Computer Science Canada :: View topic - Deep Blue Insane search!

Computer Science Canada

Deep Blue Insane search!

Author:	MysticVegeta [ Mon Jan 16, 2006 12:36 pm ]
Post subject:	Deep Blue Insane search!
Holy sh!t!! Deep Blue (the computer who beat Kasparov, 3.5-2.5) has an insane search depth, no wonder it won!! It searchs 200,000,000 chess positions per second! Thats Insane. Its a search depth of 6 in beginning game, and 10-20 in middle/end game. I found it to be quite exciting. Its predicted ELO rating is 3400! Kasparov = 2900. Although I dont think a perfect machine with infinity ELO could be made because there are 10^134 chess positions possible compared to universe which is only 10^126 nanoseconds old. Although, maybe Nanotech has a solution to make a perfect chess machine. I came across one of these article before, I will post the link if I find it. What do you guys think about this? Isnt this crazy?

Author:	Dan [ Mon Jan 16, 2006 2:41 pm ]
Post subject:
I think it is prity cool as well but i bet some day we will see ones with total deepth that will perdicted all posable moves. It is prity intresting to me since i just fished a compsci class that talked about game trees and this kind of things. It is not all about computer power ether tho, the big trick is finding what parts of the game tree you no longer need and do not need to sreach witch drastickly cuts down on the cpu time.

Author:	Andy [ Mon Jan 16, 2006 3:18 pm ]
Post subject:
i love how the word nanotechnology gets thrown around everywhere.. you guys do realize that 70% of nanotech is really about the chemical and biomedical right? nanotech at the current stage is going no where near super computing.. and it wont for a long time

Author:	Dan [ Mon Jan 16, 2006 3:35 pm ]
Post subject:
Andy wrote: i love how the word nanotechnology gets thrown around everywhere.. you guys do realize that 70% of nanotech is really about the chemical and biomedical right? nanotech at the current stage is going no where near super computing.. and it wont for a long time I did not mention nanotechnology in my post , but i was thinking more of the apications of quotom computing and other lines of resarch that whould vastatly improve cpu power. Tho nanotechnology may not be going close to it now, if it is posable then it whould have an application for supercomputing at some time. You have to rember that computers in the stage we know them have only been around for 50 or 60 years so it is not unreasnable to dream of such things since they maybe posiable in our lifetime.

Author:	Hikaru79 [ Mon Jan 16, 2006 9:16 pm ]
Post subject:
I have a copy of the documentary "Game Over: Kasparov versus the Machine", which is a very interesting look at the construction of Deep Blue, and especially about the series against Kasparov. In fact, its practically narrated by Kasparov. He's in it everywhere. It talks about a lot of things, including the controversy surrounding the second game of the match (long story short, most chess players believe Deep Blue cheated). I can send a copy if anyone would like. It's really great, trust me.

Author:	Martin [ Mon Jan 16, 2006 9:26 pm ]
Post subject:
How would Deep Blue cheat?

Author:	Hikaru79 [ Mon Jan 16, 2006 9:53 pm ]
Post subject:
Martin wrote: How would Deep Blue cheat? There's a particular move in Game 2 that virtually every Chess grandmaster agrees could not have been made by a computer, and was completely out of the league of its previous play. They figured a team of GM's (which IBM has admitted contributed heavily to book play of Deep Blue) "intervened" in that point because it was clear that the computer would have made a mistake in that situation. It seems rather iffy, but the suspicious part is that, when confronted with these accusations, IBM simply REFUSED, tooth and nail, to show anyone the computer's logs (decision trees, etc). This was only supposed to be a 'friendly' match, so this sudden defensiveness is rather ... strange. Anyway, watch the documentary and you can make up your own mind

Author:	md [ Mon Jan 16, 2006 10:33 pm ]
Post subject:
Then Deep Blue didn't cheat, it's controllers did. That just because there are 10^134 chess positions possible has nothing to do with the universe being only 10^126 nanoseconds old. Computers are getting faster and faster, already modern desktop processors are aproaching 4 instructions per nanosecond. Supercomuters are far beyond this. I can see a modern super computer being able to solve the chess game tree in a few years. Oh, and assuming the universe is 15 billion years old, that's only 4.7335389x 10^26 nano seconds. Even 25 billion years is only 7.88923149Ã—10^26 nanoseconds. 10^126 nanoseconds is 3.16887646 Ã— 10^109 years, which is a hell of a lot longer then any of the reasonable guesses as to the age of the universe.

Author:	Amailer [ Mon Jan 16, 2006 10:34 pm ]
Post subject:
I watched a documentary on this... Pretty interesting actually- it sorta portrayed the IBM guys as Evil Microsoft people, or just evil people. What Hikaru79 said- there was one move, second round was it? Either way, that move wasn't `computer like`, it didn't go after the kill- rather it made a strategic move (sacrifice I think), so Kasparov got suspicious, I think he went a little crazy too.

