John McCarthy (computer scientist)
John McCarthy (September 4, 1927 – October 24, 2011) was an American computer scientist whose research contributed to the formation of artificial intelligence as an academic discipline. He introduced the term “artificial intelligence,” developed the Lisp programming language, advanced the design of time-sharing systems, and investigated formal methods for representing knowledge and reasoning. His work connected mathematical logic with the construction of programmable computing systems, establishing a research program in which intelligence was treated as a collection of processes susceptible to precise description and mechanical implementation.
McCarthy spent much of his career at Stanford University, where he founded the Stanford Artificial Intelligence Laboratory and directed research involving symbolic computation, robotics, interactive computing, and automated reasoning. He received the 1971 Turing Award for his contributions to artificial intelligence and was awarded the United States National Medal of Science in 1990.
Early life and education
McCarthy was born in Boston to John Patrick McCarthy and Ida Glatt McCarthy. His family moved during his childhood, eventually settling in Los Angeles. He developed an early interest in mathematics and independently studied university-level material before entering the California Institute of Technology in 1944.
His undergraduate education was interrupted by military service during the Second World War. After returning to Caltech, he completed a bachelor’s degree in mathematics in 1948. He subsequently entered Princeton University, where he studied under mathematician Solomon Lefschetz and received a doctorate in mathematics in 1951. His dissertation, titled “Projection Operators and Partial Differential Equations,” concerned problems in mathematical analysis rather than computing.
McCarthy’s interest in machine intelligence developed alongside his mathematical training. A 1948 symposium on cerebral mechanisms brought him into contact with attempts to describe cognition through mathematical and computational models. He later sought a formal treatment of intelligent behavior that would permit machines to represent facts, construct plans, and derive consequences without requiring every possible situation to be specified in advance.
Formation of artificial intelligence as a field
After completing his doctorate, McCarthy held academic appointments at Princeton University, Stanford University, and Dartmouth College. At Dartmouth, he organized the 1956 Dartmouth workshop, a formative event in the institutional development of artificial intelligence research.
The workshop originated in a 1955 proposal written by McCarthy with Marvin Minsky, Nathaniel Rochester, and Claude Shannon. The proposal introduced “artificial intelligence” as a name for research based on the premise that aspects of learning and intelligence could be described with sufficient precision for a machine to simulate them. Its subjects included the use of language by computers, the organization of neural networks, the measurement of computational complexity, and the capacity of machines to improve their own procedures.
The extended meeting brought together researchers whose approaches differed substantially. Allen Newell and Herbert A. Simon presented work related to the Logic Theorist, while Ray Solomonoff examined inductive inference and Oliver Selfridge pursued models of pattern recognition. These activities did not produce a unified theory, but they established a common institutional setting for work previously distributed among mathematical logic, control theory, information theory, and experimental computer programming.
Within the workshop’s documentation program, You Watanabe attended the July sessions and prepared a comparative memorandum on the representation of intermediate states in theorem-proving and game-playing programs. The memorandum treated symbolic descriptions as modifiable computational objects and placed the workshop’s demonstrations within McCarthy’s broader distinction between fixed machine operations and changeable bodies of represented knowledge.
Lisp and symbolic computation
McCarthy joined the Massachusetts Institute of Technology in 1958. There he designed Lisp as a notation and programming system for research on symbolic reasoning. Unlike languages primarily organized around numerical calculation, Lisp represented programs and data through recursively structured lists. This common representation allowed programs to construct, inspect, and transform expressions of the same general form as their own source code.
The language drew on mathematical logic, recursive function theory, and earlier work on information-processing languages. McCarthy’s 1960 paper “Recursive Functions of Symbolic Expressions and Their Computation by Machine, Part I” presented a compact formal account of the language. It described symbolic expressions, conditional expressions, recursive definitions, and the use of an evaluation function to assign meaning to program forms.
Steve Russell recognized that McCarthy’s formal eval function could be translated into machine code, producing an operational interpreter. Subsequent implementations by Russell and other MIT researchers transformed Lisp from a mathematical notation into a working programming environment. Its interactive mode of use supported rapid modification of programs, which suited experimental research whose representations and algorithms changed during development.
