for two players, A & B, where A get's to examine B's play and B doesn't get to change at all, while A gets to adapt. What's more, Kasparov defintely "researches" chess, plays games with his friends, discusses strategy and tactics, all in private. By your definition of fair, Kasparov should have given the Deep Blue team a complete compendium of his chess knowledge and thinking and then stopped doing any thinking about chess at all while they went and adjusted Deep Blue.

Anyway, like I said in my previous post, unless you can come up with some factual basis for your arguments I'm done.

[ Parent ]

The point (none / 0) (#259)
by GhostfacedFiddlah on Sun Feb 23, 2003 at 12:30:11 PM EST

Most basically - did an unattended machine beat Kasparov?  No.  Yes, he lost by the rules he agreed to, but that doesn't mean that IBM is at all right in its assertions that machine is now better than man at chess.  Quite simply, Deep Blue wasn't good enough at adapting to a playing style, and needed some human help.  There's more to chess than just playing chess well, it's adapting, and Deep Blue was unable to do that.

[ Parent ]
in way-back retrospect (none / 0) (#262)
by dh003i on Thu Jul 10, 2003 at 10:04:28 PM EST

This is long-dead by now, but the purpose of that rule was as a safeguard for IBM if the hardware broke down or if the software had some bug in it such that it wasn't responding. And surely that was how it was advertised and how Kasparov took it; it was not for them to be able to alter the program mid-match to adjust to Kasparov's style of play.

Social Security is a pyramid scam.
[ Parent ]

Wow, someone actually rated that tripe a 5? (2.33 / 3) (#213)
by dh003i on Wed Feb 19, 2003 at 03:40:52 PM EST

I'd rank you down, except I want the rest of K5 to see how dumb you are.

The purpose of these computer v. human matches is to see how good the chess program is.  Changing it in the middle of the tournament destroys that goal.  IBM most certainly did cheat by the precedent standards of chess:  had the corresponding thing been done by a human, it would have most certainly been considered cheating.

Again, exactly what grand masters -- aside from Kasparov -- did Deep Blue play again?  A bunch of nobodies.  Deep Blue didn't play against a single GM of any significance outside of Kasparov.  Didn't play against Karpov, or any of the other former-champions known on an international scale.  IBM promptly dismantled it afterwards, perhaps to eliminate proof that their program was built and then modified specifically to play Kasparov.

Had the program been able to modify itself, that would be fine.  But it didn't.  It got outside help, which means that it's no longer Kasparov v. chess program (no longer man v. machine), but rather Kasparov v. chess program + GM's and programmers IBM use to modify code in the middle of the match.

Programs like Deep Fritz and Deep Junior are VASTLY superior to Deep Blue.  This has been proven in as much as is possible -- given that IBM destroyed the evidence -- by having these programs suggest alternate moves to the ones Deep Blue did in the Blue v. Kasparov match.  Every GM of any significance agrees that Fritz and Junior suggest better moves.

Social Security is a pyramid scam.
[ Parent ]

Yummy, tripe soup... (5.00 / 1) (#225)
by FieryTaco on Wed Feb 19, 2003 at 05:20:07 PM EST

What corresponding thing? To have a coach consult with them and give suggestions on how to change their play? It happens all the time during a multigame match. Not only does it happen, but at grandmaster/international grandmaster levels, you would be considered stupid not to have a coach/team to talk to about your games.

Who cares what masters it has played. Kasparov knew going into it that he was playing an "unknown". He knew he didn't have access to the software/system. Hell, he even knew that they could change the software (see point #13.) He knew that they were trying to beat him and would style Deep Blue's play to do so. None of this stuff was a surprise to him. He felt he would win anyway, turns out he was wrong. Then he started coming up with reasons why he lost. Well, he knew about those reasons before hand, the only person he has to blame is himself. Deal with it.

