|
< Earlier Kibitzing · PAGE 7 OF 10 ·
Later Kibitzing> |
| Apr-12-06 | | LoFarkas: When the programmer says they forgot to programme Deep Blue for eternal checks, I smell something very fishy. |
|
| Apr-12-06 | | acirce: It was simply very difficult for any machine to see the whole perpetual check line since it is so long, first the check sequence starts only with 48..Qc1+ and then there is no immediate perp either since White's king can wander about for a while (visiting squares from e1 to h2!). There is simply nothing special at all about missing this. Simple as that. Today's strongest PC programs 9 years later don't find it even nearly as quickly as Deep Blue would have had to either, try it. DB was much ahead of its time when it came to chess "understanding", but this is just about calculation. It's discussed over and over at Deep Blue vs Kasparov, 1997 |
|
| Apr-12-06 | | Akavall: <acirce>
click for larger viewIn this posiotion Deep Blue played 23. Rec1; this is a very human-like move. My fritz for example doesn't even take it seriously. 
click for larger viewThis is more interesting. Deep Blue plays 26. f4! with the right plan, but after 26...Nf6 27. fxe5 dxe5 plays a slightly inaccurate 28. Qf1!? Don't you find it very strange how Deep blue found a right plan but had trouble executing it? This is very strange for computers; computers are known for not being able to find right plans, but when they do, they execute them flawlessly. Humans on the other hand, often know what the right plan is, but executing it is the hard part. In this position Deep Blue played like a human. |
|
| Apr-13-06 | | acirce: <Akavall> This kind of argumentation is not convincing at all. I don't even get why these examples should be even slightly suspect. You could just as well look at Topalov vs Mamedyarov, 2006 and pretend that the fact that Topalov went pawn-grabbing "like a computer" with 34.Rxa6 and 35.Nxb5 instead of using the "human" plan of 34.f4 and 35.f5 is evidence that he used computer help. Nonsense as well. I'm not even sure I agree with your premises. Computers don't "plan", they make the best move they can find in each position taken individually. Why can't it have decided on both 26.f4 and 28.Qf1 (the preferred move of Rybka 1.0 Beta on my computer)? What I also don't get is who was supposed to have the chess understanding needed to overrule Deep Blue's decisions with confidence? Random GM Joel Benjamin? Or did they hire in Karpov? For the record, entering the position in your first diagram my Fritz 9 gives many moves as being of roughly the same value. After 5 minutes 23.Rec1 (+0.81) is given as 0.08 pawns worse than the preferred move 23.Ra2 (+0.89). Rybka 1.0 Beta also comes to like it after a while. |
|
| Apr-13-06 | | Akavall: <acirce> I am basing my claim on the following assumption: " 26.f4 is the strongest move; 28.Qf1 is not the strongest move(an inaccuracy)". < '26.f4! Again, simply superb.''28.Qf1?! Funnily enough this inferior move looks very much like a human misstep>
http://chess.about.com/library/week...
If those are not true, than I have nothing. However, if they are I don't see how Deep Blue could play 26. f4, if the strongest continuation he saw was 28. Qf1. What it looks like to me is that somebody played 26. f4 for Deep Blue, but later either they messed up or Deep Blue "messed up". <Why can't it have decided on both 26.f4 and 28.Qf1 (the preferred move of Rybka 1.0 Beta on my computer)?> Does Rybka want to play 26. f4 and the 28. Qf1? |
|
| Apr-14-06 | | whatthefat: <Akavall: However, if they are I don't see how Deep Blue could play 26. f4, if the strongest continuation he saw was 28. Qf1. What it looks like to me is that somebody played 26. f4 for Deep Blue, but later either they messed up or Deep Blue "messed up".> From my fairly considerable amount of time analysing with Fritz 8, I'm not so surprised by this. The continuation that the computer prefers when it makes a move, has frequently altered a couple of moves down the line. It's just a horizon effect. What looked good at 14 ply two moves ago, doesn't always look so good at 14 ply when the new position is reached (so effective depth of 18 ply from the original position). I must say, I was under the impression that the evidence for Deep Blue having been human controlled was rather stronger than this. |
|
| Apr-14-06 | | Akavall: <whatthefat> In other words, Deep Blue played 26. f4! for the 'wrong' reasons? Sure possible, but the likelihood? |
|
| Apr-15-06 | | whatthefat: <Akavall: In other words, Deep Blue played 26. f4! for the 'wrong' reasons? Sure possible, but the likelihood?> I found that to be a hard question to answer, so I set Fritz 8 to the task for a few hours. If you have the patience(!), the findings were rather interesting. Starting at the position <after 25...Qd8>, Depth 12 ply: #1 is 26.Bd1 (+1.11); 26.f4 doesn't really rate. From here however, Fritz begins to devote considerable computing time to 26.f4, and it concurrently moves up the rankings. 13 ply: #1 axb5 (+1.12); #24 f4
