After the Second World War, automatic translation - especially Russian-to-English - attracted heavy US government funding on the promise that computers would soon translate documents faster and cheaper than people. By the mid-1960s the agencies paying for it wanted to know whether the promise was being kept. The National Academy of Sciences and National Research Council convened the Automatic Language Processing Advisory Committee (ALPAC) to evaluate the state of the work, and in 1966 it published its report, “Language and Machines: Computers in Translation and Linguistics.”
The verdict was blunt. The report found “no immediate or predictable prospect of useful machine translation” and argued that human translation was faster, more accurate, and cheaper than the machine systems then available. It recommended redirecting money away from machine-translation systems and toward more basic research in computational linguistics. Whatever its fairness - critics then and since have argued it underweighted long-term research and judged the field by an unrealistic near-term standard - its practical effect was immediate and severe.
The funding that had sustained machine translation in the United States largely dried up, and the field went quiet for roughly two decades. This is the earliest clear case in AI’s history of a funding collapse driven by a gap between promise and delivery, and it predates the phrase “AI winter,” which would not become common until the 1980s. The ALPAC episode established the pattern those later winters would follow: oversold capabilities, a sober external review, and a sharp withdrawal of support.
The report is worth reading as a primary document precisely because its reputation is contested. It is often described as the report that “killed” machine translation, and in funding terms that is roughly true; but the committee did not claim translation was impossible, only that useful systems were not close and that the money was better spent elsewhere. The distinction matters for anyone reading today’s confident timelines: the question ALPAC asked - is the capability actually here yet, or just promised - is the same question every hype cycle eventually has to answer.