We live with machines. And our machines are getting smarter. They’re still very dumb, they do what we tell them to, and often not really all that well. But we’re teaching them. And I do mean “we.” When you tag your friends on Facebook, you’re teaching its facial recognition system what to look for in faces. When you click on ads or Google results you’re training its system to know what you like. Computer scientists are working on ways to make machines learn smarter, learn passively, make connections we never even thought of.
Alpha Go, an artificial intelligence trained to play Go, recently beat a Go grandmaster. This is a big deal, partially because Go is an extremely complicated game, but also because the system did things while playing that nobody expected. It made moves that seemed like mistakes, and twenty moves later were suddenly clearly not mistakes. Now imagine your computer could surprise you in that same way.