14 ply: #1 axb5 (+1.12); #23 f4
15 ply: #1 Bd1 (+1.12); #19 f4
16 ply: #1 Bd1 (+1.09); #12 f4
17 ply: #1 Bd1 (+1.10); #6 f4 (+0.75)
At this point I moved on due to the computational time involved. You can see though that it's not unreasonable to believe 26.f4 could be rated #1 given some more computational time (i.e., a few more ply). The next point of interest, is what lines Fritz was looking at <in response to 26.f4>, at various depths, 13 ply: #1 is 26...Rbc8 (+0.70) and it remains this way throughout. Kasparov's 26...Nf6 is always close behind, but it can't be claimed that it's the only critical reply (in the computer's evaluation). Thus the point of whether 26.f4 was played with the text move 28.Qf1 'in mind' loses relevance to some degree. 14 ply: #1 Rcb8 (0.76); #4 Nf6 (0.87)
15 ply: #1 Rcb8 (0.70); #4 Nf6 (0.87)
16 ply: #1 Rcb8 (0.74); #3 Nf6 (0.87)
<After 26...Nf6>,
13 ply: #1 is 27.Qf2 (+0.86) and Fritz 8 consistently prefers this move. The move played by Deep Blue, 27.fxe5 is at this point rated nowhere (#36 of 42 possible moves!!). 14 ply: #1 Qf2 (0.89); #13 fxe5 (quite the sudden improvement!)
15 ply: #1 Qf2 (0.90); #34 fxe5 (the move slips back to nowhere)
16 ply: #1 Qf2 (0.86); with fxe5 again way down.
If anything, <this seems to be the point of contention> - why was 27.fxe5 played by Deep Blue? Fritz does not seem to like it (at the depths I'm looking at). Indeed, after <27.fxe5 dxe5>, the position is no longer judged very favourably for white. At depth 14, 28.a5 is considered best (+0.43), although closing the queenside with rooks doubled on the a-file doesn't look too dynamic to me. 28.Qf1 is rated around equality (+0.1) at this depth, so again, it could reasonably come back to the top with greater depth. How similar the Fritz 8 and Deep Blue evaluation functions are is hard for me to say. If they are reasonably similar, it's fair to conclude Deep Blue was selecting moves at an equivalent ply depth of at least about 20 - assuming no human intervention... :) |
|
| Apr-15-06 | | whatthefat: I can just add to the above analysis, that at depth 18, Fritz 8 now chooses 28.Kh2, scoring it at +0.50. This is still not great for white, so I don't see a convincing explanation for 27.fxe5. |
|
| Apr-15-06 | | Dionyseus: <whatthefat> Rybka 1.1 immediately claims 27.fxe5 is the strongest move. So far up to depth 18 it still claims it as the best move, evaluating it as +0.45 after 9m:36s I decided to try it with Shredder 9.1, and it immediately claimed 27.fxe5 is the strongest move. So far up to depth 17 it still claims it as the best move, evaluating it as +1.02 after 9m:50s I decided to check out what Crafty 19.01 had to say about the position. It also immediately claimed 27.fxe5 as the strongest move. So far up to depth 14 it still claims it as the best move, evaluating it as +1.00 after 2m:41s I even tried Comet B50 and it immediately claimed 27.fxe5 as the best move. After 27.fxe5 dxe5, Rybka 1.1 immediately claims 28.kh2 is the strongest move. At a depth of 13, it calculated 28.qf1 as the 6th strongest move, but the evaluation spread is thin, a difference of less than 2/10ths of a pawn. |
|
| Apr-15-06 | | whatthefat: <Dionyseus>
Very interesting, thanks for putting in the effort. I'd say when you combine our analysis, it pretty much refutes any allegations of human intervention in Game 2.It's also intriguing that Fritz 8 evaluated the position after 26...Nf6 quite differently to the other engines, but that's a side point. Still, makes me consider getting Rybka. |
|
| Apr-15-06 | | Dionyseus: <whatthefat> I'd strongly recommend getting Rybka 1.1, at least give the free demo a try. I put Rybka up in matches against my Shredder engines (8 and 9.1) and Rybka makes Shredder look like an amateur. On all the engine rating lists, Rybka is dominating the field by over 100 points, this is unheard of. If you have Playchess, go to the engines room and at least 90% of the engines are Rybka, it's unbelieavable! Still I believe Hydra is the stronger over the board player, but I think no one can doubt Rybka is the strongest engine ever created. |
|
| Apr-15-06 | | whatthefat: <Dionyseus>
Thanks for the tip - I'm downloading the beta version now! ;) |
|
