AlphaGo能否代表人工智能的未来(2)

不过在现实中，猴子打出一篇像样的文章的概率是零，因为科学家经过反复试验后发现，猴子在使用键盘时通常会连按某一个键或拍击键盘，2003年，某个科学家做了这个实验，结果打出了5张全是‘S’的纸。最终打出的文字不可能成为一个完整的句子。这是因为现实是非常大的有限，而不是严格意义上的无限。

AlphaGo解空间示意图

AlphaGo是什么，有何厉害之处？

通过阅读AlphaGo发表在《Nature 2016》上的文章Mastering the game of Go with deep neuralnetworks and tree search[3]可以了解到：

Abstract: The game of Go has long been viewed as the most challengingof classic games for artificial intelligence owing to its enormous search spaceand the difficulty of evaluating board positions and moves. Here we introduce a new approach tocomputer Go that uses ‘value networks’ to evaluate board positions and ‘policynetworks’ to select moves. These deep neural networks are trained by a novelcombination of supervised learning from human expert games, and reinforcementlearning from games of self-play. Without any lookahead search, the neuralnetworks play Go at the level of state-of-the-art Monte Carlo tree searchprograms that simulate thousands of random games of self-play. We alsointroduce a new search algorithm that combines Monte Carlo simulation with valueand policy networks. Using this search algorithm, our program AlphaGoachieved a 99.8% winning rate against other Go programs, and defeated the humanEuropean Go champion by 5 games to 0. This is the first time that a computerprogram has defeated a human professional player in the full-sized game of Go,a feat previously thought to be at least a decade away

AlphaGo的神经网络训练模型和架构

通俗地讲，AlphaGo就是一套针对围棋周密设计的深度学习引擎，采用多种机器学习技术进行整合：增强学习（reinforcement learning），深度神经网络（deep neural network），走棋网络（policy network）、快速走子（fast rollout）、估值网络（Value Network）和蒙特卡洛树搜索（Monte Carlo tree search, MCTS），加上Google强大的硬件支撑和云计算资源，结合CPU   GPU，通过增强学习和自我博弈学习不断提高自身水平[4-5]。值得一提的是，paper两位主要作者之一的Aja Huang（一名低调的台湾教授，另一位是David Silver）也是一名围棋爱好者，目前水平已经达到业余六级（普通人一般是业余二级），整个AlphaGo主要来自两位在博士阶段及毕业以后五年以上的积累。

在未来，AI科技能否超越人脑？

自1956年AI这个概念被提出并确立以来，一共经历了符号主义→ 专家系统→ …→ 统计学习→ 神经网络→ 深度学习等几大重要阶段，尤其是近十年来对深度学习的研究使得AI在历经神经网络发展低潮过后再一次空前崛起。(责任编辑：方向)