Frank wrote:
> ATM,
>
> I have implemented your changes from Jan - Feb.
> What I am noticing now is all responses use the object
> of the input SVO as the subject of the response.
> Is that what you were intending?
Yes, but things are getting much trickier now.
In the subject-verb-object (SVO) sequence,
a (direct) object in the AI knowledge base may not
have an associated verb and downstream object
to keep the chain of thought going.
For that reason (lack of continuity in thought),
the MindForth program up to (most recent) 1feb2008
includes a "detour" variable and code for "backing out"
or detouring away from an un-finishable thought --
one where the subject has no associated verb.
However, even backing out of unfinishable thoughts
is not sufficient for demonstrating AI Mind functionality.
We need additional algorithmic code to make the
chain of thought not only abandon (detour from)
an incomplete thought, but to "jump" from a
dead--end line of thought to other food for thought
in the AI knowledge base. (If the AI Mind was installed
in a robot, the Sensorium would provide the "food".)
But I digressed in February 2008 from coding to writing
http://mentifex.virtualentity.com/m4thuser.html
-- the AI
User Manual, burgeoning up to 26 print-out pages.
Now I really want to update both Mind.Forth and
http://mentifex.virtualentity.com/Mind.html
in JavaScript,
and I suppose it is only a matter of time and labor.
> I have not walked through the data to see why this is happening.
> But I think it has to do with the values of the activation used,
> since they were changed in the past few months.
Once again I think you're right. On my agenda is a task
of experimenting with the activation-values so as to
zero in on the optimal values (least spurious thinking)
for all outset values arbitrarily set by the AI coder.
> I hope to be able to get a grip on it before the
> weekend is out.
>
> Thoughts?
>
> Frank
> [URL corrected to] http://AiMind-I.com
My thoughts are that right now we are probably both
pretty much at the same stage in our AI development.
I have not yet finished the MindForth User Manual.
http://mentifex.virtualentity.com/aiumprep.html
is
where I am still embedding links in the Manual.
FJR, this enquiry from you bolsters my motivation
to get some new work done in the near future.
I will announce whatever I come up with in c.l.f
or in the Yahoo win32forth group, or both.
Bye for now,
Arthur T. Murray
--
http://mentifex.virtualentity.com/mind4th.html
|
