So how did Google's AlphaGo beat one of the best Go players in the world?

To get an idea of why this was a common thought among researchers, it may be useful to understand how programs that play chess work (a game where machines have vastly surpassed humans), and then see why the same approach couldn't be used for the game of Go. In chess, a procedure known as minimax (along with several other clever tricks that help optimize it) is a common strategy to write programs that play the game (a.k.a. The most sophisticated of these programs use this approach at their core, including popular open source programs such as GNU Chess and Crafty. Minimax, which performs what is known in game theory as a "game tree search," can be explained in simple terms as a simulation of the game that takes into account all possible moves of one player and all counter moves of the opponent, until either the end of the game is reached or a certain prefixed number of moves has been simulated (more on this later). In essence, it's a way of simulating all possible futures of a game, and then figuring out, from the current position, which of the best futures can be forced by the player in turn to get the best possible outcome.