Which function is was used to correct bug of the wrong tower move position in Deep Blue vs Kasparov (1997)?

1

1

During the first game, in 1997, New York, between the Deep Blue supercomputer against Kasparov, there was an unexpected tower move position by supercomputer. That bug had come out the year before while training with the great master. The programmers had identified four bugs, but the fifth came out only during the first game with kasparov.
They tried in every way to understand why that error just came out that way, then, no answer, one of the programmers who created the algorithm, Joseph Hoane, corrected that bug

I would understand, mathematically, what is this error, if it as the algorithm formula, the evaluation function, the generator of moves? Can anyone tell me how usually you correct a mistake of this kind? What should you look for?

Peter Long

Posted 2017-01-04T20:09:38.850

Reputation: 21

Question was closed 2017-01-05T23:50:43.320

1I'm actually not sure whether this is on-topic or off-... but the question itself is fascinating, so I'm hoping someone manages something like an answer ;) – Tetsujin – 2017-01-04T20:20:59.373

While this is an interesting question, the answer is likely not something that anyone but Hoane and his colleagues know. I assume that the bug was with a specific scenario in the game, but without understanding the code it would be difficult to describe the fix. – heavyd – 2017-01-04T20:30:32.803

mmm..thanks, I find something math here and here but I don't undestand if is used Montecarlo method or algorithm like minimax based on evaluated functions to determinate move generator positions

– Peter Long – 2017-01-04T20:47:55.873

Answers

2

Agree with @Tetsujin - nice question!

Maybe just a legend: Did a Computer Bug Help Deep Blue Beat Kasparov?

Either at the end of the first game or the beginning of the second, depending on who’s telling the story, the computer made a sacrifice that seemed to hint at its long-term strategy.

Kasparov and many others thought the move was too sophisticated for a computer, suggesting there had been some sort of human intervention during the game. “It was an incredibly refined move, of defending while ahead to cut out any hint of countermoves,” grandmaster Yasser Seirawan told Wired in 2001, “and it sent Garry into a tizzy.”

Fifteen years later, one of Big Blue’s designers says the move was the result of a bug in Deep Blue’s software.

The revelation was published in a book by statistician and New York Times journalist Nate Silver titled The Signal and the Noise — and promptly highlighted by Ezra Klein of the Washington Post.

For his book, Silver interviewed Murray Campbell, one of the three IBM computer scientists who designed Deep Blue, and Murray told him that the machine was unable to select a move and simply picked one at random.

Quote from Nate Silver's book The Signal and the Noise: Why So Many Predictions Fail--but Some Don't:

The bug had arisen on the forty-fourth move of their first game against Kasparov; unable to select a move, the program had defaulted to a last-resort fail-safe in which it picked a play completely at random.

duDE

Posted 2017-01-04T20:09:38.850

Reputation: 14 097

1This article also quotes Nate Silver's book stating that Deep Blue was unable find a reasonable move so it just selected a move at random. – heavyd – 2017-01-04T20:32:17.173

mmm..thanks, I find something math here and here but I don't undestand if is used Montecarlo method or algorithm like minimax based on evaluated functions to determinate move generator positions

– Peter Long – 2017-01-04T20:47:17.760