Artificial intelligence--Spiral 60 years _ Artificial Intelligence

This is October 29, 2016, the report of the academician at the "Taishan Academic Forum – Intelligent Mine Information Technology" conference. Have the honor to listen, benefit. The main content is now organized as follows:

Speaker Profile: Gawain, Professor of Peking University, doctoral tutor, academician of Chinese Academy of Engineering (elected at age 55). The national CPPCC member, the current Digital Media Research Institute director, the director of the System Chip Institute. March 2013 deputy director of the seventh session of the National Natural Science Foundation of China. 2013 Elected American Computer Society (ACM fellow). Director of information and Engineering Science, Peking University. 1 About Artificial Intelligence

The three main factions of artificial intelligence:
Logic (symbolism): Symbolic inference and machine inference
Founder: Simon (CMU)
The faction derives from the computer sector

Connectivity: Neural networks and deep learning
Founder: Minsky (MIT)
The faction derives from the electronic sector

Behaviorism: Control, adaptive and evolutionary computing
Founder: Wiener (MIT)
The faction derives from control

Intelligence/Knowledge:
Biological general spiritual ability. This ability includes the following points (understanding, planning, problem solving, abstract thinking, expression of ideas, and language and learning abilities)

Three factors of intelligence theory:
Compositional intelligence: Related to the level of education (corresponding to symbolism)
Experience Intelligence: relating to the Life experience (correspondence connection doctrine)
Situational Intelligence: can be simply understood as EQ (correspondent behaviorism)

The theory of human multiple intelligences:
The ability to do things logically
Competence in language and writing
Capacity for space reform
Musical ability
The ability of limb operation
The ability to Introspect
Interpersonal skills
The ability to explore nature
The ability of graphic image understanding

The father of artificial intelligence:
Alan Turing (1912-1954, University of Cambridge)
The test that we use to determine whether a machine is intelligent is what he proposes, called the Turing Test.

Turing Award (Nobel Prize in Computer science)
The Turing Award has now been awarded to more than 60 scientists (the only Chinese, Cc Yao), among which 8 scientists are doing artificial intelligence.

MIT Professor Minsky, the first link is his proposed, and then said that the link is not the same point of view is he proposed that he is a very great AI scientists.

McCarthy, Newell, Simon, and Feigenbaum, who were all very typical symbolism representatives, were pushed to include the most advanced machine proofs, artificial intelligence, general artificial intelligence machines, and knowledge engineering, and basically they made progress with several personal push.

Another Reddy is mainly to do speech recognition, he is also Kai-fu Lee, Shen Xiangyang teacher. He awards the field is large-scale artificial intelligence, in fact he is to do multimedia.

The other two valiant is the 2010 award (Machine learning theory), Pearl is the 2011 award (probability calculation and causal reasoning), their work is the future of artificial intelligence focus direction. 2 Artificial Intelligence 60

The origins of Artificial intelligence:
The Dartmouth Summer Meeting, 1956

The first wave of artificial intelligence: 1956-1976
Symbolism prevails (functional dominant)
Leader: CMU, MIT, IBM, Stanford, Harvard
Main characters: McCarthy, Newell, Simon, Feigenbaum, Reddy
Achievements: Inference, expert system, knowledge engineering, machine proof, artificial intelligence logic language

The second wave of AI: 1976-2006
The prevalence of connectivity (deep learning has not made a breakthrough)
Main characters: Minsky
BP algorithm, perceptual machine, SOM, error back-propagation network

Ai's Third wave: 2006-present
2006 Geoffrey Hinton joint Yann LeCun, Yoshua Bengio's paper is the symbol of the third wave beginning
The prevalence of connectivity (deep learning to achieve breakthroughs)
The reason for deep learning success: the progress of hardware/model and parameter adjustment skills Progress 3 The future trend of AI:

Future directions: Non-deterministic information processing and Bayesian networks (the combination of symbolism and connectivity)
Main characters: Valiant, Pearl

Real actions
In the field of machine learning, the basic theory of deep learning is still supported
Applications (computer vision, speech recognition, computational linguistics), the use of deep learning to solve practical problems is not innovative, such projects will be killed
Deep learning
Deep Yes
Learning No
Deep machine Learning
Deep machine. Yes
Learning No

60 years ' thinking of artificial intelligence

The following three pictures, a good understanding, very good ~

Whether the artificial intelligence era is coming
Good at logic, language and graphic images
Space, music, and physical functioning are so-so.
Introspection, interpersonal and natural exploration are not entirely

Artificial Intelligence 1.0 is a statistical and inferential part of the results achieved.
The artificial Intelligence 2.0 is on the basis of 1.0 to the non-statistical can not be inferred part of the region to advance.
Artificial Intelligence 3.0, probably up to 4.0, there will be a considerable part of it is difficult to do (epiphany), but this is the direction of future development. 4 Summary