John McCarthy

Computer scientist

• Born: September 4, 1927
• Birthplace: Boston, Massachusetts
• Died: October 24, 2011
• Place of death: Stanford, California

Creator of the Lisp programming language

Born: September 4, 1927; Boston, Massachusetts

Died: October 24, 2011; Stanford, California

Primary Field: Computer science

Specialty: Computer programming

Primary Company/Organization: Stanford Artificial Intelligence Laboratory

Introduction

John McCarthy's research into artificial intelligence focused on formalizing commonsense knowledge and reasoning. His career spanned several important developments in the computer era. Like many early computer scientists, he was trained as a mathematician. By initiating experimentation into time sharing, he represented a shift into a new paradigm for computing, where researchers interacted with their own computer resources. His Stanford Artificial Intelligence Laboratory (SAIL) was funded in part by the Defense Advanced Research Projects Agency (DARPA) and was one of the first sites connected to the ARPANET. McCarthy's dedication to the field of artificial intelligence earned him the A. M. Turing Award in 1971, but equally impressive is the fact that sixteen researchers affiliated with SAIL have also been honored.

Early Life

John McCarthy has been called a red-diaper baby by historian John Markoff because his parents were members of the American Communist Party and his father, John Patrick McCarthy, was the business manager of the Daily Worker, the national newspaper of the Communist Party. (McCarthy was a member of the Communist Party for a short time in 1949.) John Patrick immigrated to the United States from Ireland. His wife, Ida Glatt, immigrated from Lithuania. They had son John in Boston but lost their house shortly thereafter, during the Great Depression, and moved to Los Angeles, where John Patrick organized members of the Amalgamated Clothing Workers union. Son John graduated from Belmont High School, Los Angeles, in 1943.

At the age of fifteen, McCarthy bought a textbook being used at the California Institute of Technology (Caltech) and taught himself calculus. By the time he entered Caltech at the age of sixteen, he had learned too much calculus for undergraduate courses and so took a graduate class. He failed physical education because he refused to attend class, left the school, and was drafted into the Army. After being stationed in Texas, McCarthy was allowed to reenter Caltech; he graduated with a bachelor of science degree in mathematics in 1948. The next year, he continued his study of mathematics at the graduate level at Princeton University, where he presented his research to John von Neumann and became friends with Marvin Minsky; he graduated two years later. In the summer of 1952, McCarthy worked at Bell Laboratories, where he met Claude Shannon and with him edited a volume of essays on intelligent machines, Automata Studies (1956). McCarthy was an instructor of mathematics at Princeton from 1951 to 1953 and at Stanford University from September 1953 to January 1955. His background as a mathematician was not unusual. In the 1950s, computer programmers were largely industry experts who learned from practical projects, such as IBM's involvement with the SAGE missile defense system. Computer programming had not yet advanced to an academic discipline, but it was beginning to be taught at the university level in mathematics departments, where it naturally fit.

Life's Work

In the spring of 1955, McCarthy was named assistant professor of mathematics at Dartmouth College. At Dartmouth, based on his acquaintance with Shannon and their shared interest in intelligent machines, McCarthy is credited for coining the term artificial intelligence in a research proposal he wrote with colleagues to the Rockefeller Foundation. The proposal resulted in the 1956 Dartmouth Conference, a two-month, ten-person symposium that marked the start of research into artificial intelligence.

Through the auspices of the Sloan Foundation, McCarthy went to the Computation Center at the Massachusetts Institute of Technology (MIT) for a one-year fellowship in 1957; he did not return to Dartmouth. From 1958 to 1961, he was an assistant professor of communication science at MIT. At MIT, he initiated research into time sharing with a memo he wrote, “A Time Sharing Operator Program for Our Projected IBM 709.” The computers at the time were used by means of “batch-processing” systems. The “batch” referred to a set of punch cards, and the implication was that only one user could work with a computer at a time. For each computer program to be run, programmers waited for the card reader to process the punch cards, the results to be calculated, and the output to be printed. It became clear that the processor was active for only a small part of that time, and given that the machines were very expensive, that idle time represented wasted money. An early form of time sharing was available for the SAGE system, where multiple users had access to the same information, but McCarthy took this idea further in his memo. He hypothesized that a user could be given the illusion of full access to a machine if the computer quickly cycled among all requests, and users could debug their programs simultaneously.

McCarthy was not comfortable proposing hardware modifications, given that his specialty was not in electronics, and so he proposed what he called minimal changes to computer circuits. His demonstration of the concept of “time stealing” (as they called it then) involved McCarthy in a computer room and the audience in a lecture hall, connected by closed-circuit television. While an ordinary batch job was running on the computer, it also collected input from a teletype machine and processed that input when the other job was completed. This inspired Fernando Corbato and his colleagues to build a prototype time-sharing operating system called Computer Time-Sharing System (CTSS) in November 1961; when this system was operational in 1963, it initiated the Project on Mathematics and Computation (Project MAC). With Ed Fredkin, McCarthy put together a time-sharing system on a Digital Electronics Corporation (DEC) PDP-1 computer at Bolt, Beranek, and Newman, which was operational in 1962. McCarthy declined to take this work further; he wanted to return to research on artificial intelligence. Instead, Alan Kotok, one of McCarthy's students, designed the first commercial time-sharing system, the DEC PDP-6. Time sharing would become a research priority for DARPA under J. C. R. Licklider, who funded several projects in that field beginning in 1963. The importance of time sharing cannot be underestimated, as indicated by the fact that what would become the ARPANET was imagined as a cooperative network of time-sharing computers.

