IEEE Robotics & Automation Magazine - June 2014 - 86

ON THE SHELF

Cybernetics

Cybernetics
F.H. George, M.A., Ph.D.,
Hodder and Stoughton,
London, England,
1971, 194 pages.
odern robots require a level
of intelligence to carry out
the operations required of
them. We now have the
capacity to build robots that are capable
of learning and adapting. Many of these
modern triumphs can be attributed to
cybernetics. By combining many disciplines together to understand communication and control, cybernetics has
made it possible for robots to perform
humanlike cognition and problem solving. For these reasons, cybernetics is an
important field
for any roboticist
For the experienced
to understand
and be able to
engineer, this
apply in robot
book provides a
construction.
F.H. George's
historical overview
book, Cybernetby showing where
ics, introduces
readers to the
we came from and
field of cybernetprevisioning
ics in cle ar,
straightforward
the future that
language, so that
has unfolded.
the newcomer
will gain a firm
foundation in this field, both its theory
and applications. For the experienced
engineer, this book provides a historical
overview by showing where we came

Digital Object Identifier 10.1109/MRA.2014.2314012
Date of publication: 10 June 2014

IEEE ROBOTICS & AUTOMATION MAGAZINE

june 2014

from and previsioning the future that
has unfolded.
George uses block diagrams and
other drawings to illustrate the theory
clearly and make the ideas easy to grasp.
He uses everyday experiences and
events to exemplify what he is expressing to the reader without complex
mathematical algorithms, so that communication of the learning can take
place. The principles he explains show
how intimately interwoven cybernetics
is with our everyday lives. For example,
the feedback and feedforward principle
is deeply part of what stabilizes progress
in our routines. So when cybernetics is
applied to robotic devices, the principles
that he taught become useful, and it is
clear how to apply them.
This small, well-written volume is
divided into 12 chapters for a quick and
enjoyable read. The first chapter
explains the basic ideas of cybernetics,
which make it possible to create artificial intelligence (AI). A key quality of
intelligence is the ability to be selfadapting and self-sufficient, and Chapter 2 points out the different ways this
problem is addressed. Some cybernetic
systems attempt to simulate human
intelligence by building models based
on human functioning. However,
another approach is to synthesize intelligence based on the materials and
models used to form it, which may
result in cognition that is different from
the human mind. Chapter 3 clarifies
historic cybernetic hardware models,
including Ashby's Homeostat, Grey
Walter's tortoises, Uttley's classification
machine and conditional probability

machine, and Chapman's classification
machine. The author also describes
some chemical hardware models and
maze-running machines along with
partial, partitioned, and adaptive
growth types of classification models.
If you have ever been confused
about logic, Chapter 4 gives a concise
review to sharpen your foundation in
this area, including the famous Turing
machines. George shows how Boolean
algebra, the calculus of classes, manipulates logic in a symbolic form, to produce machine thinking algorithms,
known as automata. Neural nets are a
particular kind of automata, based on
the human nervous system. Chapter 5
summarizes the simplest forms of neural nets and explains how the network
approach can minimize errors.
Computers were in their infancy at
the time of the writing of this book, and
yet, from the early perspective described
in Chapter 6, we get clarity about programming and computer language as
they relate to cybernetics and AI.
Whereas logic and neural nets give a
means for constructing simulations of
human processing, information theory,
described in Chapter 7, offers mathematical theories of communication, such as
stochastic processes and Markov chains.
Models of intelligence can draw
upon psychological theories of learning
and perception. Chapter 8 explains how
the two main forms of conditioning
work, classical based on punishment
and instrumental, based on reward, and
then shows how these methods are
applied to artificial systems, such as neural nets. In addition, learning involves

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2014