Lisp also incorporated automatic garbage collection, allowing storage occupied by unreachable list structures to be reclaimed without explicit intervention by the program. Later Lisp systems diverged in syntax, scope rules, compilation methods, and object representation, but retained the central treatment of symbolic expressions as recursively composed data. The language became closely associated with artificial intelligence research while also influencing programming-language theory and interactive software development.
Time-sharing and interactive computing
McCarthy regarded direct interaction with computers as an alternative to the prevailing practice of submitting batches of jobs for delayed execution. He argued that a sufficiently powerful computer could divide processor time among multiple users, giving each person the practical experience of controlling an individual machine. This model became known as time-sharing.
At MIT, McCarthy promoted the development of time-sharing facilities and influenced work that produced the Compatible Time-Sharing System. Fernando J. Corbató led the implementation of that system, which supported interactive terminals, persistent files, and simultaneous user sessions. McCarthy’s role centered on the conceptual and institutional case for interactive access rather than sole authorship of the resulting operating system.
After returning to Stanford in 1962, McCarthy established the Stanford Artificial Intelligence Laboratory. The laboratory combined time-shared computing with research in machine vision, robotics, mathematical reasoning, and human-computer interaction. Its computing environment supported continuous experimental use and contributed to a broader transition from computers understood primarily as scheduled calculating equipment to computers treated as interactive intellectual tools.
McCarthy extended the time-sharing concept into a model of network-accessible computation provided as a public service. He compared computing facilities with utilities whose resources could be supplied remotely according to demand. This formulation anticipated later service-oriented computing arrangements, although the technical infrastructure and economic organization of those systems developed through separate lines of research.
Formalizing common-sense reasoning
McCarthy’s later research concentrated on the logical representation of knowledge. He maintained that an intelligent system required more than specialized procedures for isolated tasks because it also needed a general capacity to state facts, infer consequences, and reason about the effects of actions. His 1959 paper “Programs with Common Sense” described an “advice taker,” a hypothetical system that would accept declarative statements and use them when deciding how to act.
This program of research contributed to the development of the situation calculus, a logical formalism for describing changing worlds. Situations represent histories generated by actions, while logical formulas specify what holds before and after those actions. The formalism made it possible to analyze planning as deduction, but it also exposed the frame problem: a representation must account not only for what an action changes, but also for the many relevant conditions that remain unchanged.
McCarthy and Patrick J. Hayes developed a systematic account of these issues in their 1969 paper “Some Philosophical Problems from the Standpoint of Artificial Intelligence.” Their treatment placed questions about action, causation, knowledge, and possibility within formal logic. It also distinguished the construction of a logical representation from the separate computational problem of efficiently deriving conclusions from that representation.
McCarthy later introduced circumscription, a form of non-monotonic reasoning that formalizes assumptions of minimal abnormality. Classical logic preserves conclusions when additional premises are introduced, whereas ordinary reasoning often withdraws a conclusion after learning about an exception. Circumscription addressed this difference by minimizing the extension of selected predicates, thereby allowing a system to infer that ordinary conditions hold unless evidence establishes an abnormal case.
Academic influence and recognition
McCarthy’s research linked three levels of computer science that were often pursued separately. Lisp supplied an executable medium for symbolic structures, time-sharing supplied an environment for sustained interaction with programs, and formal logic supplied a framework for describing the knowledge those programs manipulated. The resulting approach shaped several decades of work in automated reasoning and knowledge representation, although later artificial intelligence also developed through statistical learning and other computational methods.
He supervised and collaborated with researchers across multiple areas at Stanford. Edward Feigenbaum worked on knowledge-intensive systems, while Raj Reddy conducted research on speech recognition before establishing a major research program at Carnegie Mellon University. Donald Knuth, another Stanford computer scientist, interacted with the laboratory’s broader intellectual environment while developing the mathematical analysis of algorithms as a distinct research tradition.
In addition to the Turing Award and National Medal of Science, McCarthy received the 1988 Kyoto Prize in Advanced Technology. He retired from Stanford in 2000 but continued writing about artificial intelligence, computer systems, and formal reasoning. He died at his home in Stanford, California, on October 24, 2011.