Computational Intelligence (CI)? In my humble opinion, before we
may ask 'quo vadis?', we should discuss the term CI itself. To be
honest, I am not very fond of it. All the paradigms that are
considered to be contained in CI like 'Fuzzy Set Theory', 'Neural
Networks', 'Evolutionary Computation/Genetic Algorithms', and so
forth, are purely descriptive. The word 'Intelligence', however,
suggests a rather uncritical judgement of the things comprised
under CI. I personally consider the word 'intelligence' (in this
context) as ambiguous and dangerous. I very well remember an
anecdote that was spread in some newsgroups a few months
ago. Even if it was a fake (what I don't know), it might remind us
all of situations we have already experienced ourselves. It was
said that the following dialog actually took place at a PhD defense
(the thesis was about something using neural networks) at an
American university:
Referee: Why do you call your method 'intelligent'?
Candidate: Because it uses neural networks.
Referee: If I managed to resemble what your ANN does with some other method, would this then be intelligent?
Candidate: No!
Referee: Why not?
Candidate: Because it would not be using neural networks!
In any case, it's still the human expertise that makes fuzzy
systems, ANNs, GAs, etc. work, it's not any method itself. The
methods themselves only do number crunching. 'Intelligence' is a
highly sensitive and subjective term that should not be used/abused
too carelessly.
I did not particularly like the term 'Soft Computing' (almost
nobody understood quite well what the 'soft' should have meant; but
everybody had at least some imagination what it could stand for),
but at least it did not contain any tendency of subjective
judgement like the term 'CI' does.
In any case, I do not feel that we need a special term for gluing
disciplines together that are quite independent of each other, not
even if they can be used together synergetically. It is, for
instance, quite a good deal to apply numerical methods to tuning of
fuzzy controllers, but would ever anybody come up with the idea to
say that numerical analysis is or should be a part of CI or Soft
Computing? Surely not!
(Ulrich Bodenhofer)