|
Book Reviews Archive: July 2000 to October 2002 |
This book proposes various explanations of human and animal behavior
based on "situated activity," interactive emergence, and history of use.
Topics covered: natural selection, artificial intelligence (AI), and
learning. In the final chapters, the author argues that human
behavior and thought can be explained using these terms, and he cites
recent studies of the interactive behavior of new-born infants and the
role such behavior plays in concept formation and language development. Over the past few decades, it has become evident that traditional AI
is limited as an engineering tool for building systems that can respond
in real time to open-ended, ever-changing environments of the kind in
which intelligence is really needed. Some suggest that the various
disciplines under the heading of Artificial Life may be of great value
in creating robots that can respond in "creative" ways. For example,
recent work in situated robotics has revealed that meaningful (although
primitive) behavior can emerge without the need for internal
representations. In other words, it is possible to build a robot that
follows walls without having to put any representation of a wall inside
its controlling mechanism. There are no explicit programs inside the
robot that instruct it to follow walls. Walls need not be formally
defined in order to produce this behavior. Even though the robot won't
follow the same wall in exactly the same way each time it nears the
wall, it can be respond to the wall with behavior that singles out walls
as a meaningful feature of its environment. This interactively emergent
behavior of wall-following may lead to more complex behavior.
Catching Ourselves in the Act suggests scientific explanations of
development and learning that link various disciplines such as
sociology, anthropology, and situated robotics. It provides an overview
of autonomous agent research and artificial life, and explores the
impact of situated robotics in understanding human psychology. The book is not light reading but should be of great interest
to those interested in computational models of the mind.
|
copyright © 2004 ISAST