games chess

Games Chess

Share and you will unmask!
Funny name, real matters
[ Sign up | Log in | Guest ] (beta)
olympio 34 ( +1 | -1 )
Blue Gene I would really like to see the kind of chess that a machine like Blue Gene would be capable of playing. It's somewhere around 7,000 times the computing power of Deep Blue. Clearly it would be the strongest chess playing entity on the planet and it's too bad the machine is predestined to fold proteins instead of push pawns.
macheide 101 ( +1 | -1 )
olympio Dear friend,

As a matter of fact, the main goal of the Kasparov- Deep Blue matches was to promote selling of this machine to big pharmaceutical companies and other giants that need that kind of procesing speed to work in vital tasks as to decipher how proteins are structured, among other things.

Kasparov played horrible. I couldn't even recognize who was playing against the machine,..., and lost his second match. Kasparov demanded another match (I'm sure Kasparov would win it), but IBM fellows was affraid of that and very proud the publicity that that last match gived to IBM so they refused.

I work with computers and don't know much about chess software. Today chess programs are almost invincible in blitz and rapid chess, but I trust in the best player's human hability to recognize subtelties of the positions and to evaluate them better than the best machines,..., at least for the time being.

regards,

Andrés
macheide
olympio 52 ( +1 | -1 )
deep blue had great hardware but really crappy software. software is so much stronger now that the simple 8 processor machine that played kramnik with deep fritz was considered about as strong as deep blue on it's 256 processors. now imagine a chess algorithm as sophisticated as fritz running on a machine with 1 million processors (blue gene) at this point I think it would prune far enough ahead to make up for 90% of the positional shortcomings that would have hurt it's performance before
macheide 108 ( +1 | -1 )
olympio Dear friend,

Your standpoint is very interesting. Some months ago I recived, by mail, an exemplar of of a very good magazine titled TWIC (I think that's the name), it was a promotional gift. In this magazine I founded an artistic composition by the great GM Vassily Smyslov. Smyslov is retired from the active play but he spend many time composing chess problems (I recommend you to read "Vassily Smyslov, Endgame Virtuoso" by himself).

The point is that in his newest compositions he tries to find positions where the human intuition, chess knowledge and experience can be superior to machine's brute force and evaluating algorithms.

I'm at work at this moment but tonight I'll expose this compositionin the problems and puzzles section.

If some of our GK fellows who have strong software test this problem in their machines it would be very interesting to see the outcome, solution time, etc. I solved it in less than a minute. But it could be an interesting way to compare how humans and machines are different in this respect.

Hopefuly you are interested.

Andrés
macheide
olympio 11 ( +1 | -1 )
interested I would be delighted to see these problems. Would be kind of you to post here and some of us can give it a shot.
anaxagoras 40 ( +1 | -1 )
An interesting point about computer play is that openings are by the book--a book composed by humans! There is yet to be a chess program that could play an adequate opening from scratch, that is, from the computational power it relies upon for analyzing middle and end game positions. Someday, someday.... until then, I fear chess programs are still infantile.
olympio 38 ( +1 | -1 )
blue gene
a typical computer does about 1 million nodes per second. that's to about 15 ply in 3 minutes.
since blue gene is as many times the processingp ower of the typical computer as the typical computer does nodes/sec it would do 30 ply in 3 minutes.

so 15 full moves fully force searched.. it's pretty likely that the opening moves would be fairly good
thumper 13 ( +1 | -1 )
Wait till they start using biological computers with neuron connections instead of silicon. AI is not that far off.
More: Chess
bogg 129 ( +1 | -1 )
olympio There is a maximum number of cpus that can be assigned to one task before the overhead of managing the cpus outweighs the speed gained by having subprocessors evaluate the positions. I don't know if anyone has determined what that number is but all that should be necessary is to have enough evalutors to return the position evaluations as fast as the main cpu can walk the tree. Any more cpus than that are mostly wasted computing power.

Or using a simpler design, one drive processor could generate a 4 ply deep tree, roughly 1 million nodes, and pass each position out to a seperate cpu and have each one process the position completely. That method would net 4 additional plys using 1 million cpus. Roughly 997600 of the cpus would be looking at trash!

Its no big deal though. If things keep going like they have been it will only take about 34 years for a single cpu to be about a million time faster than what we can get today. That will add about 7 full width ply to the search or using a program that does heavy pruning about 11 ply. Either way that should add about 1400 rating points making our futuristic desktop computer about a 4000 player.
olympio 7 ( +1 | -1 )
that That would be unbeatable at even correspondence time controls.
olympio 25 ( +1 | -1 )
bogg Btw, I read that Deep Blue was only using 30% of it's full processing power just because of a problem that is similar though not identical to the one you mentioned. Their problem was that networking protocals were not that efficient yet.
macheide 37 ( +1 | -1 )
Smyslov's composition was posted Dear olympio,

Yesterday night I posted Smyslov's composition in the Problems and Puzzles section.

I'm interested in your opinion and in that of our fellows that can test the composition with a powerful software.

As I said you yesterday, the solution is very easy to see for a human, but I'm intrigued to see what a powerful chess machine "thinks" about it.

regards,

Andrés
macheide
olympio 48 ( +1 | -1 )
I took a look It was indeed pretty obvious. And I doubt a program would find this to be a draw unless it had so much capacity to recognize this type of position as a draw that it significantly sacrificed it's other faculties. Has anyone done a test?
I would think the typical program would have to go all the way out to 50-move rule to see it's a draw and not a huge advantage for the side with the queen. And 50-move rule is just impractical for a program to reach through brute calculation.
olympio 48 ( +1 | -1 )
I took a look It was indeed pretty obvious. And I doubt a program would find this to be a draw unless it had so much capacity to recognize this type of position as a draw that it significantly sacrificed it's other faculties. Has anyone done a test?
I would think the typical program would have to go all the way out to 50-move rule to see it's a draw and not a huge advantage for the side with the queen. And 50-move rule is just impractical for a program to reach through brute calculation.
finduriel 52 ( +1 | -1 )
I ran the thing on my machine I have an AMD Athlon 1,4 GHz, Fritz 8 software and a 144MB Hash. Your predictions were right, the program didn't recognize the draw until I let it calculate for about two hours. Maybe a different engine would do better.

However, Fritz was also able to achieve the draw when we played the position out (although he might have resigned if I hadn't switched this option off). So I don't think Fritz's inability to recognize the draw as early as a human player harms him a great deal in practice.

Fin
macheide 27 ( +1 | -1 )
Thanks a lot olympio! Dear friend,

Thank you for taking the time to analyse this problem!. What I read is fantastic. Did I understand well? Does the machine was unable to give a definitive (correct, of course) diagnostic of this position?

Your friend,

Andrés
macheide
macheide 9 ( +1 | -1 )
finduriel Dear finduriel,

Thanks a lot for your time (2 hrs!), too.

Best regards,

Andrés
macheide
finduriel 18 ( +1 | -1 )
No problem You're welcome, Andrés! It wasn't really my time (I cooked and ate while I let the computer work for me), so that's no problem at all.
My program was definitely unable to assess the position correctly!

Best,

Fin
macheide 6 ( +1 | -1 )
finduriel Good point, my friend :))) Thanks anyway!

macheide