Page 86 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

74 CHAPTER 3 Third Gen AI as Human Experience Based Expert Systems

from which we have verified Donald O. Hebb’s learning rule, in the Ergodicity
ensemble average sense, which was formulated six decades ago, in the field of
brain neurophysiology. Given a time asynchronous increment ¼jDtj, the learning
.
plasticity adjustment is proportional to the presynaptic firing rate S i and the post-
.
synaptic glue force g .
j
Theorem Asynchronous robot team and their convergence proof: If and only
there exists a global optimization scalar cost function H int. known as Helmholtz
Free Energy at isothermal equilibrium to each robot member, then each follows
asynchronously its own clock time in Newton-like dynamics at its own time frame
“t j ¼ ε j t”; ε j 1 time causality with respect to its own initial boundary conditions
with respect to the global clock time “t.”

d W i;j vH int:
; (3.17)
¼
dt j v W i;j
Proof: The overall system is that force changing synaptic first order Hebb rule as
the acceleration which is convergent guaranteed by a quadratic A.M. Lyaponov
function:

2

dH X vH d W i;j X
0; ε j 0 time causality: Q:E:D:
vH
¼ ε j ¼ ε j
dt v W i;j dt j v W i;j
j j
.
0 1

v W i;j vH vH vD j . .

D W i;j ¼ h ¼ h ¼ . $ @ A hh g S i h (3.18)
j
vt j v W i;j v W i;j
DD j
This Hebb learning rule may be extended by chain rule for multiple layer
“Backprop algorithm” among neurons and glial cells

old . .
j
W i;j ¼ W i;j þ g S i h (3.19)
We can conceptually borrow from Albert Einstein the space-time equivalent spe-
cial relativity to trade the individual time life experience with the spatially distrib-
uted experiences gathered by asynchronously massively parallel distributed
(AMPD) computing through Cloud databases with variety of initial and boundary
conditions. Also, Einstein said that “Science has nothing to do with the truth (a
domain of theology); but the consistency.” That’s how we can define the glial cells
for the first time consistently (Eq. 3.8).

4. CONCLUSION

We shall illustrate a smaller size feature processing after the back of our head Cortex
17 area V1eV4 layers of feature extraction, these feature feed is underneath the con-
trol of hypothalamus-pituitary gland center; there are two walnut/kidney-shape