In Part Five of his “Discourse on the Method” (1637), Rene Descartes drew a sharp line between bodies and minds. He argued that the bodies of animals, and even much of what the human body does, are nothing more than elaborate machines, organs and fluids arranged like the gears of a clock. Animals, on this view, are automata: they react, but they do not think. Living in an age fascinated by mechanical figures and moving statues, Descartes took the comparison seriously and pushed it as far as it would go.
But he stopped short of saying a machine could ever be a genuine mind, and to mark the boundary he offered two tests, remarkable for anticipating debates that would only become concrete three centuries later. First, he said, a machine could never “use words or other signs arranged in such a manner as is competent to us in order to declare our thoughts to others,” meaning it could not use language flexibly and appropriately across situations. Second, even if a machine did some things as well as or better than we do, it would inevitably fail at others, because it acted from the arrangement of its parts rather than from understanding, whereas reason “is a universal instrument that is alike available on every occasion.”
The episode is a striking piece of AI prehistory because Descartes framed the core question, what would distinguish a real mind from a very good imitation, and proposed language and general, flexible reasoning as the deciding criteria. When Alan Turing proposed his imitation game in 1950, he was in effect taking up Descartes’s first test and turning it into an operational experiment. The primary source used here is the Gutenberg edition of the Discourse.
For a general reader, Descartes shows that the modern worry about telling humans from convincing machines is not new; it was posed clearly the moment people started taking mechanical imitations of life seriously.