10th aniversery of deep blue win: interesting news article

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Terry McCracken

Re: 10th aniversery of deep blue win: interesting news artic

Post by Terry McCracken » Fri May 11, 2007 5:45 am

Leto wrote:
Terry said:
Some engines have found Deep Blues moves, not faster than Deep Blue.
False. Game 1, after 36.Kg1 :
[D]3rr1k1/1p6/2p3P1/4nP1p/pP2p3/P1B1NbP1/2P3B1/5RK1 b - - 0 36

[log] 8(4) #[Rd7](-30)[Rd7](-30) -30v T=0
Rd8d7 rf1f2 Pb7b5 bg2f3B Pe4f3b bc3e5N Re8e5b rf2f3P Kg8g7 pg3g4 Ph5h4 kg1g2
8(6) #[Rd7](-64)############################# -64 T=4
Rd8d7 rf1f2 Pb7b5 bg2f3B Pe4f3b bc3e5N Re8e5b rf2f3P Kg8g7 pg3g4 Ph5h4 kg1g2
9(6) #[Rd7](-53)###########################[b5](-48)# -48 T=13
Pb7b5 rf1f2 Rd8d6 bc3e5N Re8e5b bg2f3B Pe4f3b rf2f3P Kg8g7 kg1f1 Rd6d2 rf3f2 Rd2c2p
10(6) #[b5](-40)############################# -40 T=39
Pb7b5 rf1f2 Rd8d6 bc3e5N Re8e5b bg2f3B Pe4f3b rf2f3P Kg8g7 kg1f1 Rd6d2 rf3f2 Rd2c2p
11(6)<ch> 'kg1'
#[b5](-38)####################[TIMEOUT] 566 sec]#[Kf8](-36) -36 T=209
Kg8f8 bg2f3B Pe4f3b bc3e5N Re8e5b rf1f3P Kf8g7 kg1f2 Rd8d2 kf2f1 Pb7b5 rf3f2 Rd2f2r kf1f2R Kg7f6 [/log]

According to the log, it took Deep Blue 209 seconds to see that 36...Kf8 is strongest.

Rybka on the other hand on my two year old computer takes only 138 seconds to see that 36...Kf8 is strongest :