Time sharing, however, was not McCarthy's primary interest; it was merely a tool to aid research on artificial intelligence (AI). When Minsky joined McCarthy at MIT, they founded the MIT Artificial Intelligence Laboratory. His Teddington Symposium paper presented his early thinking on how to incorporate commonsense reasoning into AI and was published in 1959 as “Programs with Common Sense.” In it, McCarthy described how a program could use sentences in a logical language, acquire new knowledge, and then draw inferences from the stored information. He and his students created a computer program to play chess, later known as Kotok-McCarthy, soon after teaching what Markoff says was the first undergraduate course in computer science. In 1958, McCarthy invented the programming language Lisp, the second-oldest programming language (the first being Fortran), which would be used for symbolic processing needed for artificial intelligence research. The first paper describing Lisp was published in the Communications of the ACM in 1960, and its examples would be widely used in Lisp textbooks. McCarthy's 1959 “advice taker” paper, “Programs with Common Sense,” proposed that reasoning for everyday problems was just as worthy of study as abstract programs, and both could be represented in mathematical logic.

After six years at MIT, McCarthy returned to Stanford's mathematics department in 1962, where he remained until he retired in 2000. He created the Stanford University AI Project in 1963, which after 1971 was known as the Stanford Artificial Intelligence Laboratory (SAIL), and served as its director from 1965 to 1980. From the start, SAIL was an innovative place. In a symposium honoring his achievements, McCarthy's students and colleagues praised his flexible, inquiry-based mentoring style and his willingness to secure research positions for them.

Starting in 1965, McCarthy made several visits to the Soviet Union. In 1965, he brought the chess-playing program written by MIT undergraduates, and a computer chess match against the program, written by researchers at the Institute for Theoretical and Experimental Physics in Moscow, was the result. Moves were transmitted via telegraphy from 1966 to 1967. In 1968, McCarthy taught in Akademgorodok in Novosibirsk, and in 1975, he helped persuade the Soviet Union to allow cybernetics researcher Alexander Lerner to speak at the International Joint Conference on Artificial Intelligence. McCarthy is said to have smuggled a fax machine and copier to the Soviet linguist Larisa Bogoraz.

Although student researchers at SAIL had access to a time-sharing system, undergraduates at Stanford had to use a batch-processing system with punch cards until 1976. In order to justify the cost of a time-sharing system for them, McCarthy set up a research project that would provide a time-sharing system for undergraduates, calling it the Low Overhead Time-Sharing System (LOTSS). By 1979, years before personal computers were widely available, undergraduates had access to a computing system, helping to create a community of undergraduate computer scientists.

As an established leader in the field, McCarthy continued to do research, published papers, mentor graduate students, and participate in professional conferences. At the turn of the century, he was working on a new programming language, Elephant. He hoped it would be a natural language interface between computers and humans, useful for commercial transactions but also research. Although it was based on other popular AI languages, the Elephant proposal indicates he was hoping to integrate past knowledge more directly into the processing of information. Also, in 2004, he wrote a sardonic science-fiction story, “The Robot and the Baby,” that gives credibility to reports of McCarthy's irreverent attitude and demonstrates his awareness of the interactions among government, science, and industry bureaucracies that shaped the development of computer science. The story features the commonsense reasoning of R781, a household robot who finds itself at the center of a child-care controversy.

McCarthy has many times been honored for his work. In 1971, the Association for Computing Machinery honored him with its A. M. Turing Award. He was elected to National Academy of Engineering in 1987. He won the Kyoto Prize from the Inamori Foundation in November 1988. He was elected to the National Academy of Sciences in 1989 and was awarded a National Medal of Science in 1990. The California Institute of Technology has named him a distinguished alumnus.

Personal Life

McCarthy retired from his full-time position at Stanford in 2001 and was named professor emeritus. He was married three times: to Martha Coyote, Vera Watson, and Carolyn Talcott. His second wife was an IBM programmer who died in a 1978 climbing accident in Nepal. He had three children, two from his first wife and one from his third. He died in his Palo Alto home at the age of eighty-four from complications related to heart disease.

Bibliography

Lifschitz, Vladimir. “John McCarthy (1927–2011).” Nature 480 (2011): 40. Print. An obituary by a well-known professor of computer science and artificial intelligence at the Univesity of Texas at Austin.

Markoff, John. What the Dormouse Said: How the 60s Counterculture Shaped the Personal Computer Industry. New York: Penguin, 2005. Print. Examines the interconnectedness of the 1960s drug and computing subcultures, including McCarthy and his accomplishments.

Morgenstern, Leora, and Sheila A. McIlraith. “John McCarthy's Legacy.” Artificial Intelligence 175 (2011): 1–24. Print. A detailed overview of what McCarthy contributed to the AI field.

Patrick, J. Hayes and Leora Morgenstern. “On John McCarthy's 80th Birthday, in Honor of His Contributions.” AI Magazine 28.4 (2012): 93–102. Print. A tribute that looks back on McCarthy's career and achievements.