The ancient game of checkers (or draughts) has been pronounced dead. The game was killed by the publication of a mathematical proof showing that draughts always results in a draw when neither player makes a mistake. For computer-game aficionados, the game is now "solved". Draughts is merely the latest in a steady stream of games to have been solved using computers, following games such as Connect Four, which was solved more than 10 years ago. The computer proof took Jonathan Schaeffer, a computer-games expert at the University of Alberta in Canada, 18 years to complete and is one of the longest running computations in history.
Siby Abraham is a computer scientist specialising in artificial intelligence. He is an associate professor and head of the department of mathematics and statistics at Guru Nanak Khalsa College, Mumbai. How many years does it take for a child that does not know anything about English to master it at a Shakespearean level? Suppose also that there is no book, no help and no support (human or non-human) at all times. Recent research in artificial intelligence (AI) suggests that, if you replaced the child with a learning computer, the time taken could be in the order of a few days.
Recently I posted about the phenomenal performance of the AlphaZero algorithm in computer chess. For the first time in history, an algorithm displayed human-like understanding of chess. AlphaZero seemed to understand what moves were best and spent its time focusing only on them. It didn't mechanically crunch through millions of possible positions, run out of time, and then select the best move. The best moves emerged from its computer neural network, like a human grandmaster.
It's man vs machine this week as Google's artificial intelligence programme AlphaGo faces the world's top-ranked Go player in a contest expected to end in another victory for the machine. Google's artificial intelligence program AlphaGo took on the Chinese world number one of the ancient board game today in the first of three planned games, beating its human opponent by a narrow margin. It is the second time the AI has gone head-to-head with a master Go player in a public showdown, after stunning the world last year by trouncing South Korean grandmaster Lee Sedol four games to one. Google's artificial intelligence programme AlphaGo (right screen) will face the world's top-ranked Go player, China's 19-year-old Ke Jie (left), in a contest expected to end in another victory for rapid advances in AI. AlphaGo, part of Google's DeepMind project, competed against Ke Jie, currently ranked as the top player in the world, at an event held in the eastern Chinese water town of Wuzhen.
ResNet architectures are a popular way to train deep networks, often for image recognition. For AlphaZero's network, the input is the board state, and there are two outputs: The network is initialized randomly. After some self-play has occurred, training data consists of randomly chosen board positions from played games, in the form of (board state, game result, child visit counts from MCTS). This video does an excellent walk-through of standard MCTS. To choose a move given a board position, we perform 800 iterations of the following algorithm.