IBM Deep Blue to Google AlphaZero, a brief history of Chess AI

Deep Blue was a supercomputer developed by IBM specifically for playing chess. It was the first computer to defeat a reigning world champion, Grandmaster Garry Kasparov, in a six-game match in 1997. This was a historic achievement for artificial intelligence (AI) and chess programming, as it demonstrated the ability of a machine to outsmart a human in a complex and strategic game.

Deep Blue was based on an earlier system called Deep Thought, which was also developed by IBM and was the first computer to beat a grandmaster in tournament play. Deep Thought was described in a paper published in Scientific American in 1990, which presented the design and implementation of the system. Deep Thought used a powerful processor that could evaluate up to 720,000 chess positions per second, and a large database of chess games and openings to guide its search and evaluation.

Deep Blue improved on Deep Thought in several ways. It used a parallel supercomputer that consisted of 256 processors, each with a special-purpose chess chip that could evaluate up to 200 million positions per second. It also used more advanced AI techniques, such as search algorithms, evaluation functions, and machine learning, to explore the possible moves and outcomes of a chess game, and to choose the best move according to some criteria. It also had a team of human experts who helped to fine-tune its parameters and provide feedback after each game.

The paper that describes the architecture, algorithms, and performance of Deep Blue was published in Artificial Intelligence in 2002. It provides a detailed account of the challenges and lessons learned from the project, as well as the analysis of the six games played between Deep Blue and Kasparov. The paper also discusses the implications and limitations of Deep Blue, and the future directions for AI and chess programming.

AI has advanced significantly since the time of Deep Blue. Nowadays, AI systems can perform tasks that are not only more complex and strategic, but also more creative and diverse, such as generating art, music, poetry, code, and more. AI systems can also learn from large amounts of data and from their own experiences, without relying on human guidance or predefined rules. AI systems can also interact with humans and other agents in natural and social ways, using natural language, speech, vision, and emotion.

One example of a modern AI system that plays chess is AlphaZero, developed by Google’s DeepMind. AlphaZero is a general-purpose system that can learn to play any board game, such as chess, shogi, and Go, by playing against itself. AlphaZero does not use any human knowledge or data, but only the rules of the game. It uses a deep neural network and a reinforcement learning algorithm to learn from its own moves and outcomes, and to improve its performance over time. AlphaZero can defeat any other chess program, including Deep Blue, with ease.

AI is still evolving and improving, and there are many challenges and opportunities ahead. AI can be used for good or evil, depending on the intentions and actions of the users and developers. AI can also raise ethical, social, and philosophical questions, such as the nature and value of intelligence, creativity, and consciousness. AI can also inspire and challenge humans to learn more about themselves and the world, and to achieve new heights of excellence and innovation.