Arthur Samuel

Who Arthur Samuel was

Samuel was born on 5 December 1901 in Emporia, Kansas. He studied at the College of Emporia (graduating in 1923) and completed a master's degree in electrical engineering at MIT (1926), where he also taught as an instructor until 1928. From 1928 to 1946 he worked at Bell Labs, developing vacuum tubes and contributing to radar improvements during World War II, including the TR tube for antenna switching. In 1946 he moved to the University of Illinois at Urbana-Champaign as a professor of electrical engineering, where he started the ILLIAC project, one of the early university computers in the United States.

The checkers program on the IBM 701 and the 1956 TV demonstration

In 1949 Samuel moved to IBM in Poughkeepsie, New York. There, in 1952, he began developing a checkers program for the IBM 701, the company's first commercial computer. The first version with a learning ability was completed in 1955, and on 24 February 1956 the program was first demonstrated publicly, on television. IBM's president at the time, Thomas Watson, arranged for it to be shown to shareholders and predicted that the demonstration would raise the company's stock by 15 points, and that is exactly what happened, overnight.

"Some Studies in Machine Learning," 1959: the birth of the term

In 1959 Samuel published the paper "Some Studies in Machine Learning Using the Game of Checkers" in the IBM Journal of Research and Development, and in it coined in writing, for the first time, the phrase "machine learning." The program he described combined "rote learning," saving every board position already seen together with its final reward value, with a minimax algorithm for choosing moves, and with alpha-beta pruning to cope with the severe memory limits of the computers of the time.

The famous game against Robert Nealey, 1962

In 1962 Samuel arranged an exhibition game against Robert Nealey, a blind checkers player from Connecticut whom IBM's publicity materials described as "a former Connecticut checkers champion, and one of the nation's foremost players." The program won the game, and the result was widely publicized as proof of a computer's superiority in a strategy game. A later review by the Chinook research team at the University of Alberta found that Nealey was not yet Connecticut champion at that point (he won the title only in 1966), and that his loss came from a single bad move. In a six-game rematch in 1963, Nealey won decisively, with one win and five draws. In 1966, when the program was matched against Walter Hellman, the reigning world champion, and against Derek Oldbury, four games against each, it lost all eight games, fairly easily. Still, the 1962 game remains a milestone in the public's view of machine learning history, even if its image was exaggerated in hindsight.

Stanford, awards and legacy

In 1966 Samuel moved to Stanford University, where he continued to research machine learning and also contributed to the documentation of the TeX typesetting system developed by Donald Knuth. He received the Computer Pioneer Award in 1987 for his work on adaptive non-numeric processing, and in 1990 he was named a founding fellow of AAAI (the Association for the Advancement of Artificial Intelligence). He died on 29 July 1990 at Stanford, California, at age 88, of complications of Parkinson's disease. The term he coined, "machine learning," has since become the name of the central approach to building modern artificial intelligence systems.