342

24

The Nervous System

5. What are emotions?

6. What is intelligence?

7. How is time represented?

8. Why do brains sleep and dream?

9. How do the specialized systems integrate?

10. What is consciousness?

11. How do brains balance plasticity and retention?

all of which are still very much under discussion. At one time it was thought that

explicit simulation in silico of the brain would help to answer these and other

questions. ⁷

It should be emphasized that there is no evidence that biological neural systems

manifest the well-known principles of digital logic circuits. The same applies to

biological memory. It may be that neural computing is not even algorithmic. ⁸ In

computer terms, its operations appear to be partly analogue (depending critically on

the degree of excitation in the nerve fibres) and partly digital (depending critically

on the identity of which fibres are activated).

One is continuously confronted with the fact that our knowledge of the brain

is essentially privative. Some of the most pressing questions are as basic as how

information is coded in neural activity, how memories are stored and retrieved, and

what the baseline activity of the brain represents. Regarding this last question, it

appears that certain (marine) animals are able to analyse noise in their sensory inputs

sufficiently well to inform their hunt for prey, possibly involving stochastic resonance.

The brain is the supreme example of an object being observed—and doubtless

being influenced during observation—by itself. von Foerster (1973) has proposed

the postulate of epistemic homeostasis: “The nervous system as a whole is organized

in such a way (organizes itself in such a way) that it computes a stable reality”. The

self-referential character of the brain has already been noted in an earlier chapter.

24.3

Artificial Neural Networks

Artificial neural nets (ANN) are inspired by the study of real networks of neurons

but they have diverged from the latter and should not be considered a model of the

former. They are used for computation: for example, given a set of essential features,

one wishes to compute the identity of the object possessing those features.

An ANN typically consists of a number of individual cells (“neurons”) arranged

in layers and connected as follows (there are no intralayer connexions): in the first

layer there are as many cells as there are input parameters for the calculation that one

wishes to carry out. In the second and third (“hidden”) layers there should be a large

7 This was the goal of the ten-year Human Brain Project when it was launched by the European

Union in 2013; the original goal was later modified to become curation of existing data about the

brain. See also Fan and Markram (2019).

8 See Ramsden (2001) for more discussion.