How we recognize good ideas

Harry Erwin erwin at trwacs.fp.trw.com
Mon Nov 1 18:04:28 EST 1993


I've been thinking about how we recognize a 'good idea.' If that portion
of the brain works like the olfactory system (which I think it does, in
outline), it goes as follows:

First of all, assume there is a single array where current ideas are
processed by the frontal cortex. This is probably invalid, but can serve
as an approximation for what really happens. Note that this array and all
the others discussed are 'content addressible memories' (CAMs) and consist
of subarrays that compete to 'recognize' signals. The current 'concepts' are
presented to this array in parallel. There is a second array that provides
a signal that is added to the 'concepts' signal. This added signal
consists of a large number of parallel mutually chaotic channels. The sum
signal is competitively evaluated by the subarrays of the frontal processing
array with one eventually winning out. The mechanism involving the frontal
processing array and the second array is a variant of 'stochastic
resonance' with the chaotic signal serving to complete the 'concept'
pattern presented to the frontal processing array. A third array interacts
with the frontal processing array and the second array to generate the
chaotic signal and evaluate the pattern identified by the subarrays of the
frontal processing array. The frontal processing array and the third array
can learn, but the second array does not. The frontal processing array is
trained by a combination of association (hebbian) and habituation
(non-hebbian). If the pattern identified by the frontal processing array
is not interesting, habituation occurs, leading to extinction of the
connections that identify it. If the pattern is interesting, associative
learning takes place. When a known pattern is identified, the signal from
the frontal processing array becomes synchronized, although it remains
relatively chaotic to unrelated signals.

The implication is that creativity is a chaotic process, involving the
'completion' of patterns by an additive chaotic signal in a fashion
similar to 'stochastic resonance.'

Cheers,
-- 
Harry Erwin
Internet: erwin at trwacs.fp.trw.com
          herwin at cs.gmu.edu
Working on Freeman nets....



More information about the Neur-sci mailing list