Bill has commented on how connectionists et al. basically throw up their
hands and give up on understanding what is going on under the hood and how
their systems accomplish what they evolve to accomplish.
This was my take too when I first started looking at connectionism, and I
still share that view, though on a less informed basis than Bill.
It occurs to me that in PCT the same sort of indeterminacy gets shunted
into the development of input functions. A given neural signal may
contribute to the input functions of several very different elementary
perceptual control systems (ECS). That's where our conception of what goes
on in the brain gets especially mushy. This may apply as well to output
functions, but that is not so clear to me. The output from a given ECS may
be distributed to the reference inputs of diverse other control systems or
to effectors.
Bill has commented on how connectionists et al. basically throw up their
hands and give up on understanding what is going on under the hood and how
their systems accomplish what they evolve to accomplish.
...
It occurs to me that in PCT the same sort of indeterminacy gets shunted
into the development of input functions. A given neural signal may
contribute to the input functions of several very different elementary
perceptual control systems (ECS). That's where our conception of what goes
on in the brain gets especially mushy. This may apply as well to output
functions, but that is not so clear to me. The output from a given ECS may
be distributed to the reference inputs of diverse other control systems or
to effectors.
Actually this is pretty well worked out in PCT; Rick's spreadsheet models
demonstrate how the hierarchy can work when _every_ system at one level
receives copies of perceptual signals from _every_ next-lower system, and
sends reference signals to _every_ next-lower comparator. In my Byte
articles, I showed the same thing, but with fewer systems.
Of course we're still mushy about what really goes on in the brain. But at
least we know one way it could work.