DaviddesJ
Deity
Here's proof that you're wrong:
http://en.wikipedia.org/wiki/Deep_Blue_-_Kasparov,_1997,_Game_6
As it happens, I was in the audience for this game.
Soren Johnson: The Chick Parabola
- Thread starter alexman
- Start date
Here's proof that you're wrong:
http://en.wikipedia.org/wiki/Deep_Blue_-_Kasparov,_1997,_Game_6
As it happens, I was in the audience for this game.
MadRat
Cheese Raider
I think we also have to look beyond the opening - with data capacity so high today the databases can incorporate full games not just openings. GMs and high level players dont just look at openings they look at full games. I have met a few GMs and they eat things like Informants and other chess publications of major tourneys like candy (or have staff that do so for meta-analysis).
This was Kasaparov's big beef - Deep blue was loaded with a DB with almost all of his games so it essentially was programmed to "beat Kasparov" not simply just to play chess well in general. IN their defense GMs do this as well - they analyze their opponents for the major tourneys and interzonals much as boxers and other atheletes do.
To be fair openings do help novices avoid simple blunders. You know I havent played much chess in years - I forgot most of my openings and started falling for many of the trap variants I forgot. The problem with this style of play is that many amateur players dont really THINK about openings much and where they are headed and just like PCs stumble badly in the mid and end game transitions. The best thing to teach players first are endgames not openings.
Rat
This was Kasaparov's big beef - Deep blue was loaded with a DB with almost all of his games so it essentially was programmed to "beat Kasparov" not simply just to play chess well in general. IN their defense GMs do this as well - they analyze their opponents for the major tourneys and interzonals much as boxers and other atheletes do.
To be fair openings do help novices avoid simple blunders. You know I havent played much chess in years - I forgot most of my openings and started falling for many of the trap variants I forgot. The problem with this style of play is that many amateur players dont really THINK about openings much and where they are headed and just like PCs stumble badly in the mid and end game transitions. The best thing to teach players first are endgames not openings.
Rat
I don't think that is a good example to prove what I have stated.
The Russian GM's no5 move in that match is arguably a mistake, right?
He just made a move that he believe is the best against the program. (I suppose it is based on his own study and could be done with a reason only the GM himself knows). i.e. he has his own reason to do that in that match. You can't claim that move is a openning mistake simply because at the end he loss the match.
DaviddesJ
Deity
I wouldn't call move 5 a mistake. It's a line many grandmasters choose. Black's move 7 is a mistake, even though it is not "silly". Kasparov makes the mistake because he has a finite human memory and forgets, in the heat of the moment, a specific refutation that is in opening books. Deep Blue takes advantage of his lapse because it has an effectively infinite computer memory and follows the refutation that's in its opening books. This game precisely illustrates how opening books are important, how GMs and computers do use them to reach stronger rather than weaker positions, and how having a larger and deeper opening book is a real advantage.
You claimed it is impossible for a grandmaster to make a mistake and play a bad opening, but this game proves you are wrong.
Jolly Rogerer
Prince
- Joined
- May 16, 2008
- Messages
- 402
Good point, but this cuts both ways too. The internet added a lot to my enjoyment of SMAC, Civ4, MoM, Morrowind, MOO2 etc. because these games had so many vectors to experiment with that I would not have tried if others hadn't opened my eyes to them. And of course with some of them new content via mods.
Ah! I am sorry, wrong figure, I mean move 7 not 5 in my last post.
No. I am not wrong! Particularly when you intend to prove that according to the example you have quoted.
Remember, it is the article that said Kasparov has made a mistake in choosing that move, did Kasparov himself admit that? I repeat, that move is arguably a wrong move!
1. The wrong move 7th (as claimed by you) is not WRONG. It is simply not the best move because there is a better move (to replace it) and according to the article, the GM has play with that better move in two other matches earlier.
2. The reason why I say it is not an absolute bad openning move is, if play against another program (not deep blue) or when play against a human player, Kasparov might still win at the end. Further more, as you can find in the article, Kasparov made further mistake after the "openning phase" (that is when deep blue no more using an openning library), that again means, move no7. is arguably the sole reason that cause him to loss.
3. Let me ask this, say if Kasparov win the match at the end simply because Deep Blue is weak after it start thinking instead of using the openning library, would you still call that move a WRONG move?
4. My point is, you can only claim a move is wrong in the openning stage, if it will absolutely lead you to a lossing end and there isn't any chance at all for you to change the result if your opponent follow strictly a serial of known moves. In your example, that match is obviously not one that fulfill this condition.
I don't mean to insist on "I am not wrong"
It is obvious, openning library only serves "to save time" (for few initial moves) which otherwise a computer program will spend quite an amout of time yet still arrive at the same move (again and again in different matches when come to the same state). It is definitely not a crucial part that make a good AI. It doesn't even need logical calculation.
You and the other MR. are over-emphasizing its role! that is something I can't keep myself quiet for you are misleading others...
Hail
Satan's minion
reading comprehension fail
DaviddesJ
Deity
You're just wrong. You really don't know much about chess, I'm afraid.
