The Ancient Game of Go: Human vs. Artificial Intelligence

Google's neural network DeepMind, running the software AlphaGo, has secretly won over 60 games online played against some of the world's top Go players (REUTERS/Google/Yonhap) Image source: ibtimes.co.uk

In March of 2016, it was said that there was an unexplainable sadness that the groundbreaking artificial intelligence program, AlphaGo beat Lee Sedol, the world’s best player over the last decade,  in the ancient Asian strategy board game, Go.

After chess, Go became the next Artificial Intelligence frontier. Yes, the game is simple enough and can easily be learned in a matter of a day. However, because of Go’s simple rules yet elaborate possibilities, no computer could have competed on the professional player level and had become one of the most sought-after field of AI development obsession.

Not So, Black And White

Go starts with an empty board, 19-by-19 for those who intend to play a full game. One player plays with black stones while the other plays with white. Placing one of their stones on a vacant point, players take turns with the object of the game to control more territory. Stones do not move once played but can be taken off the board if captured. Whoever controls the most territory at the end of the game wins.

Go has existed for two to three millennia and legend has it that the Chinese Emperor Yao invented it as a means to teach his son discipline, concentration, and balance. Another myth surrounding its origins is that it was used by Chinese warlords to map out attacks using the pieces of stones to signify positions. Whatever its true origins, the game of Go hasn’t changed since it was first documented. The rules are minimal; the aim is deceptively simple. However, it differs from games like chess as it has been described as more of a game of instinct rather than strategy.

Go game board much more complex than chess with more than 200 possible moves for every turn. Image source: inspirationroom.com

A Beautiful “Deep” Mind

Perhaps this is why AI AlphaGo’s triumph over its human opponent is so fascinating and a notable achievement. AlphaGo combined the old technologies with new algorithms, developing its own intuition. With the deep learning technique, AlphaGo’s “DeepMind” — a British AI company founded in 2010 and acquired by Google in 2014 — discovered for itself how a human player would execute moves and countermoves, acquiring the kind of wisdom that professional Go players take years of practice to accumulate.

“It’s not a human move. I’ve never seen a human play this move,” Fan Hui; three-time European Go champion said as he watched the 37th move in the second game of the match between AlphaGo and Lee Sedol. “So beautiful!”

Fan Hui kept repeating the word “beautiful” as the game approached the end of its first hour at the Four Seasons hotel in Seoul. To the shock of the entire audience, AlphaGo directed its human assistant to place a black stone in an area on the right-hand side of the 19-by-19 grid. The move prompted Lee Sedol to stand up and leave the match room, perhaps to recover. When Lee Sedol returned, it would take him a whole 15 minutes to consider his next move.

Ultimately, AlphaGo won the game after the next 3 hours, becoming another machine to have beaten the best humans who were considered professionals in games like chess and checkers. But as everyone can recognize, Go is a far more complex game than the mentioned making this win historical. The sadness felt by so many humans that day may have been the realization of our inferiority and weakness compared to a machine that was human-made in the first place.

Facebook
Twitter
LinkedIn
WhatsApp
Email

Leave us your question.