> Of course, most of the limitations of NN can be avoid by generalizing
> the concept to such a level. However, at the same time, such a general
> notion does not support the claims of advantages of NN, either. How
> can someone argue that this concept can compete with the classical
> symbolic model of cognition? Can semantic network be counted as a kind
> of neural network? Once again, when the extension of a concept gets
> larger and larger, its intension gets smaller and smaller.

The way I think about it, a neural net is a dynamical system composed
of connected components that roughly model neurons.  The system's
dynamics have got to take place via equations that update the
quantitative parameters of the simulated neurons and synapses, and the
system's interface with the outside world has got to take place via
input and output neurons receiving "raw perceptual stimuli" and
outputting "raw actuator controls."

Of course, there are some NN architectures that don't fulfill this
definition, but IMO they are cheating and are just "NN-ish"...

If an NN architecture like this creates an internal "semantic network"
via self-organization, that's fine.

Also, if you understood the dynamics of the NN well enough to
explicitly encode knowledge in the NN in a way that worked effectively
within its neuronlike dynamics, that would still be within the NN
paradigm.

Typical semantic network architectures are not based on closely
neuronlike components, so they don't qualify as NN's in my view.

I believe that probabilistic logic operations can be boiled down to NN
operations in some way, and I've posted some partial ideas in this
regard before (Google "Hebbian Logic"), but this line of reasoning has
not been effectively completed by me or anyone else yet, and IMO until
one has either

a) theoretically understood how uncertain logic is grounded in NN's, or

b) figured out how the brain represents and manipulates complex knowledge

then as I said I think NN-based AGI is a non-starter.

-- Ben

-- Ben

-------
To unsubscribe, change your address, or temporarily deactivate your 
subscription, 
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]

Reply via email to