The Lighthill Report

The Science Research Council, wanting an unbiased view of a field it was funding, commissioned it from a famous scientist who worked in a different one. Sir James Lighthill's "Artificial Intelligence: A General Survey" was published early in 1973 in a Science Research Council volume titled Artificial Intelligence: a paper symposium, alongside replies from Stuart Sutherland, Roger Needham, Christopher Longuet-Higgins and Donald Michie — the volume's own structure conceding that the survey would not go unanswered. Its verdict on the field's record: "In no part of the field have the discoveries made so far produced the major impact that was then promised."

The report's structure was its argument. Lighthill divided AI into three categories: A, advanced automation — machines doing recognizably useful industrial, military, mathematical or scientific work; C, computer-based research on the central nervous system; and B, deliberately labelled a bridge — building robots — whose justification depended on the traffic it carried between the other two. A and C, he found, deserved support on their own merits. B, the part that was distinctively AI, he judged a failure, and the mechanism he named was the "combinatorial explosion" — the explosive growth of search that in his account stands as a general obstacle to any self-organising system working over a large knowledge base.

John McCarthy's review says exactly what the classification cost: "his three categories have no place for what is or should be our main scientific activity - studying the structure of information and the structure of problem solving processes independently of applications and independently of its realization in animals or humans." And the consequence: "missing these particular points is a consequence of missing the existence of the AI problem as distinct from automation and study of the central nervous system." The two were not arguing about results. They were arguing about whether the thing named at Dartmouth in 1956 existed as a science at all.

The argument then did something no AI dispute had done before or has done since: it played out on national television. The BBC broadcast a debate from the Royal Institution in June 1973 — Lighthill on one side, Donald Michie, Richard Gregory and John McCarthy on the other. The recording runs 81 minutes and survives; Edinburgh's AI institute hosts it.

The consequences are what the report is remembered for, and the archive that now hosts the full text — the Chilton Computing history site, a computing-heritage project, not the funding body — states them plainly: the report "provoked a massive loss of confidence in AI by the academic establishment in the UK including the funding body," with effects that persisted "for almost a decade." Recovery, by the same account, began only with the September 1982 Research Area Review Meeting on Intelligent Knowledge-Based Systems, whose findings fed the IKBS component of the Alvey Programme.

Two details keep the story from flattening into martyrdom. First, the report did not recommend abolishing the work: it supported research related to automation and to computer simulation of neurophysiological and psychological processes — categories A and C — and was "highly critical" only of the foundational middle. Second, nothing in it instructed anyone to stop funding that middle either. The collapse came from how a funding culture read one distinguished outsider's survey. A field, it turns out, can be defunded by taxonomy.