Author:	Hikaru79 [ Mon Jan 16, 2006 11:05 pm ]
Post subject:
Cornflake wrote: Then Deep Blue didn't cheat, it's controllers did. When I say "Deep Blue", I'm not reffering to just the hardware, but rather the team behind it as well. They would be the cheaters in that situation. Amailer wrote: What Hikaru79 said- there was one move, second round was it? Either way, that move wasn't `computer like`, it didn't go after the kill- rather it made a strategic move (sacrifice I think), so Kasparov got suspicious, I think he went a little crazy too. Yeah, that one move in game 2 was what triggered the suspicion -- it was IBM's reaction to that suspicion (extreme hostility and secrecy and defensiveness) that really got a lot of people wondering.

Author:	codemage [ Tue Jan 17, 2006 1:32 pm ]
Post subject:
DB wouldn't have to computer all possible positions to be a perfect chess player. Some positions (and therefore any resulting positions in the decision tree) are so sub-optimal that they don't need to be computed. Brainless example: even though there are 20 possible opening moves, most of these are not optimal to play.

Author:	MysticVegeta [ Tue Jan 17, 2006 3:28 pm ]
Post subject:
Hikaru, you have the documentary in what? Video or book?

Author:	Hikaru79 [ Tue Jan 17, 2006 4:00 pm ]
Post subject:
MysticVegeta wrote: Hikaru, you have the documentary in what? Video or book? It's a video. If there's a book out there, I'd be interested to hear about it.

Author:	MihaiG [ Tue Jan 17, 2006 5:50 pm ]
Post subject:
yes a lot of contreversy is still evident trhough many chess servers but most programs now are just as power full as DB, such as Fritz 9 and Deep Fritz 8 (multi-proccesor ver.) an new more power full engine made and tested against british gm michael adams named Hydra beat him....

Author:	MysticVegeta [ Wed Jan 18, 2006 3:36 pm ]
Post subject:
Hikaru79 wrote: MysticVegeta wrote: Hikaru, you have the documentary in what? Video or book? It's a video. If there's a book out there, I'd be interested to hear about it. dam I would like to watch it, where did you buy it? ahem, cough, Can you copy and send it to me. LOL. [/idiotic question] El commandmente: Are you the one who Cervantes hates?

Author:	Andy [ Wed Jan 18, 2006 5:12 pm ]
Post subject:
hes the one we all hate

Author:	Cervantes [ Wed Jan 18, 2006 6:07 pm ]
Post subject:
MysticVegeta wrote: El commandmente: Are you the one who Cervantes hates? Double-you tee eff? Don't be putting words in peoples mouths, now. Hate is a strong word.

Author:	[Gandalf] [ Wed Jan 18, 2006 7:45 pm ]
Post subject:
Actually, Elcomandate is right. There are a multitude of improved chess engines available nowadays, Fritz, Deep Fritz, Hiarcs, Hydra (although it's not available to the public), and some others which would easily beat Deep Blue on the same kind of hardware. If I remember correctly, after Kasparov beat Deep Blue for the first time the IBM team came to him for help. They wanted to adapt the program for just his kind of playing style. In return he got to see the search logs and everything like that which Deep Blue could give. Could this have something to do with the unexpected move? I doubt Deep Blue's ELO would be 3400, especially in competitions against other supercomputers. edit Sigh... I really should stop waiting hours before posting...

Author:	MihaiG [ Wed Jan 18, 2006 9:03 pm ]
Post subject:
yes....personaly Fritz 9 the newest one is the most powerfull single proccesor engine...until they come out Deep Frtiz 8 (multi-prossecor) is the strongest for multi-proccesor home based systems.... ive played amyng ames with F9 agsint DF8 and its even

Author:	MysticVegeta [ Wed Jan 18, 2006 11:05 pm ]
Post subject:
But the intention of my post was the level it was at that far back at time... but poeple sitll aurge taht hmuans are bteter tahn cmoupters [/dan]

Author:	codemage [ Thu Jan 19, 2006 12:52 pm ]
Post subject:
Only the highest chess-playing percentile of humans. All but the weakest chess software can beat the masses.

Author:	[Gandalf] [ Thu Jan 19, 2006 5:12 pm ]
Post subject:
There are also some people who train specifically to beat computers. One guy was rated 1800 real life, and he beat computers rated around 2500+. I didn't know Fritz 9 was out, but I do have Deep Fritz 8. As I recall, last time I went to the computer vs computer section, there was something (possibly Hiarcs?) which was usually beating other people's Deep Fritz's.

Author:	MysticVegeta [ Thu Jan 19, 2006 8:41 pm ]
Post subject:
I only have chessmaster 9000, its rated at around 2900. It was interesting to make computer vs computer, the games were draw.

