Using Concepts in a function seems like an interesting idea.  However,
there are a few things about the idea which makes it different from
the application of mathematical functions.  A single result for an
interaction of two concepts is a little too hard edged.  In a
recognition problem, for example, a single restricted result might be
too restricted to be of much help.  But the idea of a function does
not have to be limited to processes which produces a single value.  So
a Conceptual Function might create a list of many different values.
This is similar to things that were tried in the early days of AI.
However, using the model of types, operations, operands and functions
in computer programming as an analogous generator to this idea of
Conceptual Functions I was able to realize that I might take
Conceptual Types and use them to produce different kinds of lists
depending on the situation.

Often we need to use knowledge in different kinds of way.  For
recognition problems, which I consider to be a major obstacle to
producing simple AGI, the use of Conceptual Types in Conceptual
Functions which can produce lists of possible 'meanings' (including
other kinds of useful results) might be very effective.  So my idea of
a Conceptual Type, which was drawn from linguistics theories, is
innovative because I am thinking of employing it to a much greater
extent then it was typically used in computational linguistics.  And
recognizing that there is a need to use knowledge (of a subject) in
different ways for different situations I am also thinking of using
the idea to produce different kinds of lists relative to the stage of
the cognitive task that it 'thinks' it is in.  For example, there are
different kinds of generalizations.  There are generalizations which
refer to a well defined categorical class and then there are
generalizations which are vague and so on.

The other things that make my ideas innovative are that I would use
this along with:
Imaginative projection.
Trial and error learning.
Conceptual structural learning (which is based on an extensive
collection of learned Conceptual Types).
Anaphoric-like analysis across temporal interactions.
User behavioral analysis (very lightweight behavioral analysis based
on some tracking of discussions or other interactions). - Jim Bromer


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to