ALPHAGO | AlphaGo | 2016
The Day a Machine Played the "Hand of God" / Google DeepMind "AlphaGo"
Seoul, March 2016. The world's strongest Go player, Lee Sedol, faced off against Google's Go-playing AI, AlphaGo, in a five-game match. Go had long been called a game AI could never solve. In just five days, that belief was turned upside down.
The problem
Go has far more possible board patterns than chess, and winning was said to come down to "feel" more than calculation. That's why it was considered humanity's last stronghold. AlphaGo had already beaten Europe's champion in 2015, but the real wall to climb was Lee Sedol, an 18-time world champion.
The insight
What DeepMind did wasn't make the machine calculate every correct answer. Instead, they had it study past games, let AlphaGo play on its own, and learn from winning and losing. It started by imitating humans, then went on to discover moves no human would ever make. It was about giving a machine a "feel" for Go.
The idea
On move 37 of the second game, AlphaGo played a move no one would ever play. Commentators and spectators alike were stunned, and that single move decided the match. Lee Sedol later called it "incredibly creative and beautiful." In the fourth game, Lee Sedol struck back with an equally unprecedented move of his own, but AlphaGo ultimately won the series 4 games to 1.
What happened
Watched by 200 million people, the match won the Grand Prix in the Innovation category at Cannes Lions 2016. After the match, Lee Sedol said he "felt powerless," and later retired from professional play. Meanwhile, the unconventional moves AlphaGo made are now standard strategy in the world of Go.
Afterthought
Ask me if this counts as advertising, and honestly, I'm not sure. But it put the question "can a machine be creative?" in front of 200 million people. That alone carried more power than any slogan ever could. Even now, I feel like that 37th move is still asking me something.
Awards
- Cannes Lions Grand Prix 2016
