The Chinese Room thought experiment was first proposed by the American philosopher John Searle in the early 1980s. This experiment requires you to imagine a person who speaks only English in a room where he carries a book with a Chinese translation program. There is enough papers, pencils and cabinets in the room. Papers with Chinese questions are sent into the room through a small window. People in the room can use his book to translate the text and reply in Chinese. Although he does not speak Chinese at all, through this process, people in the room can make anyone outside the room think he speaks fluent Chinese.

Can machines have ideas? This is an old problem. Turing has designed a Turing test. If the machine passes this test, we should admit it is thinking. The Turing test is like this: Put a computer and a normal person in two rooms, then ask questions. You analyze the answers to find out which one is the machine. If you can't distinguish, then this machine pass the test to be regarded having the same thought as people. Searle created the “Chinese Room” thought experiment to refute the idea that computers can really think. People in the room can't speak Chinese, but he can make people think he can speak Chinese fluently. Computers cannot really understand the information they receive, but they can run a program and give you a smart impression.

What is thinking? What does understand mean? Does the computer understand knowledge? To answer these questions, we need to understand what is the essence of knowledge? What is the ultimate source of human knowledge? What are the scientific methods of understanding? These questions are fundamental questions of philosophical epistemology.

In fact, it is not difficult to answer these questions. To understand where human knowledge comes from, one should first think about where people come from. People evolved from animals, so human understanding must have evolved from animal recognition. Then how do animals recognize the world? Animal recognition comes from conditioned reflex, so almost all of human knowledge comes from conditioned reflex. There is no essential difference in the process of conditioned reflex between human and animal.

When Pavlov trains a dog with a bell, the simultaneous appearance of the bell and food is a new environment. This new environment makes the dog establish a new neural connection. This new neural connection makes the dog establish a conditioned reflex. The dog salivates when it hear the bell, even if it don't see food. When a newly established conditioned reflex can benefit a dog, it becomes a dog's knowledge. Does the dog really understand this knowledge? The dog doesn't know the story behind the bell at all. It doesn't know what Pavlov wanted to explain through this experiment, but the association caused by the bell has become a dog's knowledge. For human, establishing a useful conditioned reflex is to add a kind of knowledge. To dogs, adding a kind of knowledge is to add a biological character. The connection between bell and food is dog's interpretation of the world.

Lower forms of animals can only establish simple conditioned reflex, advanced animals can establish complex conditioned reflex, that is, the connection between multiple phenomena and changes in interest. If the dog observes that changes in self-interest do not occur at the same time as the phenomenon A, but at the same time as the phenomenon A+B, then in the dog's nervous system, A+B will be linked to the change in the interest. This is the earliest analysis.

For example, A is a ringtone, B is a breeder, and food is only available when the ringtone and the breeder appear at the same time; if there is only a ringtone, no breeder (the ringtone may come from a visitor), there is no food, then in the nervous system of the dog, A+B will be contacted with food. If the dog continues to observe that the food does not always appear at the same time as A+B, but always at the same time as the phenomenon A+B+C, then A+B+C will be contacted with food. This is called further analysis. Induction is the process of establishing conditioned reflexes, analysis is the process of establishing complex conditioned reflexes.

A newborn child does not have any concepts and knowledge, but he will slowly learn knowledge. This learning process is the process of gradually establishing conditional reflexes. For example, when learning to use chopsticks, he will be praised for doing it right and be punished if he do something wrong. There is no essential difference between this process and that of training dogs. The process to establish conditioned reflex is to strengthen associations of thinking. New conditioned reflex generate new associations that associate new stimuli (including images, words and sounds, etc.) with decisions, just as dogs associate ringtones with saliva. It produces science and technology to combine weeding and harvesting. Human complex decision-making, judgment, theory and social practice do not seem to have similarities to conditional reflex. In fact, all of these are complex conditional reflex.

The Materialism believes “Human cognition is a reflection of man's objective world.” This statement does not discover the nature of cognition. The reflection of the cell phone in the mirror is totally different from the reflection of the cell phone in the human eyes. After seeing the mobile phone, people can associate and act, and can pick up the mobile phone, and the reflection of the mirror will not cause any association and operation. Wittgenstein believes that the meaning is to use.

The so-called understanding mainly refers to the association of thinking. When you are too close to a painting, you often don't understand what it is drawn. When you are far away, it's easy to understand. Because what you see closely cannot make you association. Sometimes when you hear a word, you will immediately nod your head, “I have heard of it.” Actually, you do not understand its precise meaning. For example, even if you do not know the definition of life, you can understand what life is because you can associate this word to many things. The more association, the deeper the understanding is. Generally when people say, “I do not understand”, he just means that he cannot associate. Every time you read a book, you will get a new understanding. This mainly means that every time some new associations is created.

It will come to light sooner or later if the person in the Chinese room can only answer questions by looking at translation book. He cannot answer all questions correctly. Only after long-term practical use of Chinese can we establish correct complex associations and truly under- stand Chinese. So this thought experiment did not defeat the Turing test. True or false understanding can be measured by the Turing test. If someone in the Chinese room can really answer all the Chinese questions by looking at the translation book, we can think that he really understands Chinese.

There was a good comment on zhihu.com, “The Chinese House explains nothing but that there is no guarantee that anyone who looks smart and flexible (either a machine or human) is 100% conscious. I can use Chinese House logic to doubt that any people have ability to think, because there is no physiological evidence that his nerve activity is not mechanically retrieving a ready-made reaction, but these reactions just happen to correspond to reasonable answers. If anyone think the machine that pass the Turing test had no consciousness, he also had the reason of exactly the same intensity to question the unconsciousness of the human being in the test (control group). If he decides to ignore the possibility of the latter, this conclusion also supports him to ignore the former. In any case, the effectiveness of the Turing test is not reduced. The discussion of Daniel Dennett from another point of view achieves the same result.” “If the behavior cannot be used to test consciousness, how can you ensure that other people in reality are not naturally evolved zombies?” (Fu Tieqiang answered the Chinese room thought experiment is what? Is it successfully rejected the effectiveness of the Turing test? https://www.zhihu.com/question/39447528/ answer/237092278)
In short, understanding is association, association comes from conditional reflex, and the purpose of establi- shing conditional reflex is for practice. The learning process of artificial intelligence is based on the genetic algorithm of neural network programs, which is basically the same as the biological process of establishing conditional reflex. Association is the basic function of the computer. When machines can solve problems through conditional reflex and association, we can say that machines understand knowledge.

The essence of knowledge is the ability to change the world. When AlphaGo triumphs over man, no matter whether we admit it or not, it has a capacity to change the world beyond humans. It is a superfluous issue or scholastic issue whether the machine has knowledge and understand knowledge. It is people who are having trouble with themselves. No matter whether people recognize it, machines will surpass people and dominate the future world.