John McCarthy (September 4, 1927–October 24, 2011) was an American researcher in computer science and one of the founders of artificial intelligence (AI). He coined the field’s name in 1955, helped organize the Dartmouth summer research project of 1956, and created the Lisp programming language. His research also encompassed interactive computing and the formal representation of commonsense knowledge. He received the ACM A. M. Turing Award in 1971. (heidelberg-laureate-forum.org)
Education and academic career
McCarthy was born in Boston, Massachusetts. He earned a bachelor’s degree from the California Institute of Technology in 1948 and a doctorate from Princeton University in 1951, both in mathematics. His mathematical background informed his subsequent interest in describing computation and intelligent reasoning through precise formal systems. (heidelberg-laureate-forum.org)
He joined Stanford University as an assistant professor in 1953, moved to Dartmouth College in 1955, and worked at the Massachusetts Institute of Technology (MIT) from 1958 to 1962. At MIT, he and Marvin Minsky started an artificial intelligence research project. McCarthy returned to Stanford as a full professor in 1962 and founded the Stanford Artificial Intelligence Laboratory (SAIL) in 1963. He became professor emeritus on January 1, 2001, and died at his Stanford home in 2011. (engineering.stanford.edu)
Naming and organizing artificial intelligence
The proposal for the Dartmouth Summer Research Project on Artificial Intelligence was dated August 31, 1955. Its four authors were McCarthy, Minsky, Nathaniel Rochester, and Claude Shannon. It proposed a two-month research meeting at Dartmouth during the summer of 1956 and requested support from the Rockefeller Foundation. McCarthy’s introduction of the term “artificial intelligence” provided a name under which several approaches to machine intelligence could be investigated together. (www-formal.stanford.edu)
The proposal treated the possibility of precisely describing and mechanically simulating intelligence as a research conjecture, not an established result. Its subjects included machine use of language, abstraction, problem-solving, self-improvement, neural networks, and the efficiency of computation. The resulting workshop became an important institutional landmark in AI, although its ambitious research program extended far beyond what could be accomplished during one summer. (www-formal.stanford.edu)
Lisp and symbolic computation
McCarthy developed Lisp in the late 1950s, with implementation beginning at MIT in autumn 1958. His paper “Recursive Functions of Symbolic Expressions and Their Computation by Machine, Part I,” published in April 1960, presented the language as both a practical programming system and a mathematical formalism. Unlike languages oriented primarily toward numerical calculation, Lisp made symbolic expressions and list structures central objects of computation. (www-formal.stanford.edu)
Its design combined recursive function definitions, conditional expressions, and operations for constructing and selecting components of lists. Programs could themselves be represented as Lisp data, allowing software to manipulate expressions with the same mechanisms used for other symbolic information. These features made the language useful for developing systems that processed formulas and other structured representations. (www-formal.stanford.edu)
Lisp also incorporated garbage collection, which automatically reclaimed memory no longer reachable from active program structures. McCarthy’s account describes the selection of this approach as an alternative to explicitly deleting structures or maintaining reference counts. The language’s early development was collaborative: Steve Russell recognized that McCarthy’s evaluation function could serve as an interpreter and implemented it. Lisp subsequently became an important language for AI research and developed along multiple independent paths. (www-formal.stanford.edu)
Logical AI and commonsense knowledge
McCarthy was a major contributor to symbolic AI, particularly approaches grounded in logic. In “Programs with Common Sense,” presented in December 1958, he proposed the Advice Taker: a program that would represent knowledge as statements and derive conclusions about what actions to perform. The proposal distinguished knowledge expressed declaratively from information embedded within a particular program’s procedures. (www-formal.stanford.edu)
This approach helped establish knowledge representation and reasoning as a central AI research area. General knowledge and facts about a particular situation would be represented explicitly, allowing a reasoning system to combine them when deciding what to do. Commonsense reasoning, however, introduced difficulties involving incomplete information, unstated exceptions, and the consequences of actions. McCarthy’s work sought formal mechanisms for addressing these problems rather than assuming that ordinary deduction alone was sufficient. (www-formal.stanford.edu)
In their 1969 paper “Some Philosophical Problems from the Standpoint of Artificial Intelligence,” McCarthy and Patrick J. Hayes developed the situation calculus, a logical framework for representing actions and changing properties. They also formulated the frame problem: how to specify what remains unchanged after an action without enumerating an unwieldy collection of separate statements. These questions connected formal reasoning with the requirements of systems that select and execute actions. (www-formal.stanford.edu)
Nonmonotonic reasoning and circumscription
McCarthy’s research on nonmonotonic reasoning addressed conclusions that may need to be withdrawn when additional information becomes available. In classical logical deduction, adding premises does not invalidate an existing consequence. Commonsense reasoning often behaves differently: an assumption justified by the available facts may cease to be justified when an exception is discovered. (www-formal.stanford.edu)
His circumscription formalism, published in 1980 and extended in 1986, expressed certain default assumptions through minimization of selected predicates. It could support reasoning that no additional exceptions or relevant entities should be assumed beyond those required by the stated facts. McCarthy used such methods to investigate the qualification problem—the difficulty of listing every condition necessary for an action to succeed. (www-formal.stanford.edu)
Interactive computing
McCarthy also helped advance computer time-sharing, in which multiple users interact with a shared computing resource. A memorandum dated January 1, 1959, proposed time-sharing to the director of MIT’s Computation Center. This work concerned the practical organization of computing as well as AI: researchers needed interactive access to machines for developing and testing programs. His historical writings document both this proposal and his recollections of time-sharing’s development. (www-formal.stanford.edu)