Machines Like Me

As for the intuitive artificial mind, it was pure urban legend, begun in early 1968 when Alan Turing and his brilliant young colleague, Demis Hassabis, devised software to beat one of the world’s great masters of the ancient game of go in five straight games. Everyone in the business knew that such a feat could not be accomplished by number-crunching force. The possible moves in go and chess vastly exceed the number of atoms in the observable universe, and go has exponentially more moves than chess. Go masters are unable to explain how they attain their supremacy beyond a profound sense of what feels right for any given situation on the board. So it was assumed that the computer was doing something similar. Breathless articles in the press announced a new era of humanised software. Computers were on the threshold of thinking like us, imitating our often ill-defined reasons for our judgements and choices. In a counter-move and in a pioneering spirit of open access, Turing and Hassabis put their software online. In media interviews they described the process of machine deep learning and neural networks. Turing attempted layman explanations of the Monte Carlo tree search, an algorithm elaborated during the forties’ Manhattan Project to develop the first atom bomb. He became famously irritable when he attempted, overambitiously, to explain PSPACE-complete mathematics to an impatient television interviewer. Less well known was his loss of temper on an American cable channel as he described a problem central to computer science, P versus NP. He was in front of a combative studio audience of ordinary ‘folk’. He had recently published his solution, which mathematicians around the world were then checking. As a problem it was easy to state, formidably difficult to resolve. Turing was trying to suggest that a correct positive solution would initiate exciting discoveries in biology as well as in concepts of space and time and creativity. The audience did not share or understand his excitement. They had only a dim awareness of his role in the Second World War, or of his influence on their own computer-dependent lives. They regarded him as the perfect English gentleman egghead and enjoyed tormenting him with stupid questions. The unhappy episode marked the end of his mission to popularise his field.

Before the confrontation with the nine-dan Japanese go master, the Turing–Hassabis computer played thousands of matches against itself continuously for a year. It learned from experience, and it was the scientists’ claim – reasonable enough – to have come a step closer to approximating human general intelligence that gave rise to the legend of machine intuition. Nothing they said could bring the untethered story back to earth.

Commentators who suggested that the computer’s victory would make the game extinct were wrong. After his fifth defeat, the elderly go master, helped by an assistant, stood slowly, bowed towards the laptop and congratulated it in a trembling voice. He said, ‘The mounted horse did not kill athletics. We run for joy.’ He was right. The game, with its simple rules and boundless complexity, became even more popular. As with the post-war defeat of a chess grandmaster, the triumph of the machine could not diminish the game. Winning, it was said, was less important than pleasure in the intricacies of the contest. But the idea that there was now software that could eerily, accurately ‘read’ a situation, or a face, a gesture or the emotional timbre of a remark was never dislodged and partly explained the interest when the Adams and Eves came on the market.

Fifteen years is a long time in computer science. The processing power and sophistication of my Adam was far greater than the go computer. The technology advanced and Turing moved on. He spent concentrated time looking at decision-making and wrote a celebrated book: we are disposed to make patterns, narratives, when we should be thinking probabilistically if we want to make good choices. Artificial intelligence could improve on what we had, on what we were. Turing devised the algorithms. All his innovative work was available to others. Adam must have benefited.

Turing’s institute drove forward AI and computational biology. He said he wasn’t interested in becoming richer than he already was. Hundreds of prominent scientists followed his example on open-source publication which would lead, in 1987, to the collapse of the journals Nature and Science. He was much criticised for that. Others said that his work had created tens of thousands of jobs around the world in diverse fields – computer graphics, medical scanning devices, particle accelerators, protein folding, smart electricity distribution, defence, space exploration. No one could guess the end of such a list.