A great man and founder of the discipline of Artifi cial Intelligence which he named at the 1956 Dartmouth College Summer Conference, has recently left us.

 

John McCarthy, had stints as a Professor at MIT, Dartmouth College, Princeton University, and Stanford University. He spent most of his career at Stanford where his is Professor Emeritus.

 

He is credited with the invention of the LISP programming language which had been standard for developing AI programs for many years, particularly in this country. With a strong aptitude for Mathematics, McCarthy received a B.S. in Mathematics from Caltech in 1948 and then a Ph.D. in Mathematics from Princeton University in 1951 under Solomon Lefschetz.

 

Professor McCarthy’s interests and contributions were profound and wide-ranging, covering many areas of AI, including, for example, publications on diverse areas of logic, natural language processing, computer chess, cognition, counterfactuals, common sense, and a number of philosophical problems from the standpoint of AI. His home page (http://www-formal.stanford. edu/jmc/) is a great treat in itself where most of his publications in these areas are available. There you can also fi nd more recent pages on “The Sustainability of Human Progress” which he is quite optimistic about. Some other well known titles include: “Circumscription – a form of Nonmonotonic Reasoning”, “Artifi cial Intelligence, Logic and Formalizing Common Sense”, and “The Little Thoughts of Thinking Machines”.

 

As a founding father of AI McCarthy often used his papers as a mean of commentary on what AI systems need in order to be practically useful and effective, such as “Some Expert Systems Need Common Sense (1984) and “Free Will – Even for Robots”

 

He was the recipient of the prestigious A.M. Turing Award for his contributions to AI in 1971. Other awards he received include The National Medal of Science in Mathematical, Statistical, and Computational Sciences (1971), and the Benjamin Franklin Medal in Computer and Cognitive Science (2003). Like many early AI researchers, (Turing, Newell, and Simon) McCarthy had an early interest in mastering chess via AI methods. Between 1959 and 1962, students of John McCarthy at MIT, Alan Kotok, Elwyn Berlekamp (1960), Michael A. Lieberman, Charles Niessen and Robert A. Wagner, wrote a chess program for the IBM 7090. The program was known as the McCarthy –Kotok Program and lost a match to a Soviet counterpart, 3-1 in 1966.

 

McCarthy stated: “We humans are not very good at identifying the heuristics we ourselves use.” I believe that the early AI researchers did not appreciate how much of chess is knowledge and patternbased. They had more inclinations towards chess being solved by “search, calculation, and logic.” The pattern-based nature of chess was later more formally demonstrated by the research of Chase and Simon (1973). However success at chess programming leading to the superstrong programs of today has been mainly accomplished through deep search and statistical techniques. In the 1990’s McCarthy recognized chess as the “drosophila of AI”, but recognized that the next drosophila was likely to be the game of Go. 

 

 

---