Author:	Hikaru79 [ Thu Jan 19, 2006 11:59 pm ]
Post subject:
MysticVegeta wrote: Hikaru79 wrote: MysticVegeta wrote: Hikaru, you have the documentary in what? Video or book? It's a video. If there's a book out there, I'd be interested to hear about it. dam I would like to watch it, where did you buy it? ahem, cough, Can you copy and send it to me. LOL. [/idiotic question] Come to the IRC channel. A few people have already had their wish granted, if you know what I mean. nudge nudge wink wink

Author:	[Gandalf] [ Sun Jan 22, 2006 4:22 pm ]
Post subject:
MysticVegeta, the ratings that Chessmaster gives itself are probably inflated. Also, I'm fairly certain that it uses a different rating system (ie. not ELO). Chessmaster is a good program though, since it offers a lot of flexibility on playing strength. I made a "personality" on CM10000 that is stronger than "Chessmaster" in short games, and a bit weaker on longer games.

Author:	cool dude [ Sun Jan 22, 2006 7:42 pm ]
Post subject:
i was thinking of programming chess on turing or VB in the summer. i just don't know where to start on my algorithms. does anyone have any hints. P.S. i got Kasparov's chess computer. it cost a shit load! i learn quite a lot of new openings and its awsome. unfortunately, so far i only manages to beat level 1 intermediate. Advanced is crazy. it says that it looks 8 moves ahead which is pretty good, considering average players like most of us think only 4 to 5 moves ahead.

Author:	blaster009 [ Mon Jan 23, 2006 1:53 am ]
Post subject:
I've got that chess computer too ...I got it for my birthday waaay long ago. It's freaking awesome. WARNING: CHESS PROGRAMMING SPOILERS The following description is very specific...If you wish to figure it out for yourself, I suggest you skip ahead ---BEGIN SPOILERS--- Anyhow, if you're gonna be making a chess game, start simple with the board, evidently. An 8 * 8 multidimensional array of strings should be great. Then, store "pwnblk" or "qwnwht" in each of the squares, and use "null" as an empty square. If you keep the names of your pieces to 6 characters, you can easily use the last three to identify piece ownership, and thus control where units can/can't move (only to a square with "null" or the opposite colour's piece). GUI-wise, draw a grid of tiles (50x50 or w/e) and then divide the user's click position on the screen by the tile size to find out which tile the user's clicking on...For piece movement, it's a simple matter of making the sure the user is moving their own piece, and the tile they're moving it to mathematically matches the movement patterns of the units (slope comes in really handy here...m = [y2-y1]/[x2-x1]). As far as AI goes, that's a little bit trickier. My personal strategy would have the computer examine possibilities, assign points to each one, and then choose the play with the highest point value. To avoid taking forever, it would force a move if the computer takes over 10-20 seconds. ---END SPOILERS--- Anyhow, that's my take on chess. As you can see, I haven't given programming a chess game any thought at all. Lol. Oh, and I think there should be a spoiler HTML tag that hides certain portions until you click on it. You could use it for forum quizzes and such.

Author:	MysticVegeta [ Mon Jan 23, 2006 11:14 am ]
Post subject:
blaster009 wrote: As far as AI goes, that's a little bit trickier. My personal strategy would have the computer examine possibilities, assign points to each one, and then choose the play with the highest point value. To avoid taking forever, it would force a move if the computer takes over 10-20 seconds. I am in process of making one and I gotta tell ya, its a lot more than that; When you are saying assign points to each one, you are talking about algos people already made, they make life really easier, so what I would suggest, is first brainstorm what you are gonna implement in the AI then go to google and search up AI algorithms that could be used for chess. Offtopic: No periods ^

Author:	blaster009 [ Mon Jan 23, 2006 5:00 pm ]
Post subject:
Lol...I said I was making AI. I didn't say I was making SMART AI.

Author:	MysticVegeta [ Mon Jan 23, 2006 5:25 pm ]
Post subject:
lol. Why not make it smart?

Author:	MihaiG [ Mon Jan 23, 2006 10:00 pm ]
Post subject:
hiracs is good...but it also depends on ur how much ram you give to run ... proccesor speed...and the oepning book.. personaly i use the default one that came with F9 + F8 there other ones u can buy from chessbase but there usualy expensive and really big!>.....sometimes a dvd! but ive played hiracs with f9...not to good...F9 creamed it with a two pawns in the end game... (Free client)you can download that chess server and play ur engines...its compatible with all fritz programs,hiracs,junior and CM(needs a special 3rd party loader ) unfortunetly i dont reccomend using Chessmaster...itsa horrible engine.... chess ai...would take you quite a while....F9 is over 500,000 for the engine.....your best chance is to find an opening book that can go deep(15-20 moves).. and then try to make a brute forcing program that searches (2-3 moves ahead) ....otherwise i dont reccomend it

Author:	MysticVegeta [ Mon Jan 23, 2006 10:50 pm ]
Post subject:
Game trees algos are so efficient. Really good,work like a charm.

