Kozminski Techblog

Reading Time: 3 minutes

In 1997, a momentous event occurred that forever changed the relationship between humans and machines: the supercomputer IBM Deep Blue defeated the reigning world chess champion, Garry Kasparov. Deep Blue’s victory was a defining moment, showcasing the potential of machines to tackle complex intellectual challenges.

The Origin of Deep Blue

The journey of Deep Blue began with a research project at Carnegie Mellon University in the late 1980s. Initially called Deep Thought, it was developed by a team led by Feng-hsiung Hsu. After joining IBM, Hsu continued his work, and Deep Thought evolved into Deep Blue. IBM’s goal was to create a machine capable of defeating the best human chess players in the world.

Deep Blue’s development was driven by a combination of hardware innovation and software sophistication. The team faced numerous challenges, including designing specialized processors for parallel computation and creating software capable of evaluating millions of chess positions per second with strategic accuracy. Balancing computational speed with decision-making quality was a critical hurdle that required constant iteration and refinement. Unlike general-purpose computers, Deep Blue was designed specifically for chess. It featured custom-built processors optimized for parallel computation, allowing it to evaluate millions of potential moves in mere seconds.

How Deep Blue Worked

Deep Blue’s strength lay in its ability to analyze a staggering number of positions per second. At its peak, it could evaluate 200 million positions per second, a feat unimaginable for any human player. The supercomputer achieved this through a brute-force search algorithm, combined with advanced evaluation functions that helped it assess the relative strength of different positions on the board.

Key techniques used by Deep Blue included:

1. Brute-force search: Deep Blue systematically examined a vast number of possible moves and countermoves up to a certain depth.
2. Evaluation functions: These functions allowed Deep Blue to score positions based on various chess heuristics, such as material advantage, king safety, and control of key squares.
3. Opening and endgame databases: Deep Blue had access to extensive pre-programmed databases that covered thousands of well-known opening sequences and endgame scenarios, giving it an edge in these critical phases of the game.

The 1997 Kasparov vs. Deep Blue Match

The showdown between Kasparov and Deep Blue took place in May 1997. It was a six-game match that drew worldwide attention. Kasparov, known for his aggressive and creative style, had previously defeated an earlier version of Deep Blue in 1996. However, this time, the upgraded machine proved too powerful.

The turning point came in Game 6, before which the match was tied 2,5-2,5. Kasparov has decided to go into the Steinitz variation in the Caro-Kann defence playing with the black pieces. In the pretty known line Kasparov played the move 7. h6 which allowes 8. Kxe6. This lines was known to be dubious already at the time, but Kasparov didn’t believe the engine would sacrifice a piece without clear compensation. This might seem like a stupid risk by the world champion, but back in the day computers valued material over any positional advantage. He expands on the topic in a book “Deep Thinking” which I recommend to anyone interested in the topic. The world was shocked, when the Russian resigned the game on move 19. losing the match. Deep Blue won with a final score of 3.5–2.5. Kasparov’s defeat marked the first time a reigning world champion lost a match to a computer under standard chess tournament conditions.

Kasparov later accused IBM of unfair play, suggesting that human intervention might have influenced some of Deep Blue’s moves. IBM denied these allegations, and while the controversy remains a topic of debate, it does not diminish the significance of the achievement.

The Aftermath and Legacy

Deep Blue’s victory had a profound impact on both the chess world and AI research. For chess, it marked the beginning of an era where computers became indispensable tools for analysis and training. Today, even amateur players can access powerful chess engines far more advanced than Deep Blue.

In AI, the match underscored the potential of machines to solve highly complex problems through brute-force computation and expert knowledge. However, it also highlighted the limitations of this approach. Subsequent AI developments, such as Google’s AlphaZero, shifted towards machine learning and neural networks, which focus on teaching machines to learn and adapt rather than relying solely on pre-programmed knowledge.

Chess computers in the modern era

Since Kasparov’s loss computers have improved so much that there is no debate if they are better then people for more than 20 years. The conversation has changed to computers ruining top level chess showing refutations to any new creative opening ideas and making grandmasters memorise 40 moves for each side instead of playing on their own. Let me know in the comments what your opinion is and what aspect of our lives computers will conquer next.

Written with help of Ninja AI.

Sources:

-Garry Kasparov “Ostatni Bastion Umysłu”

-https://www.youtube.com/watch?v=EsMk1Nbcs-s

https://en.wikipedia.org/wiki/Deep_Blue_versus_Garry_Kasparov#Game_6_2