Analysis by Rybka 2.2 mp x64 2CPU:
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (-0.10) Depth: 2 00:00:00
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (-0.10) Depth: 3 00:00:00
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 4 00:00:00
36...Rd7 37.Bxf3
= (0.00) Depth: 4 00:00:00
36...Rd7 37.Bxf3
= (0.00) Depth: 5 00:00:00
36...Rd7 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2
= (0.10) Depth: 6 00:00:00 2kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 6 00:00:00 3kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 7 00:00:00 4kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 8 00:00:00 4kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 9 00:00:00 4kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 10 00:00:01 4kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 11 00:00:01 4kN
36...b5 37.Bxf3 exf3 38.Bxe5 Rxe5 39.Kf2 c5 40.bxc5 Rxc5 41.Rb1 Re5 42.Re1
= (0.00) Depth: 12 00:00:01 9kN
36...b5 37.Re1 Ng4 38.f6 Kf8 39.Nf5 Bxg2 40.Kxg2 Re6 41.g7+ Kf7 42.Ne7
= (0.00) Depth: 13 00:00:01 215kN
36...b5 37.Re1 Ng4 38.f6 Kf8 39.Nxg4 hxg4 40.Bf1 Bd1 41.Kf2 e3+ 42.Rxe3 Bxc2
= (-0.04) Depth: 14 00:00:03 624kN
36...b5 37.Re1 Ng4 38.f6 Kf8 39.Nxg4 hxg4 40.Bf1 Bd1 41.Kf2 e3+ 42.Rxe3 Bxc2 43.g7+
= (0.00) Depth: 15 00:00:04 810kN
36...b5 37.Re1 Ng4 38.f6 Kf8 39.Nxg4 hxg4 40.Bf1 Bd1 41.Ba1 Bf3 42.Bc3 Bd1 43.Ba1
= (0.00) Depth: 16 00:00:05 1244kN
36...b5 37.Re1 Kf8 38.Bh3 Kg7 39.g4 h4 40.Kf2 Kg8 41.g5 Kf8 42.Ng2 Nc4 43.Ne3
= (0.00) Depth: 17 00:00:10 2382kN
36...b5 37.Re1 Kf8 38.Bh3 Kg7 39.g4 h4 40.Kf2 Kg8 41.g5 Kf8 42.Ra1 Rd6 43.Re1
= (0.00) Depth: 18 00:00:15 3532kN
36...b5 37.Kf2 Kf8 38.Bxf3 exf3 39.Rh1 Ng4+ 40.Nxg4 hxg4 41.Rh8+ Ke7 42.Rh4 Kf8 43.Rxg4
² (0.39) Depth: 19 00:00:37 8742kN
36...Ng4 37.f6 Re6 38.Bh3 Rxf6 39.Bxg4 Rxg6 40.Bh3 Rxg3+ 41.Kh2 Rg5 42.Be6+ Kh7 43.Bf6
= (0.24) Depth: 19 00:01:43 23779kN
36...Kf8 37.Bh3 b5 38.Kf2 Rd6 39.Re1 Rdd8 40.Kg1 Kg7 41.g4 h4 42.Kf2 Kg8 43.g5
= (0.00) Depth: 19 00:02:18 31798kN
36...Kf8 37.Bh3 b5 38.Kf2 Rd6 39.Re1 Rdd8 40.Kg1 Kg7 41.g4 h4 42.Kf2 Kg8 43.g5
= (0.00) Depth: 20 00:02:55 40671kN
36...Kf8 37.Bh3 b5 38.Kf2 Rd6 39.Re1 Re7 40.Bf1 Ng4+ 41.Nxg4 Bxg4 42.f6 Ree6 43.Ke3
= (0.02) Depth: 21 00:04:15 59533kN
36...Kf8 37.Bh3 b5 38.Kf2 Rd6 39.Re1 Rdd8 40.Ng2 Bd1 41.Nf4 Bxc2 42.Bxe5 Rxe5 43.Ne6+
= (0.04) Depth: 22 00:07:15 102mN
36...Kf8 37.Bh3 b5 38.Kf2 Rd6 39.Re1 Rdd8 40.Ng2 Bd1 41.Nf4 Bxc2 42.Bxe5 Rxe5 43.Ne6+
= (0.10) Depth: 23 00:10:34 148mN
:roll:

Tony Thomas

Re: 10th aniversery of deep blue win: interesting news artic

Post by Tony Thomas » Fri May 11, 2007 6:31 am

mschribr wrote:
bob wrote:
mschribr wrote:
bob wrote:The thing that most amazes me is that supposedly Kasparov claims he has never been able to look at the deep blue logs. Even though they were (and still are) available on the web. I could email 'em to him if he would shut that claim up. :)
Those logs have a problem. The logs on the ibm web site have no independent certification as being created by deep blue during the match. For the logs to be valid the judges would need to see them being created by deep blue during the game. The logs need to given to the judges during the match. The judges did not verify the logs on the ibm web site.
Sorry, but that's wrong. The logs _were_ verified _during_ the match. I even saw a copy of the key position's log emailed to me by Ken Thompson within 24 hours of the "fiasco" happening.

You seem to demand that they disprove a negative, which is impossible. There are way too many ways to actually pull off cheating without detection. But for every move DB found in that match that was questioned by Kasparov, other programs have since shown that the moves were not surprising at all and are reproducible by several different programs.

But the logs are available, they match the games, there's no unusual aspect to them whatsoever...
Bob, you were not an arbiter and Ken Thompson was not an arbiter. So neither of you could verify that logs were correct. Only the arbiter, Carol Jarecki, could verify the logs. This was not done. The logs should have been created in front of the arbiter and given to the arbiter after each game for same keeping.