| Apr-15-06 | | Akavall: <Dionyseus><After 27.fxe5 dxe5, Rybka 1.1 immediately claims 28.kh2 is the strongest move.> I am going to go ahead and take Rybka's word for it that 28. Kh2 is in fact the right plan. If Rybka sees the right plan(28. Kh2, it is not suggesting 28. Qf1 does it?), it is not surprising that it choses 26. f4 and 27. fxe5. Pretty amazing though, maybe I should download it too :). The problem that I have with Deep Blue, is that it obviously didn't see the right plan, but still played the strongest move. <whatthefat> Nice analysis, Thanks. <You can see though that it's not unreasonable to believe 26.f4 could be rated #1 given some more computational time (i.e., a few more ply).> No, not unreasonable to this that, but I think this has mostly to do with Fritz finding a right plan, if he was able to find a right plan with a few more ply, than of course 26. f4 could become #1 choice. Btw, my Fritz gives 28. Kh2 and 28. R3a2 the same reading of +.41, though 28.Qf1 is close with +.38 ; I didn't have it on for very long though. <Dionyseus> Do all your engines suggest 26. f4? Do they all want to play 28. Kh2? |
|
| Apr-15-06 | | Dionyseus: <Akavall> At a depth of 18, Rybka 1.1 wants to play 26.Bd1 with the idea to play Qa2 next, evaluates it as +0.55 after 10m:02s. It claims 26.f4 as the 17th strongest move at a depth of 16, evaluates it as +0.29. |
|
| Apr-16-06 | | Akavall: <Dionyseus> Thank You. |
|
| Apr-28-06 | | Elixir of Life: Maybe Kasparov's Advanced Chess Idea came from IBM? IBM may have been combining the strategic skill of human with the tactical skill of a computer, so, in effect, IBM might be playing advanced chess. |
|
May-02-06
 | | BishopBerkeley: Britney Spears' Guide to Semiconductor Physics is a pretty good introduction to the subject: http://britneyspears.ac/lasers.htm
For those who would like to build their own parallel cluster supercomputer, you might check out the Beowulf Project: http://www.beowulf.org/overview/how...
And for those who are looking for a substitute for sleep, you might take a look at: http://www.donotsleep.com/
(But whatever you do, please do NOT fill a co-worker's cubicle with packing peanuts! Such a joke is not always laughed at: http://www.donotsleep.com/images/pr... ) (: ♗ Bishop Berkeley ♗ :)
P.S. Should anyone dispute the participation of Ms. Spears in her eponymous semiconductor physics website, please point them to the following photographic evidence: http://britneyspears.ac/wallpaper/b... |
|
| May-02-06 | | SnoopDogg: <I think there is simply no evidence at all> Open your eyes.
Although the controversy may never be fully understood, I think all chess players can agree on one thing. Deep Blue in no way during the match, proved superiority over Kasparov. Probably one of the biggest overreactions since the 1938 radio reading of "The War of the Worlds" was the Deep Blue Match. Resigning in a drawn position and an absurd blunder in opening theory to me, just reveals a psychologically hurt Kasparov rather than a chess playing silicon monster. Yet you have the Deep Blue team saying, "I think people should consider this match between the greatest player in the world, and Garry Kasparov." |
|
May-03-06
| | BishopBerkeley: This picture (presumably from the early days of computer Chess) is certainly worth viewing: http://www.freewebs.com/itschessjim...
Seems appropriate somehow that a pre-Staunton Chess set should be used! The person who posted this image ("mack") says the following: === quote ===
Hello Ninjas...
I bought a book cheap from my library today, 'Mathematics' (David Bergamini Time Life Books, 1972), and flicked onto the page about computers... On p.34 there was the following curious picture [the picture linked above]... Besides it is the text:
CHESS BY COMPUTER
A chess-playing computer makes its move on sheets printed with a chess board pattern. The best human players can still beat computers, but possibly not for long: machines are now being programmed to "remember" the consequences of every mistake they ever made. So, who's playing here? What's going on? Which computer is that? === end quote ===
A writer called "xplor" responds:
=== quote ===
In 1962 about the time the book was written the first MIT chess program was written. It was the first chess program that played regular chess credibly. It was written by Alan Kotok for his B.S. thesis project, assisted by John McCarthy of Stanford. The program ran on an IBM 7090, looking at 1100 positions per second. David Bergamini 1963. Mathematics. "The mathematics of beauty in nature and art"... === end quote ===
http://www.chessninja.com/cgi-bin/u... (Both messages were posted on May 19, 2004)
(: ♗ Bishop Berkeley ♗ :)
|
|
Jun-06-06
| | BishopBerkeley: It has been called the greatest TV commercial ever made, though it aired only once nationwide in the U.S. (during the Super Bowl of 1984). Directed by Ridley Scott (of "Blade Runner" fame) it shows a swift, hammer-wielding heroine as she confronts an enormous monolithic talking face that has mesmerized all the others within its hearing. It was the commercial that introduced the Apple Macintosh to the world, and you may see it here (it should take about 40 seconds to load on a fast connection): http://www.uriahcarpenter.info/1984...
