Artificial intelligence - Approach and Method
- 2. OVERVIEW OF CHAPTERS Content Synopsis Definition of Artificial Intelligence The birth and development of Artificial Intelligence Method of Artificial Intelligence research Descriptive content of artificial intelligence research
- 3. Introduction To AI Human natural intelligence is ubiquitous with human activities, such as solving problems, playing chess, guessing puzzles, discussing problems, preparing plans, and programs, and driving cars. Therefore, intelligence is closely linked with human activities. Most of the problems are complex and ill defined. Human mind have the ability to solve such complex problems. The challenge arise when we talk about computers solving such complex problems. Is it possible for computers to solve such problems?
- 4. “Our intelligence is what makes us human, and AI is an extension of that quality.” – GREAT SAYING - Yann LeCun Professor, New York University
- 5. AI Approach Humanoid Approach Kurzwell proposed artificial intelligence that artificial intelligence is a technology that creates machines that can complete certain tasks, and when we human process these tasks, we need certain intelligence. Method: For better intelligent tasks done by humans, let computers complete them The most famous is the Turing test Human Like Thinking Bellman proposed that artificial intelligence is the automation of activities related to human thinking, decision-making, problem solving, and learning. The main method used is the cognitive model---a testable theory about the working principle of human thinking. If a program can think like a person, then it must be determined in some way how a person thinks. In order to determine how the inner workings of the human mind works, there are two methods: through introspection-mastering one's own ideas in the process of thinking; or through psychological experiments Rational thinking In 1985, Charniak and McDermott proposed that artificial intelligence uses computational models to study intellectual ability. This is a rational way of thinking. A system is rational if it can make the best decision based on the information it knows (knowledge, time, resources, etc.) When knowledge is complete and resources are unlimited, it is called logical reasoning. When knowledge is incomplete or resources are limited, it is rational behavior.
- 6. Turing Test in AI Tester A, subjects B and C. A is a person, B and C are a person, and the other is a computer. A asks questions, B and C answer separately. If the answers of B and C make it impossible for A to distinguish between a human's answer and a computer's answer, then the computer has intelligence. The Turing test gave the first philosophical statement to test whether a computer is intelligent.
- 7. 1940s-foundation laying - In 1943, American neuroscientist Warren McCulloch and logician Water Pitts proposed a mathematical model of neurons. 1950s-the beginning - People call 1956 the first year of AI. As a discipline, artificial intelligence started during this period and achieved early success. 1960-ElizaIn 1961, - Unimate, the world's first industrial robot, was put on trial at the General Electric factory in New Jersey, USA. In 1966, the first mobile robot Shakey came out, that is, the robot that can smoke. Eliza was born the same year as Shakey. 1970s-Robotic consultation- In 1970, the world's first humanoid robot WABOT-1 was born at Waseda University in Japan. 1980s-"Terminator"In 1984, Princeton University professor, physicist, molecular biologist and neuroscientist Hopfield used analog integrated circuits to realize the neural network model he proposed two years ago. Deep learning is hot and has made breakthroughs. 1990s-chatbot - In the late 1990s, the combination of artificial intelligence and robots and human-machine interfaces produced intelligent agents with emotions and emotions. Emotion/emotional computing (that is, the evaluation of changes in emotions and then reappearing on the machine) developed rapidly, especially dialogue agents (chat robot). 21st century-deep learning - Robots have the capability now to recognize and imitate human emotions. They can behave as a human friend with all human emotions. As a result, the application of AI has blossomed everywhere and soon entered all areas of human life. The birth and development of artificial intelligence
- 8. Symbolism Symbolic processing as the core method Also known as top-down and symbolism, originated from GPS, used to simulate the psychological process of human problem solving, and gradually formed into a physical symbol system The goal of AI is to achieve machine intelligence, and the computer itself has symbolic processing functions, and it contains reasoning capabilities, so it can easily simulate the logical thinking process Symbolism believes that the basic unit of human intelligence is symbols, the cognitive process is the process of symbolic manipulation, and thinking is symbolic calculation. Connectionism Connection mechanism method based on network connection Also known as bottom-up and connectionism, it belongs to the category of non-symbolic processing. In reality, people do not just rely on logical reasoning to solve problems, sometimes non-logical reasoning also plays a very important role Connectionism : Artificial intelligence can be realized by the structure of the bionic human brain, and the content of its research is neural networks. Behaviorism Behaviorism, also known as evolutionism or cybernetics school, is a school of artificial intelligence based on cybernetics and "action-perception" control systems. It is a non-symbolic processing method The basic view of behavior can be summarized as: 1. Knowledge, formal expression and modeling methods are one of the important obstacles to artificial intelligence; 2. Intelligence depends on perception and action. After the machine acts on the environment, the response of the environment to the action is the original shape. Core Methods
- 9. Artificial intelligence is a science of knowledge. Take knowledge as the object to study the acquisition, expression and use of knowledge Data processing -> knowledge processing, data -> symbols. Symbols represent knowledge rather than numerical values or data. There is inspiration, there is derivation. Characteristics of artificial intelligence research Artificial intelligence is one of the most controversial sciences Focus: Should the current research on artificial intelligence focus on the universal laws of human thinking, or the processing and application of specific knowledge? What is the nature of intelligence? Can machines reach human level? Conclusion: Artificial intelligence research is very difficult