It doesn’t sound like it means much to declare that Lee Se-dol has announced his retirement as one of the top Go players the world has ever seen.
Perhaps it isn’t his retirement that matters. But his reasoning.
Go, for those who are unaware, is an ancient Chinese abstract board game that far exceeds chess in both simplicity and complexity. By that I mean that the rules are far simpler, but the game itself is magnitudes more difficult to understand and master. It has been said there are more positions that you can have on a Go board than there are atoms in the universe. Without going into detail, I will vastly simplify the concept and tell you it is about placing dots on a grid to contain larger areas than your opponent.
Lee Se-dol is the only human to have ever defeated AlphaGo, an AI Go player created by Deepmind, a subsidiary of Google. In the latest tournament, AlphaGo won four out of the five games played.
And Lee Se-dol quit playing Go.
His reasoning is that “there is an entity that cannot be defeated”.
In 1986, I was working for a company called Institutional Computer Development Corp, one of five or six employees at a place that essentially sold chess computers and helped develop new software. Now it’s a gamestore and website called USAChess.com. This was before home PCs had the necessary horsepower to run chess programs to any decent strength. And each year, computers got faster and programmers pushed further. Chess became closer and closer to being solved. We had no doubt that there would be computers no one could beat.
Go was something else. The play was almost artistic and intuitive. This was not something where a human being could be beaten just by the sheer math power of a computer. The depth of play dwarfed chess tremendously. For your first two moves in chess, there are 400 possible positions. In the first two moves of Go, there are over 150,000 possible positions. If we remove symmetrical rotation of the board, still almost 38,000 possible board configurations. After a dozen moves, chess is something that a computer can power through seeing each possibility, A dozen moves in Go… that number becomes tremendous beyond understanding.
That’s why we weren’t sure a computer could ever beat the top players. There was an almost mystical quality as to the way top Go players played. It wasn’t calculation, it was more.
in 1996, it happened. Chess champion Garry Kasparov lost a game to Deep Blue, sending shockwaves through the chess and chess computer communities.
As a guy selling chess computers, I watched a very weird phenomena happen.
People would come into the retail store and say it was pointless, now that a computer had won. They felt no reason to play. This was stunning for many, many reasons.
Reason 1: Most people couldn’t beat the computers we were selling as it was. Why would you change your own appreciation of the challenge of the game because someone else had failed to beat another opponent? No one cared if Karpov beat Kasparov. It didn’t make people suddenly quit. But when a non-biological won at a game far above the level of play of most players in the world, for some reason, people felt it changed their game.
Reason 2: Race cars exist. And yet people still sprint. The contest is not how fast can an object make it to the finish line, it is how fast a human can.
Reason 3: We only improve through learning from our mistakes. Our defeats. If an AI comes along that can beat you, don’t you try to step up, further your game, find a weakness and attempt to beat it?
No one is going to run as fast as a race car. And still they train. Very few people will compete in the Olympics. And still, people run.
Finding “An entity” you cannot defeat does not destroy the game. The game is a social interaction and challenge between humans. Whether or not you can defeat the top non-human is inconsequential. No one is running at the landspeed record of 763MPH.
I enjoy running. I’m not great at it; I don’t compete. Why would I let the fact that people or machines are faster than me take away from my enjoyment of running? I do it to improve myself and because I enjoy it.
So, Lee Se-dol. If somehow this reaches you. Don’t quit. And when a human beats you consistently, congratulate them. If you are inspired to come back and beat them, or try and beat an unbeatable computer, do so. Look at AlphaGo in awe as an achievement in human understanding of intelligence. Learn from it.
You may never beat it.
This does not mean that you have been beaten.