Moving-Wave Algorithm of Mentifex Mind.Forth AI
Mentifex Mind.Forth AI is adopting a Moving-Wave Algorithm
of activation spreading over the robot mindgrid.
Activation Rules -- Moving Like a Wave Through the Mindgrid
1. With A New Concept
For new concepts being learned as input, the HCI module has the
"uract" variable with a value of, say, 32, that functions in the
Parser module to assign activation "act" going into Instantiate.
It may be possible to stop using "uract" and use only "act".
Since a new concept may be learned at any time, both old concepts
and new concepts should receive the same general activation level.
In a radical departure from previous lore, we should perhaps
consider giving new concepts only a partial, incomplete activation.
After all, what the Mind itself is thinking, is more important
than what a keyboard user may be saying to the Mind. Any input
idea is already receiving attention from the Mind, but must
compete with thoughts-in-progress for domination of the
sovereign chains of thought of the Mind.
When the robot Mind listens to the human keyboard user,
the input information is duly recorded as associatively
tagged engrams in the auditory memory channel. Whether
or not the information of human origin sways the Mind of
the robot, is a matter of the state of mind of the robot.
2. With Known Concepts
http://mind.sourceforge.net/oldcept.html oldConcept module
recognizes a known word.
http://mind.sourceforge.net/activate.html Activate module
imposes __ units of activation.
passes __ units to "pre" and "seq" concepts.
3. During Generation of a Sentence of Thought
Activation only really matters during generation of thoughts,
not during the reception of input from a human keyboard user.
A case could be made for the notion that a question entered by
a human user should receive a higher activation than a mere
statement entered by a human user. Various question words or
auxiliary verbs could trigger the higher activation.
A threshhold of activation must be surpassed for inclusion of concept
in a sentence. The psiDecay module constantly decrements activation
of unused concepts. "Bulge" occurs when both noun and verb pass
activation to a direct object. If multiple successive inputs use
different subjects but the same verb, routing of activations must
still select the proper associations among concepts.
The generation process itself may spur the activation
of each next word in a sentence. It is not necessary
to pre-activate all the words in a sentence. The
process of NLP generation "chases" activation from
http://mind.sourceforge.net/svo.html concept to concept.
The general idea for sustaining chains of thought is that winning
concepts selected for thought-generation shall be assigned such
transitory-but-high activations that the winning concepts manage to
do their work before they, too, succumb to the psiDecay assaults.
The notion of the Moving Wave Algorithm conforms to the general
that everything is inhibited unless specifically activated.
The http://mind.sourceforge.net/psidecay.html module tries
to suppress the activation of concepts, but the meandering
chain of thought "goes against the grain" by passing the
baton of activation from concept to concept.
http://doi.acm.org/10.1145/1052883.1052885 -- Assoc. for Comp. Mach.
http://mind.sourceforge.net/mind4th.html is the free AI source code.
http://www.914pcbots.com has the A.I. Zone of "Mentifex Communicat".
To unsubscribe, change your address, or temporarily deactivate your
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]