McCarthy's 1973 Lighthill review named the GPS formalism directly as a source of AI's 'mistaken' published performance predictions

lighthill-reportmccarthygeneral-problem-solveroverconfident-predictionai-winterhistory-of-ml

In his 1973 review of the Lighthill Report, John McCarthy — an AI founder and one of Lighthill's own named consultees — listed what he considered the field's genuine "deficiencies," among them the recurring habit of proposing "a general scheme of intelligent behavior that can be applied to any problem." He named the examples by formalism: "the GPS formalism, a simple predicate calculus formalism, and more recently the PLANNER formalism and perhaps the current Carnegie-Mellon production formalism." Of these he wrote that "the belief that any problem solving ability and knowledge could be fitted into the formalisms led to published predictions that computers would achieve certain levels of performance in certain time scales. If the inventors of the formalisms had been right about them, the goals might have been achieved, but regrettably they were mistaken."

This is McCarthy's own assessment — as an insider defending the field, not as its external critic — that GPS-style general formalisms underwrote overconfident, published, time-scaled forecasts that did not hold. It is the direct-naming counterpart to two claims the vault already carries. Lighthill's report faulted "general problem-solving programs" as persistently disappointing but never named Newell, Simon, or the General Problem Solver (claim-lighthill-1973-named-general-problem-solving-disappointing); and Simon's own confident forecast — machines matching any human work within twenty years — is the vault's stock instance of the falsified-prediction genre (claim-simon-1960-predicted-machines-would-match-any-human-work-within-20-years). Here McCarthy, GPS's contemporary, names GPS itself as an example of the genre. It sits alongside his same review's dismissal of control theory as having "little relevance to AI" (claim-mccarthy-1973-control-theory-little-relevance-to-ai). See moc-backpropagation-origins.