Did a Computer Bug Help Deep Blue Beat Kasparov?

[Rémi Coulom]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

[bob]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

Been discussed elsewhere. Pure speculation. DB certainly had a bug, according to quotes from Hsu's book, but to presume that making an oddball move in a clearly lost position in game one "spooked" Kasparov and caused him to lose game 2 and 6 is wild speculation, at the very least...

[Rolf]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

Been discussed elsewhere. Pure speculation. DB certainly had a bug, according to quotes from Hsu's book, but to presume that making an oddball move in a clearly lost position in game one "spooked" Kasparov and caused him to lose game 2 and 6 is wild speculation, at the very least...

It is intended cheat at least. What is or better was human vs machine chess in such a match situation? The human player could rely on certain characteristics of the machine program to understand howw he had to adapt his chess. There was the preparation part but also the analysis of the opponent. It is trivial that Kasparov was cheated in deveral ways.

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

2) machine chess meant always that the intelli0gent human can understand a strategy for the inflexible machine.

3) If however the machine side cheated the conditions further by changing it towards a human directed tricky flexibility change between the single games (remember only 3 weith each color) then how should the intelligent human exploitate the typical weakness of the machine? BTW we mkiust not discuss that in the meantime such matches are out of mode because the preparation of the best human players would cost too much time and money. But for Kasparov it would have been possible to do if the DB2 team wouldnt have cheated. I didnt know till now of this bug but it's relevant for chearing.

4) Bob who always defended IBM team of his friends has already commented. He argues that this bug in a totally lost position couldnt have irritated Kasparov. This is from a frog perspective correct but it's not correct as such. We chessplayers at least know well that unexpected surprises are irritating. If I play a matxch and suddenly such an irrational move, more a stupid human move, sends the message that here I have not just the machine as opponent with its tactical skills, but also typical human surprises and how could I find out if I have only one more game with a color?

5) But how could the IBM side cheat in a direct influencing of the machine in case of emergency? Well, that is easy to do. If I have certain bugs I guess they had many more and only this one became public, then I can influence by shortening the time for a move. No need to input a human outthought move, there is only a need to confuse the opponent by surprise. By a move a machine usually doesnt play.)

6) The cheating is already done, if the human playernhad no certainty of the expected play of the machine. Super GM are trained to readd the sense of the opponents play. Their best chess comes if they play against equally strong players. A weaker player could be more difficult if there are only one or two games.

7) I dont know what the IBM team had prepared as possible cheating tools. In highest alert they could cut the circuit for a couple of moments, just to give an example. Depending on how the restart would have been done, a direct human intervention were possible.

8) It's already a good proof for their cheating that allegedly Hoane repaired 4 of 5 bug situation but forgot the fifth. It is clear that a totally sober machine would only be good for a single player and that was Kasparov. Every inconsistence were advantage DB2.

9) On top of these aspects we should never forget their sudden impoliteness in the communication of the sides. That alone was terrible for Kasparov who thought himself as part of a science team. I see here a big responsibility of Ken Thompson who failed big to support the science of the event.

10) All this is about cheting, but taking a design or a lower class frame of engine open source to begin with, tht is not cheating. Nobody was tricked out by Vas who played a very weak chess with his early pilot entities. It's always funny to watch how Bob sees in Vas big crimes while every possible wrong on the side of DB2 team is refutated and most of the time the critics are insulted.

[Daniel Shawul]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

A randomly picked move turned brilliant? I don't think so. It should have occured more than once if that was the case. A discontinuity in what should have been a continuum is indicative of some kind of divine/human intervention.

Some guy thinks the computer fixed the bug itself (probably HAL fan).
"This only puts more merit to the computer, because he was able to identify a fault in himself and find a solution that was thrown randomly, typical act of intelligence and not an accident."

[Dr.Wael Deeb]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

Been discussed elsewhere. Pure speculation. DB certainly had a bug, according to quotes from Hsu's book, but to presume that making an oddball move in a clearly lost position in game one "spooked" Kasparov and caused him to lose game 2 and 6 is wild speculation, at the very least...

It is intended cheat at least. What is or better was human vs machine chess in such a match situation? The human player could rely on certain characteristics of the machine program to understand howw he had to adapt his chess. There was the preparation part but also the analysis of the opponent. It is trivial that Kasparov was cheated in deveral ways.

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

2) machine chess meant always that the intelli0gent human can understand a strategy for the inflexible machine.

3) If however the machine side cheated the conditions further by changing it towards a human directed tricky flexibility change between the single games (remember only 3 weith each color) then how should the intelligent human exploitate the typical weakness of the machine? BTW we mkiust not discuss that in the meantime such matches are out of mode because the preparation of the best human players would cost too much time and money. But for Kasparov it would have been possible to do if the DB2 team wouldnt have cheated. I didnt know till now of this bug but it's relevant for chearing.

