The Model-less Neuromimetic Chip and its Normalization of Neuroscience
and Artificial Intelligence
Abstract
The conceptual basis of a novel neuromimetic chip is described. Based on
an existing computational bioelectrodynamics study and adaptive brain
signaling biophysics knowledge from neuroscience, the chip is, in
effect, a form of inorganic artificial brain tissue. This ‘physics
replication’ approach involves no abstract models of brain tissue
physics or function. Instead of the physics of a general-purpose
computer or the physics of abstract models of the brain on the chip
(analogue or digital), this neuromimetic chip has an inorganic version
of natural adaptive brain signaling physics. As a result of using the
native brain physics, the chip has functionally relevant endogenous
quasistatic electric and magnetic field systems of the form known to be
expressed by excitable cell tissue. Fully developed at macroscopic
scales it can be expected to produce an EEG/MEG-like electromagnetic
signature. This article does an extended analysis to understand an
observed generalized lack of the physics-replication approach and its
implications for the neuroscience of natural and artificial
intelligence. This is achieved through a technical comparison with the
neuromimetic chip’s closest relative, the neuromorphic chip (of the
class of general-purpose computers). The results indicate that the
physics-replication approach is a possible but neglected option. It also
reveals that the neuromimetic chip contributes empirical science, in
contrast to the theoretical science conducted using general-purpose
computers. Because of the chip’s novelty and proximity to foundational
issues, the article contributes necessary background information in
anticipation of the arrival of the first prototyping results over the
coming years.