Page 105 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

CHAPTER

From Synapses to

Ephapsis: Embodied 5

Cognition and Wearable

Personal Assistants

Roman Ormandy
Embody Corporation, Los Gatos, CA, United States

CHAPTER OUTLINE
1. Neural Networks and Neural Fields .......................................................................93
2. Ephapsis..............................................................................................................95
3. Embodied Cognition..............................................................................................97
4. Wearable Personal Assistants.............................................................................105
References .............................................................................................................109

1. NEURAL NETWORKS AND NEURAL FIELDS
Neural networks are designed to perform Hebbian learning, changing weights on
synapses according to the principle “neurons which fire together, wire together.”
The end result, after a period of training, is a static circuit optimized for recognition
of a specific pattern. There is plenty of evidence that mammal neocortex indeed
performs Hebbian learning. It turns out however that mammal neocortex does
much more than simply change the weights of individual neurons. Populations of
interacting neurons are behaving quite differently than individual neurons or static
neural networks. Theory for the role of neuron populations and ephapsis in human
cognition comes from the seminal work of Walter Freeman, Giuseppe Vitiello, and
Robert Kozma [1], as well as the ideas of Gerald Edelman [2], Gyorgy Buzsaki [3],
and many others.
In his 1975 book Mass Action in the Nervous system, Walter Freeman [4]
starts with a measurement of population of neurons (by EEG) rather than individ-
ual neurons (in neuron electrode). While an individual neuron generates trans-
membrane potential, a group of neurons generates extracellular potential field.
Whereas the path inside the dendrites is private for the neuron, the path outside
is very public. The same path is shared via ion channels for the loop currents of
all neurons in the neighborhood, so that electrodes placed outside the neurons
measure the cortical potential of the whole neighborhood. While electrodes inside

93
Artificial Intelligence in the Age of Neural Networks and Brain Computing. https://doi.org/10.1016/B978-0-12-815480-9.00005-0
Copyright © 2019 Elsevier Inc. All rights reserved.