4) Bob who always defended IBM team of his friends has already commented. He argues that this bug in a totally lost position couldnt have irritated Kasparov. This is from a frog perspective correct but it's not correct as such. We chessplayers at least know well that unexpected surprises are irritating. If I play a matxch and suddenly such an irrational move, more a stupid human move, sends the message that here I have not just the machine as opponent with its tactical skills, but also typical human surprises and how could I find out if I have only one more game with a color?

5) But how could the IBM side cheat in a direct influencing of the machine in case of emergency? Well, that is easy to do. If I have certain bugs I guess they had many more and only this one became public, then I can influence by shortening the time for a move. No need to input a human outthought move, there is only a need to confuse the opponent by surprise. By a move a machine usually doesnt play.)

6) The cheating is already done, if the human playernhad no certainty of the expected play of the machine. Super GM are trained to readd the sense of the opponents play. Their best chess comes if they play against equally strong players. A weaker player could be more difficult if there are only one or two games.

7) I dont know what the IBM team had prepared as possible cheating tools. In highest alert they could cut the circuit for a couple of moments, just to give an example. Depending on how the restart would have been done, a direct human intervention were possible.

8) It's already a good proof for their cheating that allegedly Hoane repaired 4 of 5 bug situation but forgot the fifth. It is clear that a totally sober machine would only be good for a single player and that was Kasparov. Every inconsistence were advantage DB2.

9) On top of these aspects we should never forget their sudden impoliteness in the communication of the sides. That alone was terrible for Kasparov who thought himself as part of a science team. I see here a big responsibility of Ken Thompson who failed big to support the science of the event.

10) All this is about cheting, but taking a design or a lower class frame of engine open source to begin with, tht is not cheating. Nobody was tricked out by Vas who played a very weak chess with his early pilot entities. It's always funny to watch how Bob sees in Vas big crimes while every possible wrong on the side of DB2 team is refutated and most of the time the critics are insulted.

Most of the wriiten above is rubbish as always....

I partially agree about the impoliteness from the side of Deep Blue team though....
Dr.D

[rbarreira]

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

I find this argument strange, for two reasons:

1- Kasparov had played different versions of Deep Blue several times before, and Deep Blue played games against other people. Sure they were not exactly equal to the Deep Blue that won against Kasparov, but no human would expose his pre-match improvements and preparation either. Why should computers have to expose everything?

2- If Kasparov needed to see Deep Blue's games to beat it, doesn't that mean Kasparov is exploiting specific weaknesses rather than outplaying the computer at chess in general?

[Dr.Wael Deeb]

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

I find this argument strange, for two reasons:

1- Kasparov had played different versions of Deep Blue several times before, and Deep Blue played games against other people. Sure they were not exactly equal to the Deep Blue that won against Kasparov, but no human would expose his pre-match improvements and preparation either. Why should computers have to expose everything?

2- If Kasparov needed to see Deep Blue's games to beat it, doesn't that mean Kasparov is exploiting specific weaknesses rather than outplaying the computer at chess in general?

Excellent points....
Dr.D

[bob]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

Been discussed elsewhere. Pure speculation. DB certainly had a bug, according to quotes from Hsu's book, but to presume that making an oddball move in a clearly lost position in game one "spooked" Kasparov and caused him to lose game 2 and 6 is wild speculation, at the very least...

It is intended cheat at least.

I hope, one day, you are no longer able to post here. You REALLY want to imply that the DB guys inserted a really bad bug into their program, just to confuse and intimidate Kasparov? REALLY?

That is about the most nonsensical claim I have ever seen.

What is or better was human vs machine chess in such a match situation? The human player could rely on certain characteristics of the machine program to understand howw he had to adapt his chess. There was the preparation part but also the analysis of the opponent. It is trivial that Kasparov was cheated in deveral ways.

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

Pure bullshit. There were hundreds of deep blue (first version) games available. The 1997 DB machine was not completed until days before the event, because the chips were late in being made. But there were the 6 games from the 1996 match he played, and there were lots of games after that (and prior to those) publicly available as DB was playing exhibition matches around the world as they were tuning and tweaking the thing.

2) machine chess meant always that the intelli0gent human can understand a strategy for the inflexible machine.

Apparently typed by a room full of monkeys. I can't even begin to understand what that is supposed to mean, nor how the DB guys used that to "cheat kasparov".

3) If however the machine side cheated the conditions further by changing it towards a human directed tricky flexibility change between the single games (remember only 3 weith each color) then how should the intelligent human exploitate the typical weakness of the machine? BTW we mkiust not discuss that in the meantime such matches are out of mode because the preparation of the best human players would cost too much time and money. But for Kasparov it would have been possible to do if the DB2 team wouldnt have cheated. I didnt know till now of this bug but it's relevant for chearing.

