I very much agree with this sentence: You may still not be able to accept AlphaGo like this. I think this is because when people play Go, they must first understand what "Go" is before they can operate it. But AlphaGo did not know (or was not provided with data) that "Go is a game played by two people face to face, confrontational, zero-sum, and the board is 19*19. The board is square, and the top is When playing a grid, the grid is also square. If there are two pieces, black and white, the black piece will be played first. The two pieces should be played in turn. They should be placed at the grid points instead of in the middle of the grid. There are limited time requirements and the number of chess pieces is sufficient. One piece does not need to be played. It will occupy more than one grid. The chess pieces are round and convex on both sides. It is a game invented in ancient China. It can defeat humans at any point in the game.
First of all, I think AlphaGo can understand Go.
Computers have such regulations for Go (defined by Tromp-Taylor rules). Go is played on a 19*19 grid. Two players, "Xiao Hei" and "Xiao Bai", take turns to play the game respectively. A game in which a certain grid point on the chessboard is colored black or white. Coupled with the rules of lifting and forbidden loops, as well as the final judgment, it is the complete Tromp-Taylor rule. Is there any essential difference between this kind of Go and the Go in our eyes? If you don't take the Go culture into account, then I can definitely say that there is no difference.
So why does AlphaGo attract so much attention? Some people say that the last frontier of mankind has been lost, and some even say that artificial intelligence will replace humans. Answer: Go is one of the most complex games in the world. As the saying goes, "One careless move will lead to the loss of the game." Every move may affect the overall outcome. Generally speaking, the decision-making of a chess move is divided into two steps. The first step, "point selection": give several candidate points based on experience or feeling; the second step, "judgment": make formal judgments on these points and compare them. These two steps are easy to say, but to reach the level of a top expert, the requirements for talent and diligence are no less than those required by an excellent mathematician.
AlphaGo is a Go artificial intelligence program developed by Demis Hassabis, David Silva, Huang Shijie and their team from DeepMind, a subsidiary of Google. . Its main working principle is "deep learning".
So what is "deep learning"? The concept of deep learning originates from the research of artificial neural networks. A multi-layer perceptron with multiple hidden layers is a deep learning structure. Deep learning discovers distributed feature representations of data by combining low-level features to form more abstract high-level representation attribute categories or features. Deep learning is a method in machine learning based on representation learning of data. An observation (such as an image) can be represented in a variety of ways, as a vector of intensity values ??for each pixel, or more abstractly as a sequence of edges, a region of a specific shape, etc. Tasks (e.g., face recognition or facial expression recognition) are easier to learn from examples using certain representations. The benefit of deep learning is to replace manual feature acquisition with efficient algorithms for unsupervised or semi-supervised feature learning and hierarchical feature extraction. Deep learning is a new field in machine learning research. Its motivation is to build and simulate neural networks for analysis and learning of the human brain. It imitates the mechanism of the human brain to interpret data, such as images, sounds and text. Like machine learning methods, deep machine learning methods are also divided into supervised learning and unsupervised learning. The learning models established under different learning frameworks are very different. For example, Convolutional Neural Networks (CNNs for short) is a machine learning model under deep supervised learning, while Deep Belief Nets (DBNs for short) is a machine learning model under unsupervised learning.
It is precisely because of this "deep learning" that AlphaGo has learned the profound and profound game of Go. It is precisely because of "deep learning" that AlphaGo has gone from being ignorant of Go to challenging the world's top players.
So, why do people play Go?
When people play chess, each step of the chess decision is made through similar signal additions and subtractions. We have a value judgment on the current chessboard state and a judgment on the probability of victory or defeat. Playing chess on different grids, this choice is also based on experience; for different grids, there are different estimates of the rise and fall of the probability of winning or losing. At the same time, we will also use experience to predict the opponent's moves in the next few rounds. The more experience experts have, the clearer their ability to judge value in these three aspects.
When I first came into contact with Go, I had little ability to predict these things, or was very inaccurate, because there was no chance to experience similar situations; everything was in a new state. As you become familiar with it step by step, you will retain more state memories. It will remind you of what you have experienced before. Only then can we further accurately judge the value of different options. No matter how professional people are, they don’t need to think about it deliberately, it has become subconscious. My thoughts have gone to the superstructure and the upper levels. Whenever you make a mistake or don’t know how to go, it’s because you still lack concepts at a certain level, or you haven’t experienced the situation and cannot recognize the pattern. AlphaGo also builds its judgment accuracy step by step through similar learning methods. The weights in the neural network correspond to the "rules" learned by humans. There are reasons for his level of rules and final choices. Those with experience also have "understanding". Not much different from human understanding.