The Lighthill report

The Science Research Council asked Sir James Lighthill, a fluid dynamicist of great reputation and no involvement in the field, to assess British research in artificial intelligence. His 1973 report divided the subject into three parts: automation, which was useful; the study of the nervous system, which was science; and a "bridge" between them, the general-purpose robot and the machine that understands, which he judged had produced nothing. The combinatorial explosion, he wrote, would defeat any programme that tried to reason about the world at large.

The report was debated on BBC television that summer, Lighthill against Donald Michie, John McCarthy and Richard Gregory, and the researchers lost in front of the public. The Council cut AI funding at all but two universities. In the United States the Defense Advanced Research Projects Agency reached a similar conclusion the next year, ending its speech-understanding programme and turning from open-ended grants to mission-directed contracts.

Lighthill was largely right about 1973 and wrong about the future, which is the usual shape of these reviews. The robots he doubted arrived fifty years later, running on the neural networks his contemporaries had just declared dead. The first winter lasted, by most reckonings, until the expert-systems boom of the early 1980s.