The Prehistory of A.I. Slop

During the Cold War, a period when computation and linguistics converged, having a machine that could read was useful when you were spying on your enemies. That it could learn to write was a bonus.Photo illustration by Jack Smyth; Source photograph from Getty Save this story Save this story Save this story Save this story In 1962, a programmer at Librascope, a California-based defense contractor, announced that “a computer can be programmed to write meaningful and relevant sentences in proper English.” At Librascope’s Laboratory for Automata Research, in Glendale, he’d started out by feeding into his computer—the vacuum-tube LGP-30—a vocabulary of thirty-five hundred words and a repertoire of a hundred and twenty-eight sentence patterns, and told it to do, more or less, what humans did in the nineteen-nineties when they stuck Magnetic Poetry on the doors of their refrigerators. And behold! “Broccoli is often blind,” the LGP-30 tapped out on its typewriter, and “Communism is more porcelain than albino gold.” The engineer decided to set this machine-generated text as free verse:

Was Milo mewling thrilling radishes?So, our anchovies are sad but green.

He called his program the Auto-Beatnik, cunningly deploying Kerouac, Ginsberg, and Burroughs as cover for this bilge. The ploy occasionally paid off; the London Daily Mirror described the Auto-Beatnik’s poetry as “better than most of the stuff that gets published in avant garde magazines.”