I am writing this article as I read the book “Deep Thinking” by Garry Kasparov. The book deals with his experience/experiments with AI and how it had evolved over time, from being significantly weaker than chess Grandmasters, to beating one of the greatest players, Kasparov himself.

Introduction

Chess has been one of the most commonly considered problems in AI domain. Almost anyone who has worked in AI has dwelled into chess and developed a model which can play reasonable chess. For a long time chess has been a measure of progress in AI. This probably is because of the interesting problem that chess presents, with each piece having its own unique moving pattern and having to use the pieces synchronously to win. Furthermore, the efficiency of moves is determined by the move opponent makes.

To get a bit more technical, chess is a two-player, zero sum, perfect information game. Zero sum game is where advantage for one player is disadvantage for the other. Perfect Information means that at all move the board is completely visible to both the players. Chess also follows the markov property in most cases, which means that one doesn’t need to know the exact sequence of previous moves to determine what the next move should be. This is because the next move can be determined by the current state of the chessboard and one doesn’t need to know how we got there (Except in case of avoiding three fold repetition).

Beating Thirty Two Machines

1996: Deep Blue vs Kasparov

The 6 match series between Deep Blue and Garry Kasparov was iconic to say the least, as the then greatest AI developed to play chess was challenging the world chess champion. Human Intelligence was pitted against computational powerhouse, however, it was human brain that came out victorious. The six match series concluded 4-2, with Kasparov winning three and Deep Blue 1.

While humans came out victorious this time, the defeat enabled the researchers behind Deep Blue to identify weakness and patch them up so that the machine improved.

1997: Deep Blue vs Kasparov

A year had passed by since Deep Blue’s defeat at the hands of Garry Kasparov. Researchers at IBM had a year to work and improve the machine, and they came up with what was unofficially called “Deeper Blue”. Another battle was to begin between Man and Machine, and this time Machine took the gold medal. The six match series ended 3.5 - 2.5 in favor of Deep Blue. However, it could be said with certainty that the a modern day chess engine (like Google’s AlphaZero), would easily beat Deep Blue. This is because Deep Blue didn’t so much think, rather went through all possibilities before making a move.

Machine Intelligence has come a long way since then, with advancements in neural networks and machine learning, modern machine are much more capable of thinking compared to Deep Blue. It wouldn’t be unfair to say that modern day machines are capable of computation at a pace impossible by human brain. However, human brain still seems to have an advantage compared to machines, as it is capable of thinking rather than computing. Maybe someday in the future, there will be a machine which can think the way we do.