The question is not if other chess programs could find the move. The question is did deep blue find the moves and could it do find them again. The question could easily be answered by having deep blue replay the games. Simple, setup deep blue and see if could replay the same moves. Oh, I forgot we couldn’t do that. Ibm dismantled deep blue so it could never play again. So it looks like we will never know if deep blue really played the moves.
People are always proposing conspiracy theories weather or not you believe in them its your choice. I say the invisible Man played all the moves against Kasparov.

Terry McCracken

Re: 10th aniversery of deep blue win: interesting news artic

Post by Terry McCracken » Fri May 11, 2007 1:43 pm

Leto wrote:
Terry McCracken wrote:
Leto wrote:
Terry McCracken wrote:
Leto wrote:
Terry McCracken wrote:
Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Too bad we can't put that to the test...the results might surprise you!
Who says we can't put that to the test? Even in Deep Blue's prime there were hundreds of computer systems more powerful than it. Today there are computer systems that are billlions of times more powerful than Deep Blue was.

Hydra, which calculates just as fast as Deep Blue did, is arguably stronger, and look how well it has been doing lately against the top engines (not well at all).

If you have a top chess engine, you can analyse all of Deep Blue's moves and you'll see that the engine finds the moves quicker than Deep Blue does, and can even produce stronger moves. Logically, today's top engines are clearly stronger than Deep Blue was.
What a load!

First, there are NOT computers billions of times faster than Deep Blue, and there may not be for a long time to come.

Hydra can't match Deep Blues speed. I've no idea where you're getting your information from....the top of your head, I suppose? :roll:

Hydra is in the closet so you can't test it, and it's playing no one or no machine either as of late.

Some engines have found Deep Blues moves, not faster than Deep Blue. How would you know if it found stronger moves than Deep Blue?

Sometimes I don't know why I bother answering, it doesn't do any good, and facts seem to escape most people :roll:
You want facts? Here's some facts. The Deep Blue version that beat Garry Kasparov in 1997 only evaluated 200 million positions per second (Hydra is also able to evaluate 200 million positions per second), and had a calculating speed of 11.38 gigaflops. That version of Deep Blue was only ranked the 259th most powerful computer that year.

The most powerful computer of that year was Intel's ASCI Red, which had a calculation speed of 1.06 teraflops, that's almost a hundred times more powerful than Deep Blue.

In November 2006 the most powerful computer was IBM's BlueGene, which has a calculation speed of 280.60 teraflops. That's over 24 thousand times more powerful than Deep Blue was.

So I exaggerated 'a bit' about there being a computer billions of times more powerful than DB, but as you can see Deep Blue wasn't anything special.
You forgot to take into account the the hundreds of special purpose chess chips which add to the speed significantly.

Bob Hyatt can give you details Wikipedia can't.....IBM won't...

You haven't said anything I don't already know.
You claimed that Hydra can't match Deep Blue's speed. Hydra evaluated 200 million positions a second, just like Deep Blue!
Deep Blue could hit a billion positions per sec. in some cases.

You're an expert on Deep Blue?

User avatar
Leto
Posts: 2028
Joined: Thu May 04, 2006 1:40 am
Location: Dune

Re: 10th aniversery of deep blue win: interesting news artic

Post by Leto » Fri May 11, 2007 5:27 pm

Terry McCracken wrote:
Deep Blue could hit a billion positions per sec. in some cases.

You're an expert on Deep Blue?
Deep Blue's average was 200 million per sec, that's the same as Hydra's average. Just admit you lost this argument, it's not going to be the end of the world.

No I'm not an expert on Deep Blue, and neither are you.

bob
Posts: 20555
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: 10th aniversery of deep blue win: interesting news artic

Post by bob » Fri May 11, 2007 5:32 pm

Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Your logic escapes me. It was strong enough, 10 years ago, to defeat Kasparov in his prime. But it isn't very strong by today's standards?

