Introduction - Advanced Artificial Intelligence - page 11

a new means for information processing with computers. With more and more
applications and in depth research of artificial neural networks, the researchers
have found many problems of existing models and algorithms, and even met with
some difficulties of the nonlinear theories or approximation theory. Despite these
problems and difficulties, we believe that with the in-depth and extensive
applications, neural networks will continue to develop and promote current
techniques. The theory of neural field we proposed is such a new kind of attempt.
Currently, integration of symbol processing systems and neural network
models is an important research direction. Fuzzy neural networks integrate fuzzy
logic and neural networks, taking each other's advantages in theory,
methodology and application, to develop systems with certain learning and
dynamic knowledge acquisition capabilities.
Behaviorism, also known as behavior-based AI, in many respects reflects the
behavior physiological views in AI. Rodney Brooks brought forward theories of
intelligence without representation (Brooks, 1991a) and intelligence without
reasoning (Brooks, 1991b), and stated that intelligence is determined by the
dynamics of interaction with the world.
These three research genres investigate different aspects of human natural
intelligence corresponding to different layers in the model of human thought.
Roughly categorizing, it can be taken that Symbolicism focuses on abstract
thought, Connectionism focuses on imagery thought, while Behaviorism focuses
on perceptual thought. The comparisons of Symbolicism, Connectionism and
Behaviorism is shown in Table 1.1

Table 1.1 Comparisons of Symbolicism, Connectionism and Behaviorism.

Symbolicism Connectionism Behaviorism

Perceptual Level Discrete Continuous Continuous

Representation Level Symbolic Continuous Behavioral

Problem Solving Level Top-down Bottom-up Bottom-up

Processing Level Serial Parallel Parallel

Operational Level Reasoning Connection Interaction

System Level Local Distributed Distributed

Basic Level Logic Simulant Intuitional Judgment

Some researchers classify AI research into two categories: symbolic
intelligence and computational intelligence. Symbolic intelligence, also known as