The Father Of Artificial Intelligence Reveals The Secrets Of Thinking Level

Zheng Lei / Wen

As a behavioral economics researcher, Herbert Simon has always been a great scholar who can not be ignored. He is praised as a rare "Renaissance scientist" in the twentieth Century. This can be verified by his award of "Turing Award", which is awarded the highest prize in artificial intelligence and computer science. He made outstanding contributions in psychology, organization theory, management operations research, politics, artificial intelligence and so on. The "bounded rationality" put forward as a pioneer of behavioral economics thought and won the Nobel prize in economics. I have seen his autobiography, the urchin and the master in the science labyrinth. Recently, because of the secret of brain cognition, I read his new book, "cognition: thinking and intelligence behind human behavior", based on his own lectures on cognitive psychology at Peking University. Simon is known as the father of artificial intelligence because of his outstanding contribution in the field of expert intelligence system. This book can be regarded as a basic research in this field. It introduces the steps of learning, remembering and answering problems in the brain, and how to simulate the problem solving process by computer. It is valuable for us to improve our learning methods and improve our ability to solve problems.

Cognition: thinking and intelligence behind human behavior. Data map

Simon introduced the school of information processing in cognitive psychology. He believed that when people were stimulated by the outside world, they would rely on their experience in the brain to decide what to do. Experience includes the state of the organism and the contents of its storage. We regard human beings as an information processing system, also known as the "physical symbol system". The symbol here is mode. A perfect symbol system includes six functions, which can show the intelligence of human beings, and the six functions can be realized by computer programs. Therefore, it is theoretically possible to use computer to simulate human activities, which is the theoretical basis of artificial intelligence.

In daily life, a large number of symbols enter sensory organs at all times. After processing, short-term memory and long-term memory are formed. These long-term memory will become the main toolbox for our response to external stimuli. Human memory is not like computers in the smallest bit units, but by coding the fragmented information into "chunks", in order to improve memory efficiency. Short time memory takes a short time, but only 4 blocks can be remembered. Only by repeating these blocks into long-term memory can they be permanently preserved, while subsequent extraction and use follow the opposite step. Cognitive psychology finds that the process of solving problems is the process of "re identification". When receiving the external information (see, hear or feel), the brain quickly searches for similar memory fragments in the long-term memory network, and carries out the information comparison test. This process is carried out under the unconscious speed very quickly. If you do not find the relevant information, start the logical thinking system and start solving this problem like a beginner. Therefore, people's learning process is very important. After learning and memorizing, once they are familiar with it, they can automatically and unconsciously distinguish it.

People's learning process needs to be repeated many times, and how long it takes to learn depends on the difficulty and learning methods of learning objects, such as memory and cognitive characteristics, and some ingenious ways to improve learning efficiency. Everyone is different in information encoding and learning strategies, so intelligence is different. So there are differences in AI's imitation of human thinking, but computer storage and computing power can eliminate some of the differences in human intelligence. We all admire the ability of experts to solve problems. In fact, experts have accumulated knowledge and clues in related fields, which help them to extract different knowledge. The amount of knowledge of an expert is 5-20 blocks. It takes at least ten years to get these, and about 1000 hours are spent on learning and memory. This is why we can concentrate on one thing and become an expert in ten years. At this time, a network structure and connection of neurons corresponding to corresponding knowledge are formed in the brain of experts. When problems are encountered, they can be analyzed and responded quickly. This expert instinct comes from very familiar fields. The author also pointed out that a concise picture is helpful in solving problems when solving problems, which is consistent with my experience in Reading Mathematics Department. Making visual representation is a central part of problem solving and half of success. When experts see problems in related fields, they can instantly extract abundant information, understand the meaning of this problem and make immediate reasoning; if temporarily unable to solve, they can continue to expand information and store it until they can solve this problem. This is a continuous process of learning (collecting and remembering information) and reasoning.

Simon has an argument that makes people think deeply. He points out that mathematics is not suitable for studying human behavior, while computer programming language can describe human psychological phenomena better than mathematics. In social science research, especially in modern economics research, overmath has not improved our understanding and explanation of human social behavior. Perhaps we can find a better solution by using computers as tools and using big data, artificial intelligence, computer simulation and so on.

?