I must be missing something...

bob
Posts: 20555
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: 10th aniversery of deep blue win: interesting news artic

Post by bob » Fri May 11, 2007 5:35 pm

Leto wrote:
Terry McCracken wrote:
Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Too bad we can't put that to the test...the results might surprise you!
Who says we can't put that to the test? Even in Deep Blue's prime there were hundreds of computer systems more powerful than it. Today there are computer systems that are billlions of times more powerful than Deep Blue was.

Hydra, which calculates just as fast as Deep Blue did, is arguably stronger, and look how well it has been doing lately against the top engines (not well at all).

If you have a top chess engine, you can analyse all of Deep Blue's moves and you'll see that the engine finds the moves quicker than Deep Blue does, and can even produce stronger moves. Logically, today's top engines are clearly stronger than Deep Blue was.
Can we come back to planet earth? In 1997 there were no "systems" hundreds of times faster than Deep Blue, with respect to chess. There was nothing even remotely close to their speed, because of the special purpose hardware. The IBM SP was certainly fast in its own right, but the chess processors are what made deep blue, and there was nothing close to that then. And even today it is very tough to get 200M nodes per second out of any existing hardware, although it is certainly possible.

And there are absolutely _NO_ systems "billions of times faster". That is pure fantasy...

bob
Posts: 20555
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: 10th aniversery of deep blue win: interesting news artic

Post by bob » Fri May 11, 2007 5:38 pm

Leto wrote:
Terry McCracken wrote:
Leto wrote:
Terry McCracken wrote:
Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Too bad we can't put that to the test...the results might surprise you!
Who says we can't put that to the test? Even in Deep Blue's prime there were hundreds of computer systems more powerful than it. Today there are computer systems that are billlions of times more powerful than Deep Blue was.

Hydra, which calculates just as fast as Deep Blue did, is arguably stronger, and look how well it has been doing lately against the top engines (not well at all).
You are not helping your case when you make an argument that is full of flaws. The "deep blue system" was _not_ ranked "259 most powerful machine." That was _only_ the SP part of the system. It says _nothing_ about the actual strength of deep blue, which was the 480 chess-specific processors that gave it its true "power"

Please, try to at least use arguments that have some technical merit, rather than posting rambling stuff...

If you have a top chess engine, you can analyse all of Deep Blue's moves and you'll see that the engine finds the moves quicker than Deep Blue does, and can even produce stronger moves. Logically, today's top engines are clearly stronger than Deep Blue was.
What a load!

First, there are NOT computers billions of times faster than Deep Blue, and there may not be for a long time to come.

Hydra can't match Deep Blues speed. I've no idea where you're getting your information from....the top of your head, I suppose? :roll:

Hydra is in the closet so you can't test it, and it's playing no one or no machine either as of late.

Some engines have found Deep Blues moves, not faster than Deep Blue. How would you know if it found stronger moves than Deep Blue?

Sometimes I don't know why I bother answering, it doesn't do any good, and facts seem to escape most people :roll:
You want facts? Here's some facts. The Deep Blue version that beat Garry Kasparov in 1997 only evaluated 200 million positions per second (Hydra is also able to evaluate 200 million positions per second), and had a calculating speed of 11.38 gigaflops. That version of Deep Blue was only ranked the 259th most powerful computer that year.

The most powerful computer of that year was Intel's ASCI Red, which had a calculation speed of 1.06 teraflops, that's almost a hundred times more powerful than Deep Blue.

In November 2006 the most powerful computer was IBM's BlueGene, which has a calculation speed of 280.60 teraflops. That's over 24 thousand times more powerful than Deep Blue was.

So I exaggerated 'a bit' about there being a computer billions of times more powerful than DB, but as you can see Deep Blue wasn't anything special.

User avatar
Leto
Posts: 2028
Joined: Thu May 04, 2006 1:40 am
Location: Dune

Re: 10th aniversery of deep blue win: interesting news artic

Post by Leto » Fri May 11, 2007 5:39 pm

bob wrote:
Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Your logic escapes me. It was strong enough, 10 years ago, to defeat Kasparov in his prime. But it isn't very strong by today's standards?

