While talking about this in the Reddit group I learned something and I
came to a few conclusions.
Shaping was a term Skinner used. But I think I probably read some
papers that were from a cognitive psychology perspective and they may
have tried to explain shaping using a non-Behaviorist approach.

Anyway I want to program some underlying behavior into my AI project.
For instance, a simple naïve game-like behavior that I am thinking
about could be used to focus the program's attention on a subject and
get it to examine relationships around that subject. (This supposes
that I would be able to use this and other behaviors in my text-based
AI project to get it to recognize words or phrases that represent
simple subjects.) There will also be other behaviors that will
sometimes occur in the naïve program. And I am hoping I will be able
to develop the program so that these simple behaviors will tend to be
used more knowledgably and become goal-driven.

But the first thing is that I do not want these behaviors to be
logically equivalent to deterministic program instructions. So,
'reinforcing' the programmed behaviors will not be perfect or
perfectly predictable. At any rate I am trying to say that there are
ways to design internal pre-programmed behaviors to allow the user to
shape some of the learning the program will do and most importantly,
to help it to attend to some simple basic ideas that might otherwise
be elusive. But this will not be just another form of programming. It
will be more like teaching, Right now I am just trying to write a
simple program to prove that the concept is fundamentally viable.

The second thing I realized is that, we don't teach very young
children the basics of the abstractions of thinking. They do need to
develop strategies which will allow them to acquire some insight
independently but these abilities have to available for
knowledge-driven (and goal-driven) interests. So while I (or another
user) should be able to recognize some primitive behaviors which can
be used to focus the program's attention on some subject, the
abstractions or thought processes which can be used with a particular
subject will tend to be subject-centric as they will be most
effectively employed for those particular subjects. There will be
similarities of these subject processes since they will be partly
composed of the same elementary processes. But they won't be
completely composed of the elementary processes because as they are
subject-centric and they will need to be specialized to be used
insightfully on those subjects.


-------------------------------------------
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