It's not a question of analog vs. digital, but one of computational metaphor.
Since every programable system is built upon a virtual machine (production
system, random access machine, turing machine, or neural network), selecting
the right virutal machine simplifies the programming task. Some problems are
easily described in terms of Horn clauses, hence one chooses Prolog. Other
problems, where one is theory-poor but data-rich, are probably best solved
using neural networks.
As for the cybernetics, one can find a resurgence in similar issues in research
underway at the Ohio State Univ. and Indiana Univ. Rather than strictly
focusing on control issues, this ongoing research has been exploring the
relationships between dynamical systems and cognition.
John Kolen (kolen-j at cis.ohio-state.edu)| Wanted:
Laboratory for AI Research | Postdoc in
The Ohio State University | Neural Networks or
Columbus, Ohio 43210 (USA) | Complex Systems