I must be missing something...
Kasparov played horrible chess in that match, he was psychologically defeated. Deep Blue's strength could have been 2650, perhaps 2700, but in Kasparov's mind Deep Blue was either 3000 or some strong Grandmaster must have been interfering.

bob
Posts: 20555
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: 10th aniversery of deep blue win: interesting news artic

Post by bob » Fri May 11, 2007 5:41 pm

Leto wrote:
bob wrote:
Leto wrote:
bob wrote:
Leto wrote:It's a foregone conclusion that Deep Blue wasn't very strong, clearly weaker than today's top engines, and that Kasparov could have played much better. Psychology definitely played a large role in that match.
"wasn't very strong" yet it blew away every program 10 years ago. Sort of a "self-contradiction" IMHO. It was far from "not very strong"...
Maybe strong a decade ago, but it certainly isn't "very strong" today. You can use today's engines to see that they find Deep Blue's move in a fraction of the time on ordinary hardware, and they can even find stronger moves.
Your logic escapes me. It was strong enough, 10 years ago, to defeat Kasparov in his prime. But it isn't very strong by today's standards?

I must be missing something...
Kasparov played horrible chess in that match, he was psychologically defeated. Deep Blue's strength could have been 2650, perhaps 2700, but in Kasparov's mind Deep Blue was 3000.
ah yes. When all else fails, let's place the blame on the loser, not on the victor...

Did you see how awfully Kramnik played in the last computer vs human match? Disgraceful.

bob
Posts: 20555
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: 10th aniversery of deep blue win: interesting news artic

Post by bob » Fri May 11, 2007 5:45 pm

mschribr wrote:
bob wrote:
mschribr wrote:
bob wrote:The thing that most amazes me is that supposedly Kasparov claims he has never been able to look at the deep blue logs. Even though they were (and still are) available on the web. I could email 'em to him if he would shut that claim up. :)
Those logs have a problem. The logs on the ibm web site have no independent certification as being created by deep blue during the match. For the logs to be valid the judges would need to see them being created by deep blue during the game. The logs need to given to the judges during the match. The judges did not verify the logs on the ibm web site.
Sorry, but that's wrong. The logs _were_ verified _during_ the match. I even saw a copy of the key position's log emailed to me by Ken Thompson within 24 hours of the "fiasco" happening.

You seem to demand that they disprove a negative, which is impossible. There are way too many ways to actually pull off cheating without detection. But for every move DB found in that match that was questioned by Kasparov, other programs have since shown that the moves were not surprising at all and are reproducible by several different programs.

But the logs are available, they match the games, there's no unusual aspect to them whatsoever...
Bob, you were not an arbiter and Ken Thompson was not an arbiter. So neither of you could verify that logs were correct. Only the arbiter, Carol Jarecki, could verify the logs. This was not done. The logs should have been created in front of the arbiter and given to the arbiter after each game for same keeping.

The question is not if other chess programs could find the move. The question is did deep blue find the moves and could it do find them again. The question could easily be answered by having deep blue replay the games. Simple, setup deep blue and see if could replay the same moves. Oh, I forgot we couldn’t do that. Ibm dismantled deep blue so it could never play again. So it looks like we will never know if deep blue really played the moves.
If you go back and look again, you will find that Ken, Monty, Carol, were _all_ on the committee responsible for handling these kinds of problems. Ken did watch part of the games watching the remote operators screen, as this was a group of "friends". I probably would have been back there myself had I been present. Ken had a copy of the log in question the day the protest was made. I personally looked at it. Amir Ban received excerpts from the log (pertaining to the move questioned at the time) within a day or two. They were not "hidden".

It does seem ridiculous to claim that they were doctored, and demand proof they were not, when it is _impossible_ to prove they were not altered. No way to prove something did not happen.

This is just an excuse. And a bad one.

Post Reply