The clever people at Apple Computer celebrated the 20th anniversary of this commercial by providing a version of it in which their heroine is wearing an iPod. (It is most visible at the moment she throws the hammer.) An excellent job of splicing (if you like that sort of thing)! Otherwise (as near as I can tell) the commercial is identical to the original. (This is actually a crisper version of the video than the one presented at the earlier link): http://spherule.com/media/video/198...
More may be learned about this curious snippet of filmmaking and the role in played in the history of the personal computer (and the history of advertizing) here: http://en.wikipedia.org/wiki/1984_%... The heroine is portrayed by athlete Anya Major. Apparently few of the actresses who tried out for the role could wield the sledge-hammer in the way necessary for the part. But Ms. Major was an experienced discus-thrower, and she seems to have had little difficulty with the demands of the role. (The Wikipedia piece on her says, "The petite, blonde [Ms. Anya] Major, an experienced discus thrower who had been discovered at a local health club, won the role on the basis of her looks and her ability to spin with the hammer without becoming dizzy....") http://en.wikipedia.org/wiki/Anya_M...
Ah, the computer. For good or for ill, we are, indeed, here at the revolution.... "Any sufficiently advanced technology is indistinguishable from magic." -Arthur C. Clarke (Clarke's Third Law: http://en.wikipedia.org/wiki/Clarke... ) (: ♗ Bishop Berkeley ♗ :)
|
|
Jul-18-06
| | BishopBerkeley: For those who are interested in the future of computing, I encourage you to test-drive "Google Spreadsheets": http://spreadsheets.google.com/
One nice feature of Google Spreadsheets is that you can BOOKMARK your spreadsheets in your web-browser. One-click-access to spreadsheets that you can edit from home, work, or on the road is really quite wonderful, I think! (If you do not logoff your Google account, access is faster, of course.) The feature-set of Google Spreadsheets is modest compared with MS-Excel, and you probably wouldn't want to store any security-critical information "out there", but for relatively simple uses, it is really very nice. (Cell-widths and heights are fully configurable.) For those who would like to keep a note-pad of Chess notes "out there" so that you can make notes relative to your online play, you might just try setting the column width of the first column very wide, so that it becomes a bit like a line-by-line text editor / word-processor. Your Google Spreadsheets can also be backed up to your local machine easily. ("I have seen the future, and I've even created a backup copy!") They can also be shared with others (both write-access and read-only-access). It's always fun to check in with Google Labs to see what those highly-creative folks are experimenting with (definitely worth a bookmark!) http://labs.google.com/
Can a full-featured "Google Office" suite be far off?!? Are we entering the post-Microsoft age of personal computing? (: ♗ Bishop Berkeley ♗ :)
|
|
| May-06-07 | | CapablancaFan: Here's a nice video of the match between Kasparov and Deep Blue. It's about 10 mins. http://www.youtube.com/watch?v=cK0Y... |
|
| May-11-07 | | Dionyseus: <Akaval> < Deep Blue plays 26. f4! with the right plan, but after 26...Nf6 27. fxe5 dxe5 plays a slightly inaccurate 28. Qf1!? Don't you find it very strange how Deep blue found a right plan but had trouble executing it? > Actually Deep Blue didn't think Kasparov would play 26...Nf6. Here's what Deep Blue thought was best when it played 26.f4 after thinking for 234 seconds : 26.f4 exf4 25.Bxe4 Rc8b8 26.Be3 Be7 27.axb5 axb5 28.Ba7 Rxb7 26.f4 might not be the strongest move anyways. Rybka 2.2 sticks to 26.Bd1 after 35 seconds, and as of depth 22 it still thinks Bd1 is best:
click for larger viewAnalysis by Rybka 2.2 mp:
26.Bd1 g6 27.Qa2 Rab8 28.axb5 Bxb5 29.Bg4 Rc7 30.Nf1 Nf6 31.Bf3 h5 32.Qc2 Kh7
² (0.43) Depth: 22 00:19:09 238mN
|
|
| May-24-07 | | Hesam7: <“Like the victory of IBM’s Deep Blue chess-playing supercomputer 10 years ago this month, the debut of POWER6 processor-based systems proves that relentless innovation brings ‘impossible’ goals within reach,” said Bill Zeitler, senior vice president, IBM Systems and Technology Group. “The POWER6 processor forges blazing performance and energy conservation technologies into a single piece of silicon, driving unprecedented business value for our customers.”> |
|
 |
|
|
< Earlier Kibitzing · PAGE 7 OF 10 ·
Later Kibitzing> |
|
|
|
Kibitzing
