|
Building a Machine with Visual Thinking
Capabilities
ABSTRACT:
This talk is based on the book by Z
& M Les issued by Springer-Verlag. In this
talk the new results of the research on understanding and visual
understanding will be presented.
Understanding requires knowledge
that is acquired during learning process. In contrary to the existing
learning methods, learning in the Shape Understanding
System (SUS) is based on the new paradigm of the constant learning
that means learning that will “never end”. During learning process new
knowledge is added and ‘connected’ with previously learned knowledge.
Learning refers to the system of categories where the central categorical
structure is based on the category of the visual objects. In this talk a
short description of the shape understanding system will be presented and
the basic terms of the proposed method will be briefly described. The shape
categories, categories of the visual objects and categories of the
non-visual objects will be briefly described in the context of the visual
thinking and visual understanding processes. Thinking and visual thinking
will be described in the context of the philosophical and psychological
investigations and the existing systems in the area of AI and robotics that
could be called "thinking machines" will be briefly described. The
results of learning knowledge to solve the crossword puzzle as well as
learning specific knowledge to solve tasks from the different domains of
the school curriculum will be presented. Application of the new method in
solving the visual IQ test and other selected problems will be also
presented.
|