All you have to do is cite which SPECIFIC rule the DB guys violated, if they cheated. One written rule, either in the FIDE rules of chess, or in the pre-match contract Kasparov negotiated, will suffice. Of course we all know there are absolutely none that were violated. So no cheating of any kind.

4) Bob who always defended IBM team of his friends has already commented. He argues that this bug in a totally lost position couldnt have irritated Kasparov. This is from a frog perspective correct but it's not correct as such. We chessplayers at least know well that unexpected surprises are irritating. If I play a matxch and suddenly such an irrational move, more a stupid human move, sends the message that here I have not just the machine as opponent with its tactical skills, but also typical human surprises and how could I find out if I have only one more game with a color?

Would it irritate/confuse Kasparov if he played me and I played a poor move? Your opponent makes an unexpected move, your task is to decide if he is preparing some subtle trap, or did he make a tactical/positional error.

5) But how could the IBM side cheat in a direct influencing of the machine in case of emergency? Well, that is easy to do. If I have certain bugs I guess they had many more and only this one became public, then I can influence by shortening the time for a move. No need to input a human outthought move, there is only a need to confuse the opponent by surprise. By a move a machine usually doesnt play.)

There are detailed copies of the games available. They show the times, as recorded by the press from the TV monitor everyone was watching. NO discrepancies between those and the IBM logs were found. Ergo, this never happened.


[quote

6) The cheating is already done, if the human playernhad no certainty of the expected play of the machine. Super GM are trained to readd the sense of the opponents play. Their best chess comes if they play against equally strong players. A weaker player could be more difficult if there are only one or two games.[/quote]

Pure bullshit. GMs play simul exhibitions all the time and have no idea who or what they are playing.

7) I dont know what the IBM team had prepared as possible cheating tools. In highest alert they could cut the circuit for a couple of moments, just to give an example. Depending on how the restart would have been done, a direct human intervention were possible.

And it would be detectable after the fact. None happened.

8) It's already a good proof for their cheating that allegedly Hoane repaired 4 of 5 bug situation but forgot the fifth. It is clear that a totally sober machine would only be good for a single player and that was Kasparov. Every inconsistence were advantage DB2.

9) On top of these aspects we should never forget their sudden impoliteness in the communication of the sides. That alone was terrible for Kasparov who thought himself as part of a science team. I see here a big responsibility of Ken Thompson who failed big to support the science of the event.

The sudden impoliteness came from Kasparov when he accused THEM of cheating...

10) All this is about cheting, but taking a design or a lower class frame of engine open source to begin with, tht is not cheating. Nobody was tricked out by Vas who played a very weak chess with his early pilot entities. It's always funny to watch how Bob sees in Vas big crimes while every possible wrong on the side of DB2 team is refutated and most of the time the critics are insulted.

Cheating is defined as violating the rules used for a contest. He cheated, whether you like it or not.

[bob]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

Been discussed elsewhere. Pure speculation. DB certainly had a bug, according to quotes from Hsu's book, but to presume that making an oddball move in a clearly lost position in game one "spooked" Kasparov and caused him to lose game 2 and 6 is wild speculation, at the very least...

It is intended cheat at least.

I hope, one day, you are no longer able to post here. You REALLY want to imply that the DB guys inserted a really bad bug into their program, just to confuse and intimidate Kasparov? REALLY?

That is about the most nonsensical claim I have ever seen.

What is or better was human vs machine chess in such a match situation? The human player could rely on certain characteristics of the machine program to understand howw he had to adapt his chess. There was the preparation part but also the analysis of the opponent. It is trivial that Kasparov was cheated in deveral ways.

1) he had no games the machine that played him had played earlier. That is fatal in a 6 games match with only 3 White games.

Pure bullshit. There were hundreds of deep blue (first version) games available. The 1997 DB machine was not completed until days before the event, because the chips were late in being made. But there were the 6 games from the 1996 match he played, and there were lots of games after that (and prior to those) publicly available as DB was playing exhibition matches around the world as they were tuning and tweaking the thing.

2) machine chess meant always that the intelli0gent human can understand a strategy for the inflexible machine.

Apparently typed by a room full of monkeys. I can't even begin to understand what that is supposed to mean, nor how the DB guys used that to "cheat kasparov".

3) If however the machine side cheated the conditions further by changing it towards a human directed tricky flexibility change between the single games (remember only 3 weith each color) then how should the intelligent human exploitate the typical weakness of the machine? BTW we mkiust not discuss that in the meantime such matches are out of mode because the preparation of the best human players would cost too much time and money. But for Kasparov it would have been possible to do if the DB2 team wouldnt have cheated. I didnt know till now of this bug but it's relevant for chearing.

