Essay: AlphaGo and Alan Turing

Written on 2016 3.8 nights

AlphaGo and Alan Turing

If we can be called the biological version of the robot, carrying the artificial intelligence that has been perfected for thousands of years or eons of time, and as the behavioral doctrine points out, the corresponding behavior of different stimuli, then we are how to make robotic robots have the best reference of artificial intelligence.

The mechanical body is different from the human body. We have many physical limitations. Our organs will degenerate, even if they have been so sensitive. Our brains always need to rest and adjust ourselves with enough sleep. We experience some irreparable damage as we grow older. And these are mechanical bodies that don't have to worry about, and the speed with which it's calculated and simulated is hard for our brains to approach. However, robotic robots have always been considered as not thinking.

In the early chess AI, for example, Deep Blue defeated chess master Kasbarov, it uses the decision tree to find out where to go next. This is the first time I know about AI, although I still think the decision tree is very mysterious, but I think the decision tree should still use the idea like search, violent to find out where the best chance of winning. It takes full advantage of the speed of computer computing is very fast, but this brute force algorithm is not able to support the go AI, because the most places on the board of Weiqi 19*19 for the computer is feasible next, it is difficult to imagine its time complexity will be how spectacular.

Remember that year six years old, when my mother took me to the door of the Chess house, my first teacher handed me the back of the game to write the rules of the game. We, dark blue and alphago like the most simple and direct rules of chess movement to understand and study the strategy of Weiqi. I gradually understand that the process of a chess player to improve his chess is long, and he needs to face such two major problems. A mature chess player in the game, according to the opponent's Chifeng and the current situation on the board to design their own strategy, he needs to confirm that his strategy is in line with the overall situation of the best strategy, and his lazi of the local need is the best step under his strategy. In fact, the purpose of this strategy is to clearly defeat the opponent, that is, to surround the larger than the opponent. Therefore, the strategy is to take the current most promising step.

Whether we or the machine in the realization of the best step out of his strategy, is through a lot of chess to achieve. We have gained experience by competing with our opponents of equal strength. and the machine through learning is to train a large number of high-level players in the game. This can help the machine and we decide which step is the better chance. However, how should bigger picture be implemented?

I always thought that bigger picture is more intuitive, like thinking, imagination is the machine is difficult to achieve (imagine can already do Qaq). In Alphago, however, he used neural networks to solve bigger picture problems. My first contact with the neural network was in Coursera's ML class, where his assignment was to use neural networks to identify numbers (only the pro and back of the neural network were required). Whether it is the recognition of numbers, or human faces, whether or not convolution processing, neural networks are often used to process pictures, to achieve computer vision. Using a neural network to process the entire chessboard, it is possible to get a better chance of winning a bigger picture than a few steps (perhaps a range). Then continue to use the neural network more detailed calculation of their odds. Although I do not know how it is implemented, it breaks down some of my traditional perceptions. Perhaps behind our intuition is an analysis of the logic of the human brain. But in the thousands of-year evolution of the human brain, the process of logical analysis, now seemingly intuitive, is evolving, from the initial effort to the moment of inspiration, which we mistakenly believe to be intuition. This may also explain why we are struggling to count, because arithmetic is modern in comparison to human appearance, so the human brain has not evolved to be able to count fast.

It uses the Monte Carlo search tree at the same time and finally calculates an optimal step. To better get the parameters of each layer in the neural network, it plays chess with itself, gets a new chess game, and trains itself from there. Finally defeated Europe's go champion. I saw a question on the internet about playing chess with myself. Some scholars think that with their own chess in fact the amount of knowledge is still unchanged, and give examples of kindergarten knowledge of the amount of reserves can never reach the knowledge of the Nobel Prize.

I think this idea is very interesting. A good Weiqi is not like a Nobel Prize winner, he needs not a lot of professional knowledge, but experience. Experience can come from other people's chess board, the masters of the past, but also can be derived from their own board. The teacher had spent a long time teaching me how to play chess correctly with myself, to think for resist. When humans choose to go opponents, he usually chooses to match his own strength. Each other's knowledge reserves and their own is quite, under the same teacher guidance, even knowledge will have a great overlap with their own. Then, according to the above view, the human race in chess is unable to progress. In fact, the human people is not lack of master. Switch to the computer, and vice versa.

The advent of alphago symbolizes that machines or scientists can begin to learn to replicate our thinking. According to singularity theory, the level of human science and technology is growing at an exponential rate. Alphago such a clever weak artificial intelligence will be able to achieve all aspects of human intelligence beyond the strong. Maybe one day, when a robot can think about us, it will look like a gorilla in the zoo at the moment. The robot will serve, be loyal to us (it will only be loyal to the target we engrave on it), or it will destroy us (the target on the chip has not succeeded in binding it as smoothly as we envisioned it to be). This will be the moment we have a chance to witness.

If humans use mechanical bodies to replace their degenerated limbs, aging kidneys, and finally to their brains, in order to live forever. They will become the winners of evolution, perhaps derived from new species. When they choose to build a new country, their individual is highly unified, similar and better, or there are differences (such as character and philosophy).

In the first chapter of Alan Turing, "he learns to be gregarious," the Turing family believes in "the good luck of the brave." And this brave is manifested in the Turing family bravely to express themselves and the mainstream of the difference, and proud. Alan Turing in demanding public studies to show his differences with ordinary people, to try his favorite subjects. He did nothing wrong, even quite brave and persistent. His actions, however, could not give him recognition. The teacher liked him, but criticized his behavior. The students were isolated and bullied him. These embody the human nature's isolation and contempt for the heterogeneous.

I read an article before about whether genius should be a gregarious one. The article believes that the talent's non-gregarious will cause his lifelong loneliness and gradually distorted character, eventually leading to anti-social tendencies, and great harm to society. The conclusion is that geniuses should be gregarious. But if Alan Turing is in his class as a student, he gives in to the public school, learns rigid textbooks like everyone else, and he will become gregarious in textbooks that do not seem to have much meaning now. His genius will make him successful, and become the high-income social successful person that the teacher hopes. Now, then, no one will admire him so enthusiastically.

It was the Alan Turing of the spirit that made him suffer from all kinds of suffering and persecution, and kept him from the worldly habits. It was these sufferings that shaped the uniqueness of him, and made him into a shining star.

Now, we are not tolerant to the heterogeneous, if possible, in the future, in the artificial intelligence robot country, the heterogeneous will be respected and tolerant.

Essay: AlphaGo and Alan Turing