At first, nobody noticed. There had been too many false dawns, broken promises, unrealised dreams. What was heralded as something that would change everything had changed almost nothing, except for the worse. Learning had long been confused with memorising, but now knowledge had dissolved into information, skill into technique, quality into quantity, and value into price. It was not reasonable to criticise anyone for cynicism because there was so much to be cynical about. Everything, almost, except the thing that changed education, and in changing education, changed everything.
What everyone expected and supposed was that to program a computer – actually rather a lot of computers – to be good at something one first needed to be good at it oneself. Good enough at least to know what it was that one was trying to program. And then it turned out that one didn’t. All one needed to be able to do – and of course this “all” is profoundly ironic – was to know how to enable the computer to learn. And once the computer could learn, it didn’t matter how good the programmers were; all that mattered was how good the computer’s learning had become.
Nobody noticed because everyone was so focused on the result and its implications for human thought that they missed the real point. AlphaGo had beaten one of the best Go players in the world by four games to one despite the fact that none of the people who programmed it was remotely good at the game. What they were good at was enabling the computer to learn, and programming what was needed to play the game, even if they were themselves incapable of playing it very well.
And that was really the message: all we need to be able to do is to discover what needs to be done in order to be good at something and then program it. Nobody needs to be able to do it; nobody even needs to be able to understand how the computer does it; we just need to know how to tell the computer to do it.
They quickly found, or actually to be honest stole from a little-known twentieth-century philosopher of science, a name for it: The Domain of Sophistication; the place or rather the territory where computers start to be able to do things better than we can do them and where ideas we understood to start with start to get so complex that we cannot understand them any more. Not just by calculating faster, or memorising more, but by being able to operate in territory of great complexity better than any human being can or ever could operate. And the thing that really delivered the killer-punch was that these systems were interactive: they not only learned from their human interactions; they learned even better from their interactions with themselves. They learned how to learn and how to teach learning by teaching themselves how to learn.
So, at first, nobody noticed. The odd, unpredictable move was viewed as a curiosity, an interesting intellectual challenge, viewed almost as we once entertained ourselves by watching physically deformed people in freak shows: how odd; how weird; how strange; how amusing; how inhuman. But if the system could learn to play one of the most sophisticated of games by teaching itself and playing itself, why could it not learn to do anything else that its creators thought sophisticated? In particular, why could something that could teach itself to learn by learning not also teach itself to teach by teaching? Why could not a system that could learn to play a game better than any human being by playing human beings and itself also learn how to teach a subject better than any human being by teaching human beings and itself?
So, at first, nobody noticed.