But to address one other point:

...but rather Kasparov v. chess program + GM's and programmers...
The IBM team involved one international grandmaster, namely Joel Benjamin. A man who you discounted in the article as being unknown and unnoteworthy. Not multiple GM's as you claimed in this post.

Finally, it's irrelevant whether Deep Jr. or Fritz are hundreds of times better then Deep Blue. I don't think anybody has said differently, all that's been said is that Deep Blue was a strong program, which it was. The only reason I can see to so strongly point out that Deep Junior is better, is so that it can be said that Kasparov played to a draw against a program stronger than the one that beat him as some kind of revalidation of Kasparov's chess ability.

It's almost as if you are trying to say that Deep Blue wasn't capable of playing a general chess game, but rather only capable of playing Kasparov. Do you really think that a generally weak player would be able to beat one of the best players in the world (if not the best) if they had specialized their play specifically against that player? Don't be stupid.

[ Parent ]

re (none / 0) (#234)
by dh003i on Wed Feb 19, 2003 at 09:11:11 PM EST

<I>What corresponding thing?</I>

Changing the code in the middle of the program was effectively like giving Kasparov an entirely new opponent.  It would be like Kasparov switching places with another GM in the middle of the match.

<I>The only reason I can see to so strongly point out that Deep Junior is better, is so that it can be said that Kasparov played to a draw against a program stronger than the one that beat him as some kind of revalidation of Kasparov's chess ability.</I>

It's not irrelevant.  My entire point is that Blue v. Kasparov was unfair to Kasparov.  He accepted that unfairness, but that does not change the fact that it was an unfair meaningless match.  The FACT -- FACT, as confirmed by every GM who's looked at Fritz and Junior v. Blue -- is that both programs are stronger than Blue, as they both suggested superior moves to Blue in Blue's game against Kasparov.

We don't need to revalidate Kasparov's chess ability.  He's the best active chess player in the world, and has been for longer than a decade.  In the entire history of chess, when you talk about the "best chess player ever", there's only two names that you can legitimately put up for that title:  Kasparov and Fischer.

Social Security is a pyramid scam.
[ Parent ]

What were the changes? (5.00 / 2) (#235)
by FieryTaco on Wed Feb 19, 2003 at 10:57:38 PM EST

Changing the code in the middle of the program was effectively like giving Kasparov an entirely new opponent. It would be like Kasparov switching places with another GM in the middle of the match.
OK, list what changes were made. If the changes involved replacing the position database, replaced the entire evaluation function, replaced the entire endgame database, etc. then it could be argued that they "changed" players on him. However, if they recognized that when they had an advanced pawn on the e- or f-files and a rook kingside in the middle of the board their eval function suggested an inferior move and they fixed it, then no, it's not the same as "changing" players. The second situation would be exactly like if Kasparov went to talk to Dokhoian about the game and was advised that he wasn't developing his knights adequately. So, unless you can document what changes they made to the program, mid-match, your argument is bullshit. It's much more likely that there was some small tuning going on rather than a complete change, based purely upon the nature of programming.

It's not irrelevant. My entire point is that Blue v. Kasparov was unfair to Kasparov. He accepted that unfairness, but that does not change the fact that it was an unfair meaningless match. The FACT -- FACT, as confirmed by every GM who's looked at Fritz and Junior v. Blue -- is that both programs are stronger than Blue, as they both suggested superior moves to Blue in Blue's game against Kasparov.
Yes, it is irrelevant. Not only is it irrelevant, it's totally and completely irrelevant. There is absolutely no connection between the quality of Deep Junior and Fritz and the "fairness" of the Kasparov v. Deep Blue match. None whatsoever. Zero. Zip. Nada. Get it? There's no connection.

If you want to talk "fair", then consider this: For the Kasparov v. Deep Junior match he had a copy of DJ to player against at his will for six months before the match. Did the Deep Junior team have the same access to Kasparov? Nope. Kasparov could have played over a hundred games and examined DJ's evaluation of thousands of positions. What did the DJ team have? A few hundred published Kasparov games. I dont' see the Deep Junior team out there bitching about how the match wasn't "fair".

