Skip to Content
AlphaGo

AlphaGo

The Great AI Upstart

Go is an ancient strategy board game commonly played in East Asia. Known for its simple rules but extreme gameplay, Go is played by children in special classes, in order to get a good understanding of the game later on. Players place stones on a 19×19 board to control territory, requiring deep strategy and foresight.

An event that shook the Go world was the Lee Sedol vs AlphaGo match, where Lee Sedol, considered to be one of the world’s greatest Go players, lost 4–1 in five rounds against AlphaGo, an AI program that uses deep neural networks and machine learning to play the game Go. But what was the story leading up to it, and what are the ramifications of such a colossal event?

First, an explanation. AlphaGo is an AI system that combines neural networks and a search algorithm. The two neural networks suggest intelligent moves to play and evaluate the position that is reached, while the search algorithm plays the remainder of the game in its “imagination” many times over. The programmers of AlphaGo taught it how to play Go by feeding it data from amateur games, so it could understand how the game was played. It then played against itself thousands of times and learned from its mistakes—a process called reinforcement learning.

Despite AI being prevalent in chess, many thought it would take another decade before AI could beat human players in Go. This is due to Go’s insurmountably large decision tree. With 10^170 possible board configurations, it would be impossible to process every possible option and take the best move. Before AlphaGo, the best AI for Go could only reach the level of human amateurs. AlphaGo was able to counter this by finding the best moves to play based on the neural networks, thus not needing to always calculate every option. 

In October 2015, AlphaGo played against its first professional player, Fan Hui, the reigning three-time European Champion. AlphaGo won this game with a score of 5–0. People accused Fan Hui of losing due to lack of practice, which he denied. Then, in March 2016, AlphaGo went up against legendary Go player Lee Sedol—winner of 18 world titles, and widely considered the greatest player of the decade. AlphaGo won four out of the five games, with Lee Sedol winning just once. Despite AlphaGo’s significant win, people celebrated that Lee Sedol defeated AlphaGo with what many considered to be a “divine move,” an achievement that was a decade ahead of its time. Lee Sedol countered AlphaGo by playing an unexpected move that the neural networks didn’t predict, causing the AI to go haywire. Nonetheless, the round earned AlphaGo a 9 dan professional ranking—the first time an AI achieved the highest skill rank possible in Go.

The rounds between AlphaGo and Lee Sedol showed how AI could think creatively and not through predictable patterns. In game two, AlphaGo played an unexpected move that many live viewers thought was an error, as it did not fit into the traditional Go strategies of humans. People who watched it widely agreed that it was a very bad move that not a single person would have played. AlphaGo itself said that Move 37 had a one-in-10,000 probability of being played. Despite this, AlphaGo won the round, showing how AI could transcend the standard techniques of humans in order to create an artistic level of creativity.

The shock of an AI beating the world’s greatest Go player once, much less 5 times, was felt all across the world. This stark contrast from previous AI Go models highlighted the progression of AI as a whole, showing how human determination and modern technology work hand in hand to achieve previously thought unfathomable feats.

This parallels the now ancient defeat of chess grandmaster Garry Kasparov by chess AI Deep Blue. At the time, it was a revolutionary moment, demonstrating AI’s capability to surpass even a reigning world-class chess champion given proper training. Looking at the impact of Deep Blue, we can see what might happen with the future of Go. People in the chess world began using computer engines (AI systems that analyze and show the best move in the game) in order to train and get better at chess, leading to a shift in strategies and preparation. We have already seen a similar impact in the Go world because of AlphaGo. After the Go match with Lee Sedol, a worldwide shortage of Go boards was reported. Fan Hui, who had previously played AlphaGo, said, “With [AlphaGo], I will change something in my game. Maybe he just can show humans something we never discover [sic]. Maybe it’s beautiful.” After training with AlphaGo, he went on to win the 2016 European Professional Go Championship. Unlike chess, AlphaGo’s impact was felt largely in the highest levels.

Just like with Deep Blue, people thought that AlphaGo’s victory over Lee Sedol rendered human play obsolete, which would potentially diminish interest in professional Go. However, AlphaGo inspired a Go renaissance, with high-level players studying AI games for their unconventional yet revolutionary tactics. Shi Yue, a 9 dan professional world champion, said, “I believe players more or less have all been affected by Professor Alpha. AlphaGo’s play makes us feel more free and no move is impossible to play anymore. Now everyone is trying to play in a style that hasn’t been tried before.”

We are witnessing the rise of AI all across the world, a shock that seems similar to AlphaGo’s defeat of Lee Sedol. As news of AI taking over jobs comes on the headlines, it’s important to remember that the rise of AI is not a moment that will spell doom for the human race. Despite some legitimate concerns, AI has the capability to push humans to even greater heights, showing how we as humans can use outside resources in order to benefit ourselves. The rise of AI demonstrates that, even in the face of challenges, humanity can adapt and use new resources to foster personal and collective growth in an ever-changing world.