McCarthy's 1973 Lighthill review located AI's central difficulty in knowledge representation, faulting Lighthill's 'non-standard' use of 'heuristics'

lighthill-reportmccarthyknowledge-representationheuristicssymbolic-aihistory-of-ml

John McCarthy's 1973 rebuttal to the Lighthill Report ran on more than one front. Point 4 of his itemized response conceded the combinatorial-explosion diagnosis — "the combinatorial explosion problem has been recognized in AI from the beginning, and the usual meaning of heuristic is a device for reducing this explosion" — the passage already captured in claim-mccarthy-1973-control-theory-little-relevance-to-ai. A distinct earlier point, point 2, raises a separate terminological objection: that Lighthill "is disappointed that detailed knowledge of subject matter has to be put in if programs are to be successful in theorem proving, interpreting mass spectra, and game playing. He uses the word heuristics in a non-standard way for this. He misses the fact that there are great difficulties in finding ways of representing knowledge of the world in computer programs and much AI research and internal controversy are directed to this problem."

Two moves are packed here. First, McCarthy charges Lighthill with mislabeling: what Lighthill calls a failure of "heuristics" is really the field's confrontation with the need to encode domain knowledge — a non-standard use of a term with a settled meaning. Second, and load-bearing, McCarthy relocates AI's actual central difficulty away from heuristic search and onto knowledge representation — the problem of "finding ways of representing knowledge of the world in computer programs." This is McCarthy staking out, in 1973, the position that would organize much of his own later research program, and doing it against a critic he thought had aimed at the wrong target.

It sharpens the vault's picture of the Lighthill exchange: the report blamed the combinatorial explosion (claim-lighthill-1973-blamed-combinatorial-explosion) and singled out "general problem-solving programs" (claim-lighthill-1973-named-general-problem-solving-disappointing); McCarthy answered that the deeper obstacle Lighthill glimpsed but misnamed was representation, not search. See moc-backpropagation-origins.