that self-help guide regarding girls that want the rules dating book for you to lift your man.
Its origins are quite recent. Although it was first conceived in the 1950s, the technique did not work well when applied to the real world. So it was thought to be a failure. But an intellectual and technical revolution has occurred in the last decade as researchers have made promising progress using machine learning. What was missing before was data in sufficient quantities. Now that there are, the method works. Today machine learning is the foundation for everything from internet search engines, online product recommendations, computer language translation and speech recognition, and much more.
To understand what machine learning is, it helps to know how it came about. In the 1950s, an IBM computer programmer named Arthur Samuel programmed a computer to play chess. But it wasn’t a very fun game. Samuel always won, because the machine only recognized legal moves. He knew some strategy, so he developed a clever applet that, with each move, calculated the probabilities that a given board configuration would lead to a game winning or losing.
But a game between man and machine still didn’t work out well; the system was in too embryonic a state. So Samuel let the machine play against itself. In doing so, he collected new data. By gathering more data, the accuracy of their predictions improved. So he played against the computer and lost. And again. The man had created a machine that surpassed him in skill in a task that he himself had taught him.
Similarly, why do we have self-driving cars? Is the software industry better at including all traffic regulations in one code? No. Is it because of the increase in computer memory? Neither. Faster processors? No. Smarter algorithms? Not again. Cheaper chips? Neither. All of this helped, but what really made the innovation possible was that technology experts changed the nature of the problem. They turned it into a matter of data: instead of trying to teach the car to drive – something difficult, because the world is a complex place – the vehicle collects all the data that surrounds it and tries to deduce itself what it has to do: that there is a traffic light, that it is red and not green and that this means that the car must stop.