Tonight I learned about the game Go courtesy of two documentaries on Netflix: AlphaGo followed by The Surrounding Game (I should’ve watched them the other way around). I still know basically nothing about the actual game, it just looks like dots on a board, but the stories were quite captivating.
The crux of it is that Go is a super complex ancient Chinese board game which is based on logic and really hard to predict. It’s a strategy game like chess, but has far more possible permutations (over 200 possible moves per turn compared to 35). So many that a computer can’t process all of the possibilities. It requires intuition and nuance, not just pattern recognition. That’s what made it a good experiment for the Google DeepMind team. They created an AI machine called AlphaGo and trained it using machine learning.
In 2016 AlphaGo took on the Roger Federer of Go, Lee Sedol from Korea. Best of 5 games.
First game the loss is a surprise for Sedol and a relief for the AI team. Sedol and most other people realise they underestimated the AI.
Second game Sedol goes in thinking he’s learned from the first and loses again. The computer makes some “beautiful move” that a human would never play. Everyone realises that AlphaGo isn’t just a human replica, it’s coming up with its own creative solutions that the humans didn’t teach it.
Sedol takes a day off. Third game is a shambles. He’s totally psyched out and not playing his usual game. He loses the tournament.
By this point in the film you’re starting to really get a feeling for the whole human vs machines thing, and just what it is that makes us human. The empathy that everyone is feeling for Sedol is palpable. The guy that makes the moves for AlphaGo points out how difficult it must be to play against a machine because you don’t have any connection with your opponent which makes it really easy to get caught up in your head and for self doubt to creep in.
In his press conference the language Sedol uses is that of a defeated man. One who doubts that he was ever that good in the first place. Who feels like a failure to himself and others.
Seeing Sedol’s deflated sense of self made me
feel almost as emotional as seeing the young girl with the physical disability sing When I Grow Up (Tim Minchin, Matilda) in the school assembly this morning. The audience obviously felt the same. So much so that the members of the AI team don’t even feel like celebrating.
Fourth game presents an opportunity for redemption or complete defeat. Redemption it is. Sedol makes the game so complicated that the AI doesn’t know how to respond to it. He wins! The emotional response of the crowd was immense. Sedol says “the victory was so wonderful that I wouldn’t exchange it for anything in the world”. Relief.
Fifth and final game. Everyone’s excited. The machine makes some unusual moves, playing the game differently to how a human would. Apparently humans use the progressive score as an indicator of whether or not they’re likely to win. But to the computer it doesn’t matter how much you win by, it only has to be by one point. This means AlphaGo will change the way the game will be played. Sedol is cool with the loss. He won once. That’s a victory for humans.
There were two key learnings from this for me. First, AI can give computers capacity to express creativity in a way that humans aren’t. This could be really helpful in tackling some of the wicked problems that our world is facing at the moment. Second, exposure to these types of machines can help us to better see what it truly is to be human. And more humanity could be useful in this world. Especially if we’re going to responsibly manage the implementation of this sort of technology into broader society