Topic: Information Artificial Intelligence Theory
Lecturer: Li Angsheng, professor and doctoral supervisor of Beihang University
Host: Professor and Doctoral Supervisor Li Zhetao
Time: 9:00 – 11:00, November 13 (Friday), 2020
Venue: Meeting Room 307, North Wing of Engineering Building
The axiomatic mathematics research in the 20thcentury established the mathematical theory of mathematical logic (logical reasoning); the Turing machine model, a mathematical model of the electronic computer; and, the classical information theory, the mathematical theory of communication. This achieved the glory of science in the 20thcentury and guided the information age of the 21stcentury. “Can machines think?” Turing raised the question in 1950, scientifically studied the intelligence of machines for the first time, and began the study of artificial intelligence. The research of reasoning, calculation, and information really helps the research of artificial intelligence. However, independent research on information and computing has not been able to lay a theoretical foundation for machine learning and artificial intelligence for a long time.The mission of computers in the 20thcentury is to prove that they are useful, while their mission in the 21stcentury has become ubiquitous. The corresponding mathematics should be a mathematical theory of information processing.The classical information theory solved the problem of information transmission, but it failed to solve the problem of information processing. In addition, information and computing are obviously key elements of learning and intelligence. Due to their separation, we have not been able to capture the essence of learning and intelligence.In fact, the current artificial intelligence is actually a computing-centric information processing technology. Computing is a partial concept. It cannot realize the intuitive reasoning of human learning. Computing first needs to answer the question of what counts and requires people to answer it. The machine does not know what it is.It has no abstraction, especially the mechanism of layer spectrum abstraction unique to human beings. More importantly, computing has no creative mechanism of human intelligence.The goal of human learning is to acquire knowledge, reveal laws, and create based on knowledge and laws, which are naturally the goals of intelligent machines.However, these problems are natural obstacles to any computing-centric machine.
Structural information theory is the first mathematical theory of information processing, which laid the foundation for the establishment of information-centric artificial intelligence theory.We have established a learning model based on observation, association, imagination, hierarchical abstraction, coding, decoding, and knowledge tree. We have realized the combination of logical reasoning and intuitive reasoning; created the interpretation criteria of syntactic and semantic consistency; established information measurements, models, principles, theories, and methods of learning and machine intelligence; solved the problem of interpretability of artificial intelligence; and, broke the obstacles of computing-centric artificial intelligence technology. The information artificial intelligence theory has been successfully applied in a wide range of information processing with remarkable effects. The report will introduce this new theory.
Li Angsheng, male, born in March 1964, is a professor and doctoral supervisor.
1984, graduated from Yunnan Normal University;
1988, received master’s degree at Institute of Software, Chinese Academy of Sciences (ISCAS);
1993, got PhD at ISCAS;
1998.1-1999.1, visiting scholar at University of Leeds, England;
2000.3-2002.2, researcher at University of Leeds, England;
2008.9-2009.3, visited Professor Juris Hartmanis of Cornell University, the US.
1993.7-2018.7, he worked at the Institute of Software, Chinese Academy of Sciences.
1993-1995, research assistant;
1995-1999, associate researcher;
2018.8-present, professor of Beihang University.
2003, awardedthe National Science Fund for Distinguished Young Scholars;
2008, selected into the One Hundred Talents Program of the Chinese Academy of Sciences;
2012, invited to visit the Isaac Newton Institute for Mathematical Sciences of the University of Cambridge, England as a visiting researcher.
Professor Li’s main research direction is information and computational theory in cyberspace. The structural entropy measurement he proposed in 2016 solved the information measurement problems proposed by the founder of information theory, Shannon in 1953, and by the winner of Turing Award Brooks in 2003, respectively.