Garry Kasparov lost to the Deep Blue supercomputer in chess due to computer failure

One of the greatest chess games of all times and peoples is, without a doubt, the battle of Garry Kasparov and IBM's Deep Blue supercomputer in 1997. It was already the second Kasparov game with a supercomputer, a rematch of the machine.

The first game in the game was very difficult and tense, Kasparov initially had an advantage, but starting from move 44, he stopped understanding the logic of the game of the machine, and, as a result, lost the entire match. After some time, Kasparov even accused IBM of “cheating”: the manipulation of the software of the machine, which led to the defeat. After 17 years, the situation cleared up - Kasparov lost because of a malfunction in the algorithm of the computer in the very first batch of the entire battle.

What is deep blue?

At that time, Deep Blue was a really powerful system. The supercomputer was a system with a 32-core (32-node) IBM POWER2 processor, each of which was connected to eight specialized VLSI chess processors running on an RS / 6000 server platform. The Deep Blue code was written in C, and IBM AIX was used as the operating system. Deep Blue counted 200 million positions per second, and its peak performance was 11.38 gigaflops.
')

A bit about Deep Blue from its creator

Work on the creation of specialized computer systems has been conducted at IBM since 1950. In 1985, the first computer appeared for solving highly specialized chess problems. It was built by a graduate student at Carnegie University Feng-hsiung Hsu, and this system was named ChipTest.

A little later, Murray Campbell, together with Feng-hsiung Hsu, began work on the creation of Deep Blue, becoming employees of IBM. The first version of the system lost to Kasparov in 1989. The system received its second defeat in 1996. And in 1997, all games were won by the machine.

Losing man to car

On the 44th move of the critical game, Deep Blue made a generally meaningless move: he moved the rook from D5 to D1. This move did not give any prospects - neither positional nor any other. Kasparov won this game, but he stopped understanding the logic of the computer , and the next two games (2nd and 6th) passed, losing the whole battle.



Now it turned out that that move was a normal failure. Previously, IBM engineers programmed Deep Blue to perform a safe move if any failure occurred. And when this failure occurred, the computer simply moved the figure, instead of making a calculated positional move, which did not bring anything, neither gain, nor weakening of positions.

This is how the usual computer error, a bug, led to the defeat of a person in the epochal chess battle “man versus machine”.

Based on this error, a documentary short film was made, which can be viewed here (habrapriser does not miss the insertion of the code).