Page 10 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

xxii Introduction

and decision-making. Our intent is to present the concepts involved to a target audi-
ence, who are not a narrow group of specialists working in the field but rather a broad
segment of the public intrigued by recent advances in AI.
The volume presents an introduction and 15 peer-reviewed contributions briefly
described in what follows.
In Chapter 1, Widrow et al. reconsider Hebbian learning, originally proposed in
the field of neurobiology as one of the basis of (unsupervised) adaptive algorithms
directly derived from nature. Although the LMS algorithm was previously proposed
by Widrow and Hoff as a supervised learning procedure, it can be implemented in an
unsupervised fashion. The two algorithms can thus be combined to form the Hebbian-
LMS unsupervised learning algorithm, which can be the key to interpret nature’s way
of learning at the neuron and synapse level.
In Chapter 2, Grossberg presents a survey of the main principles, architectures,
and circuits proposed by half a century of researches in the field, whose aim was
to develop a unified theory of brain and mind where the psychological perspective
can be read through the emergence of brain mechanisms. The chapter describes novel
revolutionary paradigms, like complementary computing and laminar computing,
with reference to the autonomous adaptive intelligence characteristic of the brain.
The chapter reanalyzes the fundamental approach of adaptive resonance theory
(ART) as a core model for engineering and technology, as well as to abstract insights
into mental disorders such as autism and Alzheimer disease.
Chapter 3 is the work presented by the AI Working Group spearheaded by H. Szu
and coordinated by M. Wardlaw under the aegis of the Office of Naval Research
(ONR). This work provides a vista of AI from the pioneering age in the 1960s starting
with narrowly defined rule-based systems through adaptive AI approaches using
supervised and unsupervised neural networks. They elaborate on third generation
AI based on the Zaheh-Freeman dynamic fuzzy theory, in which the Zadeh fuzzy
open sets and fuzzy membership functions are not predefined; rather they evolve
as the result of self-organizing recurrent chaotic neural networks, according to
Freeman neurodynamics. Their approach is truly human-centered and has the premise
to provide breakthroughs in AI beyond today’s cutting-edge DL.
In Chapter 4, Erdi presents an insightful review on the topic of hermeneutics
applied to brain science. Brainecomputeremind trichotomy is discussed, where
downward causality is discussed as a unique feature of brain-mind as opposed to
computation. Hermeneutics is introduced next, applied to the brain, and it is argued
that the brain is in principle a hermeneutic device. One application of this idea is the
explanation of schizophrenia, which is argued to be due to a broken hermeneutic
cycle. Finally, the chapter concludes with thoughts on how to achieve algorithms
for neural/mental hermeneutics. This is a deep theoretical essay that touches upon
fundamental issues in brain and neural sciences.
In Chapter 5, Ormandy addresses crucial issues related to the limitations of main-
stream AI and neural network technologies, especially in the context of the usefulness
of the AI in developing new technologies to improve our quality of life. He describes
the work started in collaboration with the late Walter Freeman in order to capture