Gigaom AI Minute – March 12

Transcript

I have spent the past few days looking back at March 2016 when the world watched as AlphaGo defeated Lee Sedol in the game of Go. On this date, March 12, the third game of five was played.

At the beginning of Game 3, a few people still held out that AlphaGo just got lucky a couple of times. In the first game, perhaps Lee made some early errors. In the second game, AlphaGo had stumbled onto a couple of lucky moves. Game 3 removed all of these doubts. AlphaGo won so definitively and so strongly, and played with such force and vision, that it was clear that it simply was the best Go-playing entity in the world.

Judging by the commentary on television and the online posts of the professional Go players, any hope that the situation could be salvaged seemed lost. Lee was defeated in three successive games of Go when just four days earlier the idea that he would lose even one seemed unthinkable.

It was clear that the public wasn't rooting for AlphaGo, and if you read the comments by the AlphaGo team members themselves, even they had a certain melancholy associated with their victories. One cannot help but wonder what would come next. Would a system like AlphaGo master language better than humans? Or write better music or create better art? Would it make scientific discoveries without us, discoveries that we would not even understand? While these ideas had always been fertile ground for science fiction, all of a sudden they seemed like something more. This was Go after all, not tic-tac-toe. If playing Go really is an art, then wasn't the computer better at that art than anyone alive? What would be beyond its grasp?

On top of that, the power of Moore's law doubles the capabilities of computers every 18 months. In two decades, we will have computers a thousand times more powerful than those upon which AlphaGo was built. What will they be able to do? These are questions we still cannot answer.