Page 11 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

Introduction xxiii

the dynamics of embodied human cognition and incorporate it to novel wearable
personal assistants. The author reviews the literature from file theories through
embodied cognition across species. The main thesis of this work is the critical impor-
tance of ephaptic interactions between neural populations, which produce neural
fields measurable by noninvasive means, thus providing an opportunity for the devel-
opment of wearable personal assistants in everyday life, including augmented mem-
ory, stress relief, fitness training, relaxation, and other applications.
In Chapter 6, Kasabov presents an approach based on evolutionary connectionist
systems (ECOS) that are able to evolve their architecture and functionality in adaptive
data-driven modality. Evolving spiking neural networks (eSNN) are illustrated and
proposed as a third generation of artificial neural networks (ANN). eSNN can be
used for future brain-like AI, and the NeuCube architecture is presented as a
machine that can implement DL procedures. The chapter also proposes a combination
of AI and ANN approaches as a unique method derived from neuroscience.
In Chapter 7, Brown et al. focused on the pitfalls and opportunities of developing
techniques of evaluation of AI systems. This is a cool topic, considering the relevant
progresses that DL methodologies have introduced in the computational intelligence
community. However, they raise the problem of measuring and comparing perfor-
mance. The receiver operating characteristic (ROC) paradigm and the bootstrap
method are considered as well-grounded approaches for performance metric in order
to avoid overestimation that frequently limits the practical implementations of AI.
In Chapter 8, Werbos provides an impressive summary and a critical review of the
reasons underlying today’s wave of AI successes, including DL and the Internet-of-
Things (IoT). As a major exponent both in research and research funding in the past
decades, he provides an exciting insider’s view of these developments, as well as
points toward possible avenues in future research, neuroscience in particular. He
also points out the key role researchers play in applying the novel technological
development for the benefit of humanity.
In Chapter 9, Levine reviews the history of neural networks as an artificial model
of brain and mind. Neural networks are a paradigm that in principle link biology and
technology: it thus comes as no surprise that the flagship journal of the International
Neural Network Society is indeed Neural Networks. His chapter reconsiders the orig-
inal ideas that motivate the nascence of this interdisciplinary society at the light of
present developments.
In Chapter 10, Kozma touches a fundamental problem humans have focused on
for centuries, that of creating machines that act like them. The chapter investigates
various aspects of biological and AI issues and introduces a balanced approach based
on the concepts of complementarity and multistability as manifested in human brain
operation and cognitive processing. As intelligence in human brains is the result of a
delicate balance between fragmentation of local components and global dominance
of coherent overall states, the chapter elaborates on how intelligence is manifested
through the emergence of a sequence of coherent metastable amplitude patterns.
This behavior leads to the cinematic theory of human cognition that both provides