How AlphaGo Conquered the Game of Go with Artificial Intelligence
Go is an ancient board game that originated in China more than 2,500 years ago. It is a game of strategy, creativity, and intuition, where two players take turns placing black and white stones on a 19×19 grid. The goal is to surround more territory than the opponent, while also capturing or isolating the opponent’s stones. Go is considered one of the most complex and challenging games in the world, with more possible board configurations than the number of atoms in the universe.
For decades, computer scientists and artificial intelligence (AI) researchers have tried to create a program that can play Go at a high level, but with little success. The best computer Go programs could only achieve the level of human amateurs, and were easily defeated by professional players. The main difficulty was that Go requires not only logical reasoning and calculation, but also creativity and intuition, which are hard to replicate in machines.
However, in 2016, a breakthrough was achieved by a team of researchers from Google DeepMind, a London-based AI company. They developed a system called AlphaGo, which combined deep neural networks with advanced search algorithms, and trained it on millions of human and computer games. AlphaGo was able to learn from its own mistakes and improve its skills over time, using a method called reinforcement learning.
AlphaGo made history by becoming the first computer program to beat a human professional Go player without handicap on a full-sized board. In October 2015, AlphaGo defeated Fan Hui, the three-time European champion, by 5-0. Then, in March 2016, AlphaGo challenged Lee Sedol, the 18-time world champion and widely considered the best player of the decade, to a five-game match in Seoul, South Korea. AlphaGo won the match by 4-1, stunning the Go community and the world.
AlphaGo demonstrated not only superior strength and speed, but also creativity and innovation. It played several moves that were considered unconventional, surprising, and even beautiful by human experts. One of the most famous moves was Move 37 in the second game, which had a one in 10,000 chance of being played by a human. This move was later described as “a masterstroke of the highest order” and “a move that changed the history of Go”.
AlphaGo’s victory was a milestone for AI, and a testament to the power and potential of machine learning. It also inspired many Go players and fans to explore new possibilities and strategies in the game. AlphaGo was awarded an honorary 9-dan rank by the Korea Baduk Association, the highest certification for a Go player.
After the match with Lee Sedol, AlphaGo continued to improve and evolve, and played under the name Master in online games against top human players. AlphaGo won 60 games in a row, and then faced Ke Jie, the world number one at the time, in a three-game match in May 2017. AlphaGo won the match by 3-0, and was awarded a professional 9-dan rank by the Chinese Weiqi Association.
After the match with Ke Jie, Google DeepMind announced the retirement of AlphaGo, and the release of its game records and research papers to the public. The team also developed a new version of AlphaGo, called AlphaGo Zero, which learned to play Go from scratch, without any human data, and surpassed the original AlphaGo in just 40 days. AlphaGo Zero was then generalized into a system called AlphaZero, which could also play chess and shogi, and defeated the best computer programs in those games.
AlphaGo is a remarkable achievement that showcases the power and beauty of AI and Go. It is a system that can learn, grow, and create, and a partner that can challenge, inspire, and enlighten humans.