Humans will add to AI’s limitations

IN 1958 A psychologist and computer-science researcher named Frank Rosenblatt gave a public demonstration of his Perceptron, the distant ancestor of modern machine-learning algorithms. The Perceptron had been developed on a 9-tonne IBM 704, a mainframe computer with less power than a modern television remote control. Its party trick was its ability to learn, without any direct programming, to recognise cards printed on the left from those printed on the right.