last night google’s alpha go beat the second best human go player in the world, Lee Sedol last night the first game of this 5 game match was held in south korea, so it starts at 11pm EST, hard to watch it live without falling asleep.

Go players around the world held this go program in contempt. Go can’t be played by computers. Go is artistic and requires human ideas. Come on go players. Humble yourself. You really thought humans would beat computers forever?

How this program works is ridiculous. there was no way that the programmers could find to beat humans, straight up calculation doesn’t work (like used to beat Kasparov in chess 20 years ago) the calculations get too complicated, while the random walk monte carlo method tends to create weak moves in the middle game.

What did they do? They created a neural network. An artificial human brain. Then they told the computer to play itself millions and millions of times. This computer taught itself how to win at go. Remarkable really.

In the actual game, Lee Sedol who is a very aggressive player, he likes to battle early in the game took the game out of book early with a play close to the edge on the 3rd move, he seemed to be testing alphago. Alpha go then responded by fighting for the center and Sedol definitely looked to have an advantage pursuing the upper right corner. Quickly the game turned into a huge battle with both players fighting, you know how in go you can play tactically or strategically, right away Sedol took the game into a tactical battle and alpha go obliged. The game become terribly complicated and in retrospect both the human and the computer began making some small errors. Sedol then in the middle game made 2 or 3 more errors than the computer did. When the game wound down to the ending, it wasn’t clear until just before the final ending play that alpha go was ahead and Sedol resigned immediately instead of playing it out.

Sedol never thought that the computer would be this strong. That it would play “like a human” instead of like a computer. What Sedol never counted on is that 5 months after the last time alphago played a human it had played itself millions more times and had become much stronger.

Two conclusions: Good for you computers! Chess is my game, not go, and I am glad that the computers came for you. I have had to suffer the intolerable airs of humans proclaiming go is better than chess because only humans can win at go. So good for you alpha go. Humble those go players.

Second conclusion. This is going to be really good for go. You would think that when computers beat humans it means that humans will just give up. That’s not what happened for chess when kasparov lost. paradoxically it helped chess. Something about not having to worry about computers any more gave chess a huge surge. Now people only play people. And people turn to computers to help them train. So all around losing to computers will in the long run only help go become more popular.

I’ll be rooting for the computers for the rest of this match. Hail computers. Beat the human!

You might think I have some sympathy for Lee Sedol not a chance. Before the match he said he would crush alpha go 5-0. So I love seeing him get humbled.

Advertisements