I'd also be interested in the background of your massive "FACT" (by the way, when you write it in all caps like that I can see the spittle flying from your mouth and it just provides so much more weight to your argument.) Give me a reference to Deep Junior and Fritz providing alternative moves, that are accepted by the GM community as being better moves, in the same time constraints as Deep Blue was playing under. Otherwise the ability of either program to generate a stronger move, in say 100 hours instead of 5 minutes, doesn't in any way indicate that they are better programs. (As an aside, here's what Kasparov said on the subject: "Unfortunately, because Deep Blue's records were never made public, it is quite useless to discuss the strength of Deep Blue vis-ŕ-vis the strength of Deep Junior.")

In the entire history of chess, when you talk about the "best chess player ever", there's only two names that you can legitimately put up for that title: Kasparov and Fischer.
If only those two names come to mind when you talk about the best chess player ever, you're either an idiot or uninformed. Fischer definitely isn't in the running, yes he was a prodigy, but his unwillingness to play for thirty years pretty much puts him off of anybody's list (the only exception in that time was the 1992 match against Spassky.) There are numerous other players, Capablanca, Morphy, and Steinitz, to name a few, that would be valid considerations for best chess player ever.

Unless you can do something other than argue irrelevancies, this will be my last post on this subject. It's plain you aren't arguing from any kind of rational position.

All this crap aside, it's true that Kasparov is being a sore loser. He knew the conditions going in. He accepted them. Whether it was unfair, whether they put viagra in his water, whatever, it was all up front. Bitching and making excuses after the fact, that's pretty much the definition of being a sore loser.

[ Parent ]

That is so stupid (none / 0) (#226)
by nusuth on Wed Feb 19, 2003 at 05:21:40 PM EST

Did you think the computer programmed it itself in the first place? It always was (and for the foreseeable future, is) "chess program + GM's and programmers IBM use to modify code". The "middle of the match" part was by the rules. Just stop whining and insulting everyone, will you?  

[ Parent ]
I see somewhat where you are coming from... (4.00 / 1) (#223)
by amarodeeps on Wed Feb 19, 2003 at 05:10:48 PM EST

And while I see how you find this piece logically inconsistent, I think the author had a good point when he said that the algorithm used by computers is not as 'effective' as that which is used by humans--brute force is one thing, the odd methodology used by grand masters, which we don't fully understand, is another. And as of yet, computers can't use that same method in an effective way: when they are reduced to using a similar methodology, like a reduced amount of looking ahead, they will lose. I'm not sure exactly what this means, but it does teach us something about the inequality of computers and humans, and points out our continued ignorance in really figuring out what it is to be intelligent.

I guess what I'm trying to say is that: yes, maybe deep blue proved computers can beat humans, or at least one of the best humans right now, it didn't necessarily say too much about human intelligence in general or even specifically what it means to be a grand master at chess. I wouldn't be surprised if a new generation will learn to beat this current crop of computers and we will have to start over trying to figure out what's going on.



[ Parent ]
Kasparov's point (none / 0) (#222)
by tgibbs on Wed Feb 19, 2003 at 05:08:00 PM EST

Kasparov's main point is that IBM wasn't doing computer science with Deep Blue, they were seeking a public relations coup. That being accomplished, they dropped the project. Issues of "fairness" (which hardly seem relevant in a match as unbalanced in so many ways) aside, Kasparov has a point. Did IBM create a computer generically capable of playing chess at the grandmaster level? Or a Kasparov-beating computer? They aren't necessarily the same thing.

Kasparov is clearly willing to risk his reputation by playing machine opponents. We can hardly blame him if he'd like something more to come out of it than advertising.

Seems like there are a few unspoken things here. (none / 0) (#229)
by amarodeeps on Wed Feb 19, 2003 at 06:36:37 PM EST

What's the point of all this? To prove that computers can beat humans at chess? Or to understand how humans play chess vs. how computers play chess, i.e. what it is that is expressed about intelligent when chess is played? Is there something about chess that computers just can't understand?