Well
I dare not say I am a chess grandmaster. But I am serious in my attempt to correct your wrong emphasize of the role and importance of openning library (of a chess program), particularly when the program need to defeat a human GM.
I got an idea to illustrate my point better: imagine Deep Blue fight against Kasparov again tomorrow, after it finish the openning phase you take over, brother, 99.99% you will loss to Kasparov (irregardless of how is the state at the point you take over).
I suppose the above is simple to imagine and understand right?
So do you now see which components is more important in a chess program?
DaviddesJ
Deity
No one said the opening book is the most important component of a chess program. Although it did win one game of a six game match that is the most important in computer chess history, so that empirical evidence suggests that it is quite important.
DaviddesJ
Deity
(Deleted)
Perplexity
Warlord
- Joined
- Nov 4, 2005
- Messages
- 103
The top programs would crush GMs today without using any opening book at all. It's not close.
I would like to repeat, that statement is misleading for it has over-emphasized the role of openning library. (you too!)
No offense, but just imagine it the other way round and you should see even better why:
Imagine you will be playing against Kasparov tomorrow. After the openning stage, just let Deep Blue take over. I suppose it won't make any different. (I mean Kasparov will still sweat to win or could be defeated)
Read again your favorite article, I remember it took less than a dozen moves (of Deep Blue) before the program moved out from openning phase and don't tell me if you play the openning you will lose 2 horses in less than 12 moves...
MadRat
Cheese Raider
I call BS on this. I haven't heard any evidence of this; if these so called "top programs" were beating High level players so regularly we would have heard about it. Without an opening book or DB library of games an AI will get creamed by most GMS and IMs.
BTW Deep Blue (some version) and Turbo Fritz are the top programs ATM I think.
Rat
Openning stage is simply a subset of moves out of all possible moves from the beginning to the end of a chess game.
Common sense tells:
Let say, either or both sides loss one or few tokens in openning stage, if a chess AI can start thinking from that point onwards and at the end beats a human GM, I don't see why it can't if it start thinking from the very beginning (it just need to take more time)
So, the below is simply bombastic
Common sense tells:
Let say, either or both sides loss one or few tokens in openning stage, if a chess AI can start thinking from that point onwards and at the end beats a human GM, I don't see why it can't if it start thinking from the very beginning (it just need to take more time)
So, the below is simply bombastic
MadRat
Cheese Raider
This is where the flaw in your argument is apparent - you are assuming the AI is near the caliber of a GM outside of the opening. This is often not the case. In my experience the point at which the AI consistently stumbles is in the mid game and late game transitions. Indeed, the difference between both AIs and amateur players and GMs is in their ability to navigate those transitions. The primary reason AIs can even get to those transition points without outright unfavourable positions is because of their database game/opening libraries.
Give yourself this experiment, most commercial chess programs you can shut off the opening book. Play or watch (AI vs AI) games with opening books on and off. I suspect you will notice substantial differences in the quality of the game. Without an opening book and opening such as 1 h4 a5 is just as valid as 1 e4 c5 without the context of opening theory. I am yet to see an AI actually (w/o an opening book/library) understand opening theory (or even end game theory for that matter). Hence a GM can see those opening weaknesses in a non-book opening and set up a strategy by which the AI will lose in the mid/end game.
Rat
Not true!
You only need a machine say with 4 to 6-cores (you may add additional cores if you want more kick) and with a recent Chess program you may beat GM most of the time in middle game. (that is outside openning phase)
In late game, it mostly depends on the time remain for the AI.
Nowadays, computer Chess AI normally will beat a GM given enough time for it to think, even at late stage.
You are the one who should give it a try (not me
) especially with recent Chess program.
MadRat
Cheese Raider
Look, I believe a modern chess program can beat a GM; they regularly do so. This isn't the issue at all - the question is are these programs purely algorithms or do they have game libraries. The central point we are discussing is the impact that databases have on the AIs play ability. My point is simple : an AI without a database library will not be able to beat a GM consistently.
Let me put it to you this way - a modern computer a with say 8 Gig memory can probably store database which would include every published game match in the last 20 years as well as an opening library. If a PC had a basic modern chess algorithm and this library how would this glorified expert system perform vs a GM as opposed to a pure algorithm?
My hypothesis is that a pure algorithm based chess program would not win consistently versus a GM. This is NOT the same as saying a computer cant beat a GM, which you seem to imply that I am saying.
Rat
Let me put it to you this way - a modern computer a with say 8 Gig memory can probably store database which would include every published game match in the last 20 years as well as an opening library. If a PC had a basic modern chess algorithm and this library how would this glorified expert system perform vs a GM as opposed to a pure algorithm?
My hypothesis is that a pure algorithm based chess program would not win consistently versus a GM. This is NOT the same as saying a computer cant beat a GM, which you seem to imply that I am saying.
Rat
DaviddesJ
Deity
What does any of this have to do with the Chick Parabola, again?
Computers can play chess better than humans. They can't play Civilization nearly as well as humans. Isn't that all that matters, at least for people who are designing civilization-type games?
Computers can play chess better than humans. They can't play Civilization nearly as well as humans. Isn't that all that matters, at least for people who are designing civilization-type games?
Similar threads