All you have to do is cite which SPECIFIC rule the DB guys violated, if they cheated. One written rule, either in the FIDE rules of chess, or in the pre-match contract Kasparov negotiated, will suffice. Of course we all know there are absolutely none that were violated. So no cheating of any kind.

4) Bob who always defended IBM team of his friends has already commented. He argues that this bug in a totally lost position couldnt have irritated Kasparov. This is from a frog perspective correct but it's not correct as such. We chessplayers at least know well that unexpected surprises are irritating. If I play a matxch and suddenly such an irrational move, more a stupid human move, sends the message that here I have not just the machine as opponent with its tactical skills, but also typical human surprises and how could I find out if I have only one more game with a color?

Would it irritate/confuse Kasparov if he played me and I played a poor move? Your opponent makes an unexpected move, your task is to decide if he is preparing some subtle trap, or did he make a tactical/positional error.

5) But how could the IBM side cheat in a direct influencing of the machine in case of emergency? Well, that is easy to do. If I have certain bugs I guess they had many more and only this one became public, then I can influence by shortening the time for a move. No need to input a human outthought move, there is only a need to confuse the opponent by surprise. By a move a machine usually doesnt play.)

There are detailed copies of the games available. They show the times, as recorded by the press from the TV monitor everyone was watching. NO discrepancies between those and the IBM logs were found. Ergo, this never happened.

6) The cheating is already done, if the human playernhad no certainty of the expected play of the machine. Super GM are trained to readd the sense of the opponents play. Their best chess comes if they play against equally strong players. A weaker player could be more difficult if there are only one or two games.

Pure bullshit. GMs play simul exhibitions all the time and have no idea who or what they are playing.

7) I dont know what the IBM team had prepared as possible cheating tools. In highest alert they could cut the circuit for a couple of moments, just to give an example. Depending on how the restart would have been done, a direct human intervention were possible.

And it would be detectable after the fact. None happened.

8) It's already a good proof for their cheating that allegedly Hoane repaired 4 of 5 bug situation but forgot the fifth. It is clear that a totally sober machine would only be good for a single player and that was Kasparov. Every inconsistence were advantage DB2.

9) On top of these aspects we should never forget their sudden impoliteness in the communication of the sides. That alone was terrible for Kasparov who thought himself as part of a science team. I see here a big responsibility of Ken Thompson who failed big to support the science of the event.

The sudden impoliteness came from Kasparov when he accused THEM of cheating...

10) All this is about cheting, but taking a design or a lower class frame of engine open source to begin with, tht is not cheating. Nobody was tricked out by Vas who played a very weak chess with his early pilot entities. It's always funny to watch how Bob sees in Vas big crimes while every possible wrong on the side of DB2 team is refutated and most of the time the critics are insulted.

Cheating is defined as violating the rules used for a contest. He cheated, whether you like it or not.

[syzygy]

I noticed that wired news article:
http://www.wired.com/playbook/2012/09/d ... puter-bug/

I may have missed it, but it seems nobody posted it here yet. Sorry if I am wrong.

Rémi

A randomly picked move turned brilliant? I don't think so. It should have occured more than once if that was the case. A discontinuity in what should have been a continuum is indicative of some kind of divine/human intervention.

A bug was triggered at the end of game 1 which caused a random move that was not briliiant at all. It did not change the outcome, since the game was already lost anyway. The (speculative) point is that the only explanation Kasparov's team could come up with for that move is that DB was searching deeper than they had considered possible (to a depth at which the move being played was not worse anymore than all other alternatives). This false impression of DB's strength might then have had an impact e.g. on Kasparov's decision in game 2 to prematurely resign.

The bug is a verifiable fact. The kind of impact of this bug on Kasparov is speculation, as far as I'm aware.

The first comment below the Wired article seems to correct some inaccuracies in the article. This part explains where Nate Silver (who wrote a book that caused all these recent articles popping up) got the theory from:

[Adding that now I see where Silver got this. I don't have his book in front of me so I don't know if he gives the source. But Frederic Friedel's anecdote about the post-game analysis with Kasparov on game one has Garry trying to explain 44..Rd1 by wondering if it played it because it saw so deeply that it was finding checkmates. I'd forgotten about that, as history has discarded the move. So the Murray/Silver hypothesis, while still a big stretch to me, makes sense in that light. Friedel himself is clearly trying to paint that picture to explain how Kasparov missed the draw. Apologies for doubting the redoubtable Mr. Silver!]

So the theory is speculation, and it is a stretch, but it is not completely without merit. It is certainly based on facts.