I’ve recently grown obsessed again with the requirements of
accurately simulating the neurons role in *informational*
processing. We do know that a majority of a neurons
function is devoted to repair and self maintenance.
However, the question of how much of a neurons function is
directly related to cognitive processing remains in dispute.
If you ask Hans Moravec he remains content is equating each
neuron with a single transistor. He bases this claim on his
analysis of the ocular neurons role in visual processing.
The problem with this assumption is that the simple
processing of visual data near the ocular region neglects
the remaining processing that’s required to assess the
context and meaning of that data and integrating it within
the larger realm of cognitive life itself, to say nothing of
emotional or other more subtle psychological states.
My assertion is that the computation required within each
neuron is far greater than can be met by a single
transistor. A typical neuron has several thousand synapses
connected to other neurons mediated by a complex array of
neurotransmitter activity. Each neuron is responsible for
regulating the transmission and reception of signals along
these networks. Although synaptic transmission is quantal
because an integer number of vesicles must fuse to release
their neurotransmitters, the decision when to fire is
determined ultimately by the degree and types of
neurotransmitters involved. This is why various types of
psychoactive drugs from Prozac to Marijuana have remarkably
different effects on cognition and mood.
**My contention then, is that although the pattern of the
neural network determines in part who and what we are, the
complex neurotransmitter activity within the neurons
themselves also play a significant role in the quality of
our subjective states. Therefore there is no way a single
transistor has a chance of duplicating.
So the question remains what amount of computational
machinery is required to replace the current role that a
neuron plays? Any takers?
Paul Hughes
This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:09:37 MDT