To clarify: if Kasparov is the best, and he has been beaten by a computer does that mean that we've figured out everything there is to figure out about mastery of chess? I think that many people, even if they would suggest that perhaps forever after computers will beat humans—they might still not agree with the previous statement. So what does this mean for computer vs. man from now on, is the endeavor totally spent, or is it just a way for corporations with a vested interest in chess to promote their products? Or do we need to move on to something that is even harder to make a computer play well, like "go" (as another poster suggested)? Or—how about this—maybe the next generation of chess masters will, from the beginning, be playing chess programs, studying those systems' weaknesses and strengths (and perhaps here is where trhurler is right and I should not be talking about chess and computers) and at a certain point the tables will be turned again, and programmers will need new algorithms to beat humans, and this cycle repeats ad nauseum...?

I'm not suggesting I know any of the answers to these questions, it just seems like many people are talking back and forth while managing to completely ignore this larger context...what are the thoughts on this?



I nominate this for the Troll of the Year Award... (3.60 / 5) (#230)
by NoMoreNicksLeft on Wed Feb 19, 2003 at 06:51:48 PM EST

Grandmaster chess players only consider 2 moves a second? Where did you study neurology or psychology again?

Gimme a break. No one claims to understand how the human mind truly computes things of this nature, but most people who hold doctorates in relevant fields of study would agree that the human mind/brain is easily capable of more computation than that. It's just hidden from our consciousness.

Yes, Deep Blue wasn't exactly fair... but the point wasn't that it was a fair match up, rather that under some circumstances computers were becoming good enough to compete for once. Only Kasparov, and aparently you, would continue to whine about it though.... if it were me, the excuse that "the computer was able to consider every possible move, approximately 9 bazillion of them" would suffice.

Get over it.

Machines are also stronger than I am, and can certainly move faster. But we don't see Olympic sprinters at Daytona chasing the cars down, and then acting like children because the cars "cheated". Not only is Kasparov a sore loser, but you are a sore loser fan of a sore loser.

--
Do not look directly into laser with remaining good eye.

The field is "artificial intelligence" (none / 0) (#253)
by p3d0 on Thu Feb 20, 2003 at 03:23:47 PM EST

If someone claimed to make an "artificial sprinter" that used wheels instead of legs, I think that could be considered cheating, especially if the other competitors were not informed of it until halfway through the race.
--
Patrick Doyle
My comments do not reflect the opinions of my employer.
[ Parent ]
Uh... (none / 0) (#255)
by NoMoreNicksLeft on Thu Feb 20, 2003 at 05:43:57 PM EST

So Kasparov thought there was a secret GM in the big blue IBM box? He only discovered that it was a machine afterwards?

I used my analogy in such a way that it was relevant to both my point and the situation it paralleled. You've turned it upside down, twisted it inside-out, spun it until it was dizzy and then raped it with a tire iron. What point are you trying to make?

PS This would be really ironic, if the post were made by the guy's Eliza bot or something, wouldn't it?

--
Do not look directly into laser with remaining good eye.
[ Parent ]

The point I'm trying to make (none / 0) (#258)
by p3d0 on Sat Feb 22, 2003 at 10:14:12 AM EST

The point I'm trying to make is that it is a matter of degree. The question of fairness in this case is quantitative, not qualitative. You can't just say "he knew it was a machine, so nothing is unfair".

There's a specific challenge he thought he had been given, and if the task he was actually assigned was not what he thought it would be, then that is a miscommunication at the very least.

Consider this: Deep Blue's program was altered mid-tournament. Is that unfair? Does it amount to human assistance? I don't know, but just how far could we tip the table in DB's favour? Could we allow DB's programmers to alter the program before each game? How about before each move? Could we put a 0.1-second time limit on each move? Could we say that Kasparov is not allowed to move his bishops?

