Arthur Lee Samuel was an American engineer and computer scientist at IBM who is remembered as one of the earliest pioneers of machine learning. His landmark paper, “Some Studies in Machine Learning Using the Game of Checkers,” was published in the IBM Journal of Research and Development in 1959.
In that paper Samuel described a checkers-playing program that improved its own play over time. Rather than being told exactly how to play, the program adjusted the weights in its position-scoring function based on experience, including by playing games against itself, so that it could learn to play a better game than the person who wrote it. The work is also widely cited as the origin of the term “machine learning.” It demonstrated, on the hardware of the 1950s, that a computer could get better at a task through experience rather than through explicit reprogramming.
Samuel’s checkers program is one of the founding examples of learning machines and a direct ancestor of the reinforcement learning and self-play methods that later powered systems like Deep Blue and AlphaZero.