Page 30 - Modern Robotics Building Versatile Macines

10 Modern Robotics

A group of people interested in communication engineering, the engi-
neering of computing machines, the engineering of control devices . . .
and the communication and control aspects of the nervous system has
come to a tentative conclusion that the relations between these fields
of research have developed to a degree of intimacy that makes a get-
together meeting between people interested in them highly desirable.

Neural Networks

While others worked on the organization of programs and data in
computers, Wiener remained focused on communications and con-
trol. For him this meant how these processes were carried out by liv-
ing things and the similarities between neurological and electronic
structures.
Since the 1930s, Wiener had closely followed Arturo Rosenbleuth’s
work, particularly his study of nervous spasms involving a progres-
sive loss of control. Rosenbleuth had found that in these condi-
tions the nerve signals were not being accurately processed. Wiener
realized that, similar to the antiaircraft gun that was slewing and
unable to track the moving plane, these nerve circuits were suffering
from feedback problems. The same principles that could be used to
understand automatic control systems should also be applicable to
neurology.
Rosenbleuth, together with Warren McCulloch (a leading neuro-
psychiatrist) and the logician Walter Pitts, had begun to develop a
new mathematics to describe networks of nerve cells (neurons) that
made up the brain’s information processing systems. McCulloch and
Pitts further demonstrated their theories by constructing the first
“neural network,” an electronic circuit whose components behave
in ways similar to neurons.
Neural networks would help answer a difficult question: How
does the brain make sense of the images created by the eyes’ arrays
of light-sensing cells? In other words, how does the brain recog-
nize a pattern (such as the numeral “5”) from the surrounding
background? This research also helped validate Wiener’s growing
belief that a single framework could be applied to control and com-
munication in living things, computers, and a coming generation
of robots.