Where does it start to become unfair? My point is only that I don't think it's a simple question. I think Kasparov's point of view is a valid one, as is yours.
--
Patrick Doyle
My comments do not reflect the opinions of my employer.
[ Parent ]

GM level chess (4.00 / 1) (#231)
by pheta on Wed Feb 19, 2003 at 07:23:17 PM EST

It is generally accepted in the chess world that if another grandmaster (perhaps Karpov) played Deep Blue in the configuration it was in at that time they would be able to win (the GM). Deep Blue was finely tuned to play against Kasporovs playing style. It 'knew' all of the variations he likes to play, and was hardcoded to take advantage of them. Every GM has their own style, and when the computer is finely tuned to play well against styles X-Z, someone with style W would fair much better against the program.

Another point that should be made is that the Deep Blue team had access to thousands of Kasporov's past games. Kasporov had access to none of Deep Blue's. This may seem silly to those not aquainted with chess, but this is a big deal to a GM. If the computer (and the team that programmed it) did not have access to a single Kasporov game, the computer's chances of winning would be drastically reduced.

For the record, Kasporov was given a copy of Deep Junior well in advance of the match so he could study its chess and how it plays.

Grocer's apostrophe (none / 0) (#245)
by Ubiq on Thu Feb 20, 2003 at 07:04:22 AM EST

grandmaster's → grandmasters

You do not understand how the brain works... (none / 0) (#248)
by skyknight on Thu Feb 20, 2003 at 11:34:16 AM EST

It would seem that you are comparing the processing power of the human brain to that of a 286 or 386. This is not the case. The human brain is nothing like a serial processor, but rather a neural network that is capable of massively parallel computation.

While the time required for individual connections in the human brain to fire is woefully slow, measured in milliseconds, the enormous number of them operating seamlessly in parallel makes the computational capacity of the human brain blow super computers out of the water (for the moment, but not for much longer).

It just so happens that human brains are not well suited to blindingly fast number crunching, or at least not number crunching in the way you think of it. Rather, they are fine tuned for tasks such as vision, motor manipulation, and other such skills that have historically been of use during our evolution as a species.

While a human is not going to be able to perform a gross analysis on as many moves as a computer in a given time slice, he will probably be linking together possibilities in a way that is not practical, or even possible, on serial hardware.



It's not much fun at the top. I envy the common people, their hearty meals and Bruce Springsteen and voting. --SIGNOR SPAGHETTI
Humans suck at math, but . . . (none / 0) (#251)
by hardburn on Thu Feb 20, 2003 at 02:21:08 PM EST

The human brain is really good at abstracting away complex operations without the individual being aware of it.

Try this: hold your finger about six inches from your face and move it back and forth while keeping your eye on it. Unless you're schizophrenic, you'll be able to keep your finger centered in your field of vison without even thinking about it.

Now write a computer program to do the same thing. Once finished, you probably have a Masters thesis wrapped up, if not a doctorate. (Assuming this hasn't been done already, which I'm not sure of).

Computers are really good at munching numbers. Compare the games of checkers to chess. Within a few years of starting, a checkers program was written that could beat any human. It took the same ammount of time to work the bugs out of the chess program just so it could figure out what a valid move is.

From this, we can infer that checkers is relitively easy to express mathmatically, while chess is a far more abstract game. Abstraction doesn't mean it is impossible for computers to do--just that it will take a lot longer to dig through the layers to find the underlieing mathmatical system.

Getting a computer to play chess is mostly intresting because we can find out how they will play. Deep Blue was mostly a PR stunt by IBM (remeber, IBM is just the sort of company that would spend a few million dollars on a Linux-enabled coffee machine, just to show it can be done), and I don't know if we can really learn anything from it. However, more generally, we can see styles of play from a computer that have never occured to humans before. I imagine that should we ever meet a sentient alien species, we could teach them chess and get a similar result.


