On 10 February 1996 the IBM supercomputer Deep Blue beat the world chess champion Garry Kasparov in thirty-seven slow moves. Two decades later, in 2016, the British firm DeepMind – which two years before had been acquired by Alphabet, Google’s parent company – developed AlphaGo, a program which in Go defeated Lee Sedol, eighteen times world champion of the mythical Chinese board game.
On the surface this may sound like the inevitable technological evolution of digital brains, but it conceals a much larger milestone: their automatic ability to learn. The major difference between the IBM computer and the program AlphaGo is that whereas the former works through a vast database of chess matches and a huge capacity to use statistical calculation to choose the best movement, the latter – even with a similar takeoff point – learns from every move by infinitely combining it with all the moves it has already memorized. This progressively expands its knowledge of the game, so the more it plays, the better it becomes, and most importantly, with no need for human intervention...[+]