John McCarthy

In the AI Story

McCarthy's 1956 proposal for the Dartmouth Workshop articulated the paradigm's founding ambition with characteristic directness: that 'every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.' The sentence compressed the commitments Dreyfus would spend the next fifty years challenging—that intelligence is describable, that it can be reduced to precise specification, that simulation is the appropriate relation between the description and the phenomenon.

McCarthy's specific technical contributions were enormous. Lisp, which he developed at MIT in 1958, was the first programming language designed around symbolic computation and remained the dominant AI language into the 1990s. The situation calculus, introduced with Patrick Hayes in 1969, provided a formal framework for reasoning about action and change that also identified the frame problem as a formal obstacle. Circumscription, developed in the 1980s, was his attempt to address the common-sense reasoning problems Dreyfus had been identifying since 1965.

The philosophical disagreement between McCarthy and Dreyfus was genuine and deep. McCarthy believed that intelligence was in principle formalizable and that the only question was how to do the formalization. Dreyfus believed that intelligence was in principle not formalizable and that the attempt to do so rested on a philosophical error. Neither position admitted compromise, and the disagreement drove both men to refine their positions across decades.

McCarthy's response to the collapse of the symbolic paradigm and the rise of statistical methods was complex. He did not abandon his foundational commitments; he believed that statistical methods were complementary to symbolic reasoning rather than a replacement for it, and that genuine AI would require both. He died in 2011, before the large language model revolution, but his philosophical position on the foundations of intelligence remained unchanged.

Origin

McCarthy earned his PhD in mathematics from Princeton in 1951 and taught at Dartmouth, MIT, and Stanford over a career spanning six decades. He founded the Stanford Artificial Intelligence Laboratory in 1963 and remained its director until 1980. He received the Turing Award in 1971 for his contributions to AI.

His encounter with Dreyfus began in the 1960s and continued throughout both men's careers. The debates were conducted in formal settings (academic papers, conferences) and informal ones (personal correspondence, occasional public exchanges). Neither man was persuaded by the other, but both continued to engage because each recognized the other as serious.

Key Ideas

Coiner of 'artificial intelligence.' The term itself and the research paradigm it named were McCarthy's inventions.

Physical Symbol System commitment. McCarthy's foundational view was that intelligence consists of manipulating symbols according to rules, and that this view is empirically testable.

Technical rigor as philosophical response. McCarthy met Dreyfus's phenomenological objections with specific technical work designed to address them—the situation calculus, circumscription, nonmonotonic reasoning.

Productive opposition. The McCarthy-Dreyfus disagreement was one of the most sustained and productive philosophical debates in AI history, forcing both sides to sharpen their positions.