Humans have played games like chess and Go for hundreds of years, but only in the last few decades has artificial intelligence become sophisticated enough that computers can rival and even surpass the best human players. Go, in particular, has been seen as a challenge for AI because of the incredible complexity inherent in the game.
The game of Go is played by two players whose objective is to surround as much territory as possible on a 19x19 grid. Each player takes turns placing pieces, known as stones, on the intersections of the board. The player who surrounds more territory at the end of the game wins.
Invented over 2000 years ago, this deceptively simple game holds a depth and complexity unlike any other game. Even though the rules are simple, the number of potential moves for each turn is astronomical. This makes it an attractive challenge for artificial intelligence. In order to build an AI that can play Go, developers need to create algorithms capable of dealing with this immense complexity.
One such approach is Monte Carlo Tree Search (MCTS), which combines Monte Carlo simulations with a search tree data structure. MCTS works by traversing the search tree and expanding nodes randomly. It then assigns an upper confidence bound from rewards estimate to each node, allowing the algorithm to explore the tree in a semi-intelligent fashion. By repeating this process, the algorithm gradually converges on an optimal policy for the entire game.
A major breakthrough occurred in 2016 when Google's AlphaGo AI beat the world's best Go player in a 5-game match. AlphaGo used deep learning in combination with MCTS to achieve superhuman performance.
Since then, AI has gone on to challenge the world’s best players in other board and even video games. Machines are becoming ever more nimble and capable at solving complex problems, which promises to revolutionize how we interact with technology.
Artificial intelligence is an incredibly versatile field of study that has applications in nearly every field of study. As AI progresses and grows more powerful, it will be interesting to see what all it can accomplish in the future.