[Comp-neuro] spontaneously emerging cognitive capabilities

gros07 at itp.uni-frankfurt.de gros07 at itp.uni-frankfurt.de
Thu Aug 14 11:45:16 CEST 2008


Randy recently stated in the ongoing discussion thread:
   `There is no way you are going to figure out something as
    complex as the brain using a purely bottom-up strategy.'

I fully agree and I would like to generalize this systems
perspective to a meta-level:

In the long run, most of us believe, the human cognitive system 
will turn out to be just one of the many possible realizations of
fully developed cognitive architectures. The set of fundamental 
principles giving rise to the human cognitive system and their 
realizations via specific neural circuitry will turn out, in the 
long run, to be only one of possibly many governing principles
for highly developed cognitive systems, which would allow a
living being to successfully survive and master a complex and
changing world.

When taking this systems point of view, one will immediately
ask meta-questions like the following: The brain is self-active
and most of us believe, that the sensory stimuli does not
drive the overall neural activity, having in general only 
a modulating effect on the global brain dynamics. It is then
not surprising that an emerging field in both systems neuroscience
and in the study of artificial cognitive architectures is the
study of neural architectures which are spontaneously active
on their own.

A typical question from a systems point of view would then, to
give an example from my own research, which principles govern
the interactions between the neural dynamics of an autonomously
active neural architecture and the sensory input data stream.

I believe that these kind of research questions can be highly
illuminating. We found recently, that this kind of approach
lead to the discovery of `emerging cognitive capabilities'.

To be specific, we found that letting a system with a well
defined autonomous neural activity interact with sensory
inputs, results in the system performing a non-linear
independent component analysis of its own. We believe this
result to be remarkable. The standard approach is to write 
specialized codes for algorithms like the ICA, and not to
obtain them as spontaneously emergent cognitive capability,
resulting just from the general meta-principles used for the
layout of the neural architecture.

- Claudius

*****************************************
*** Prof. Dr. Claudius Gros           ***
*** http://itp.uni-frankfurt.de/~gros ***
*****************************************


More information about the Comp-neuro mailing list