Textbook of AI

The ALPAC report (Automatic Language Processing Advisory Committee), published in 1966 by the US National Research Council, delivered a damning assessment of the previous decade's machine-translation research. The committee, chaired by John R. Pierce of Bell Labs and including David G. Hays and Charles F. Hockett, found that no general-purpose machine-translation system then existed; that experimental MT systems were slower and less accurate than human translators; that there was "no immediate or predictable prospect of useful machine translation"; and recommended that funding be redirected to basic research on computational linguistics and to support for human translators.

The recommendations were largely accepted. US government funding for MT was sharply cut, and active MT research virtually ceased in the United States for nearly a decade. The episode prefigured the Lighthill report's (1973) similar effect on UK AI research seven years later, and is generally considered the first AI winter, though localised to MT rather than affecting AI more broadly.

The ALPAC report's arguments, that practical MT requires deep semantic understanding rather than just lexical and syntactic processing, have been vindicated repeatedly. The eventual triumph of statistical and then neural MT in the 2000s and 2010s was achieved by sidestepping rather than solving the problem the report identified, by learning from massive parallel corpora rather than by deep semantic understanding.