Author:	MihaiG [ Fri Jan 27, 2006 9:00 pm ]
Post subject:
heres a documentary form the computer museum at google videos about deep blue video

Author:	[Gandalf] [ Fri Jan 27, 2006 9:26 pm ]
Post subject:
Argh! What's with everyone making a chess game now? I'm making one too. And here I was hoping for some originality (only chess game I've seen done in Turing is bugz's, and that was far from being finished), still I'm well off for the moment and will have more time to finish after exams. blaster009, you're methods hardly cover everything. For a simple example, you don't have anything on not being able to skip over other pieces (I am in the process of solving this with my game), and there are countless other methods for doing what you describe. An AI will be the last part for me, and it will probably be limited to depth 2 or 3 because of Turings amazing speeds and time restrictions. Also, el commander, not every chess engine has to be as good as Fritz. As you said, Chessmaster has inferior playing strength and yet has very good sales. For recreational purposes, a depth 5/6 program could outplay a 'hobby' player easily. Opening books are a good idea, but take some time to implement (with move recognition, searching, and all that), so they can go after finishing a decent AI. Really, creating the whole chess game (interface, move restrictions, everything like that) and creating the AI are two completely different aspects and both take time.

Author:	MysticVegeta [ Sat Jan 28, 2006 12:55 pm ]
Post subject:
omg wtf? I am making Chess AI in turing and so is zylum. How many people are making??? this is insane..

Author:	cool dude [ Sat Jan 28, 2006 2:01 pm ]
Post subject:
we should have a contest on chess!

Author:	Hikaru79 [ Sat Jan 28, 2006 2:42 pm ]
Post subject:
cool dude wrote: we should have a contest on chess! I proposed that already, on this thread right here, but nobody really seemed too enthusiastic. Maybe there's enough interest now to revive this plan?

Author:	MihaiG [ Sat Jan 28, 2006 3:10 pm ]
Post subject:
well i do have over 300mb in opening books....

Author:	zylum [ Sat Jan 28, 2006 3:10 pm ]
Post subject:
i like the idea although i still think its a bit too much work. maybe we should first do a preliminary contest such as connect 4 or checkers.

Author:	MysticVegeta [ Sat Jan 28, 2006 4:09 pm ]
Post subject:
I wanted to be a unique one to do AI in turing... but.. some dreams get shattered..

Author:	Hikaru79 [ Sat Jan 28, 2006 5:28 pm ]
Post subject:
MysticVegeta wrote: I wanted to be a unique one to do AI in turing... but.. some dreams get shattered.. You can still be very unique. There's a dozen strong engines out there commercially, and yet they're each very unique in their approach. This is a giant problem, there's a trillion different ways to approach it. There's LOTS of room to be original. Don't be discouraged

Author:	[Gandalf] [ Sat Jan 28, 2006 7:56 pm ]
Post subject:
El Comandante wrote: well i do have over 300mb in opening books.... Good luck choosing a good one from those. I have 2 GB+ zylum wrote: i like the idea although i still think its a bit too much work. maybe we should first do a preliminary contest such as connect 4 or checkers. Indeed, that's why I'm making the framework (everything but the ai) first, and the ai will come last. This is why I might have a problem entering an AI contest on time, can't spend all my time on it. For connect 4 or checkers it would be quite a bit easier, although I would probably do it just using text, and would have to think of some strategies to add. Another suggestion is Othello/Reversi. MysticVegeta wrote: I wanted to be a unique one to do AI in turing... but.. some dreams get shattered.. Don't feel defeated yet. We still have yet to see who will be successful in making a working AI.

Author:	MysticVegeta [ Sun Jan 29, 2006 1:05 am ]
Post subject:
[Gandalf] wrote: We still have yet to see who will be successful in making a working AI. Yeah like that isn't obvious.

Author:	MihaiG [ Mon Jan 30, 2006 10:21 am ]
Post subject:
heh....good luck with that..1 person making an a engine.. impossible(or nearly to it) my suggestion is open a sweatshop.... put it ini the back of a tractor trailer and always keep moving...so they wont find you

Author:	zylum [ Mon Jan 30, 2006 11:21 am ]
Post subject:
MysticVegeta wrote: [Gandalf] wrote: We still have yet to see who will be successful in making a working AI. Yeah like that isn't obvious. my friend once made a chess ai in turing... it was alright for turing i guess, nothing too special

:

You can syndicate this boards posts using the file backend.php or view the topic map using sitemap.php.

Terms of Use | Privacy Policy