----
while($story = K5::Story->new()) { $story->vote(-1) if($story->section() == $POLITICS); }


checkers (none / 0) (#252)
by jjayson on Thu Feb 20, 2003 at 02:29:01 PM EST

Computers are really good at munching numbers. Compare the games of checkers to chess. Within a few years of starting, a checkers program was written that could beat any human.
That isn't quiet true. The world champion for years as Dr. Marion Tinsley and he never lost to even he best computer opponent, Chinook until he had to forfeit a match because he died. Dr. Tinsley has a great quote about game playing computers, "Chinook was programmed by Jonathan [Schaeffer], but I am programmed by God."
_______
Smile =)
* bt krav magas kitten THE FUCK UP
<bt> Eat Kung Jew, bitch.

[ Parent ]
Deep Junior "Sore Winner?" (2.33 / 3) (#256)
by rasmoh on Thu Feb 20, 2003 at 06:33:04 PM EST

While Kasparov's comments might be a bit self-pitying, this comment from his opponent does not in any way become a champion of the game we all love.

[Kasparov,] You're Such a Loser
Deep Junior

Bite my shiny metal ass!

Inexcusable. I am quite appalled.

'Twas the pride of the peaches.
Deep blue proved a lot (none / 0) (#257)
by svampa on Sat Feb 22, 2003 at 05:54:11 AM EST

except that when the computer is given every possible advantage, and the human every possible disadvantage, the human will lose.

Not too bad. Before Deep Blue, not even with such advantages, a computer was able to win the world champion (neither a Great Master).

The only thing that could be dirty play is if, like Kasparov says, there was a human behind the computer. In this case, they only proved that a geat player helped by a computer can win the world champion.

Perhaps didn't proved that computers play chess better than human beings, as they hype, but they proved that fine tuning the software for the opponent, the computer can win the best human. Something that computeres couldn't do before. A great step.

The other thing I've heard about is that now Kasparov has played true chess. Against deep blue he played "chess for cumputers", that is, a fake chess that consist in choosing movements, not because they are so good in natural chess, but because they make things harder for a computer:

  • Breaking standard openings, so the computer must leave the database soon and begin to spend time before
  • Strange movements to make the program spend time deep analyzing unexpected movements, and so reject as improbable, and not deep analized before
  • Avoid taking pieces. The more pieces are in the board, the more posibilities must analize the software.
  • So kasparov did movements that he wouldn't dare to do against a human. It worked with other software, but not with deep blue.



    Ouch (none / 0) (#260)
    by Lelon on Mon Feb 24, 2003 at 12:36:14 AM EST

    I strongly urge you to read "Behind Deep Blue" as it refutes just about everything you say. It is true, however, that Deep Blue was altered during the match. Let me be a bit more clear, it was altered between games. I have no problem with that, but some find it an issue of contention. Any chess expert will tell you that Kasaparov lost because he consistently adopted "anti-computer" chess strategy, even after it became painfully obvious that Deep Blue doesn't play chess like any other computer ever built.


    ----
    This sig is a work in progress.
    Really? (none / 0) (#261)
    by dh003i on Fri Feb 28, 2003 at 12:10:42 AM EST

    Really, since IBM destroyed all of the evidence, how can we know anything about Deep Blue, other than what they say?  And we know damn well that they will say whatever is in their best interest (publicity-wise) irrelevant of whether or not it's true.  Especially so in this case, as it's not criminal to lie about this matter.

    Sufficed to say, we can't trust a word IBM says.

    Social Security is a pyramid scam.
    [ Parent ]

    Kasparov criticised as sore loser after Op-Ed on Deep Blue | 262 comments (241 topical, 21 editorial, 0 hidden)
    Display: Sort:

    kuro5hin.org

    [XML]
    All trademarks and copyrights on this page are owned by their respective companies. The Rest © 2000 - Present Kuro5hin.org Inc.
    See our legalese page for copyright policies. Please also read our Privacy Policy.
    Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
    Need some help? Email help@kuro5hin.org.
    My heart's the long stairs.

    Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!