Machines

Imagine that a good friend of yours has fallen for a psychopath, a man or woman who seems superficially charming but you know to be controlling and cruel. The friend, reluctant to face reality, says: ‘Just give them time, they had a tricky childhood, they’ll change.’ Do you say: ‘Yes friend, hang in there’? No you don’t. You say: ‘Run for the hills, friend.’ Caring for someone who can’t care back, whose interest is only in exploiting you, is a mug’s game. So why – and several times in the course of reading Sherry Turkle’s Artificial Intimacy I put the book aside to blink and wonder – have we entangled ourselves so readily and intimately with uncaring AI?

Given the ingenuity of machine-makers, said Descartes in the 17th century, machines might well be constructed that exactly resemble humans. There would always, however, be ‘a reliable test’ to distinguish them. ‘Even the stupidest man’ is equipped by reason to adapt to ‘all the contingencies of life’, while no machine could ever be made with enough pre-set ‘arrangements’ to be convincingly versatile. But suppose it could? In 1950 Alan Turing proposed a test remarkably similar to Descartes’s. A computer and a human are asked questions, each being invisible to the questioner, and their respective responses compared.