Warren Weaver

Warren Weaver is the most consequential forgotten architect of artificial intelligence. In the summer of 1949 he typed a four-page memo in a New Mexico hotel proposing that the new electronic computers might translate between human languages; the document became, by scholarly consensus, the single most influential early text in the field, and the line of descent runs straight from it to every large language model of today. But Weaver was no credulous prophet. The same man who carried Claude Shannon's information theory to a general readership insisted, with italics, that information “must not be confused with meaning”—a warning he planted in 1949 that has become the central unsolved problem of the twenty-first century. A year before the translation memo he published “Science and Complexity,” naming the great unexplored middle between tidy simplicity and pure randomness, which he called organized complexity—problems “interrelated into an organic whole” that neither classical equations nor raw statistics could reach. He could not have known he was