An early version of MindForth from ten years ago is archived at
http://mind.sourceforge.net/mf980528.html for the purpose of
letting Forth AI enthusiasts examine the same MindForth version
described by Dr. Paul Frenger in his ACM Sigplan Notices paper,
Mind.Forth: Thoughts on Artificial Intelligence and Forth (1998).

Whosoever wishes may capture incremental releases of MindForth and
follow the same filename system (e.g., mf080901.html) to provide
public archives as a community service on the Web. The original
author of MindForth -- very truly yours Mentifex here -- always
wanted to publish only the very latest, most advanced AI code and
was therefore reluctant to maintain embarassing public archives
of his previous efforts. The trouble is, at any point in Forth AI
evolution, Mentifex here could make a horrendously wrong decision
and thoroughly ruin the quasi-official, public MindForth codebase.
There was always a certain amount of "Junk DNA" in the AI code,
anyway, which we frequently try to remove by means of housekeeping.

In the grandiose plan of Mentifex for Forth AI taking over the world,
any high-tech corporate CEO who does not assign a team of hotshot
AI programmers to "embrace and extend" Mind.Forth is missing the boat.
Whether you are the team leader or simply Joe Appcoder, it behooves
you to make your own local copies of MindForth releases, each with
a unique filename, so that you may look good making your presentation
in the corporate boardroom on the state of the art in AI for robots.
Even if your corporate AI diverges widely from Mentifex AI releases,
you may at some time be tasked with tracking some phenomenon of mind
as it lurks quietly in the Forthmind codebase before springing to life
in a burst of highly focussed development.

The 31.AUG.2008 MFPJ webpage discusses the declaration
of entelechy goals for the MindForth Article module, so that a
work-in-progress presents a statement of the target destination.
Now is a good time to draw up entelechy goals for the audRecog
mind-module, which is crucially important for all the new things
we want to do in MindForth AI, such as recognizing both singular
and plural noun-forms for the same psi concept; recognizing
multiple roots within compound nouns; and detecting prefixes,
infixes and suffixes. Accordingly we fire up 1sep08A.F MindForth
and we commence composing the audRecog entelechy goal statement.

\ The audRecog module aims for the following entelechy goals.
\ [ ] Recognize animal vocalizations as well as human speech.
\ [ ] Receive Morse and other codes as well as speech or text.
\ [ ] Detect human foreign-language accents as per language.
\ [ ] Recognize particular voices of particular individuals.
\ [ ] Switch from keyboard input to recognizing acoustic speech.
\ [ ] Recognize words despite slight variations from correct form.
\ [ ] Detect prefixes, infixes and suffixes.
\ [ ] Recognize multiple roots within compound nouns.
\ [ ] Recognize both singular and plural forms of the same noun.
\ [X] Recognize plural noun forms as a word leading to a concept.
\ audRecog (auditory recognition) comparator is called from audSTM
\ to recognize words by matching input phonemes against memory.
\ Anyone may code gusRecog, olfRecog, tacRecog, or visRecog.
:  audRecog ( auditory Recognition )  \  1sep2008

In our code of 29.AUG.2008 we cloned the verbPhrase module
into a predicate module that the ENGLISH module calls
after formulating a noun-phrase. Previously, SVO had
been calling both nounPhrase and verbPhrase to construct
a Subject-Verb-Object sentence with a transitive verb.
By getting away from SVO and by using nounPhrase and
Predicate separately as subroutines of English, we permit
either an intransitive verb of being or a transitive verb
of action to finish the thought emanating from a subject-noun.

The predicate module needs to use its selection of either
a transitive or intransitive verb as the deciding factor
in whether or not to call a direct-object noun, or to call
a noun or adjective serving as a predicate nominative.

In Predicate there could be a kind of doubly conditional
branching depending on what comes back from the search for
a motjuste verb. If a highly active, transitive verb is
found, the search will continue for a direct-object noun.
In a secondary conditional branching, failure to find a
highly active verb of any sort, transitive or intransitive,
shall default to choosing an intransitive verb of being.

Before we introduced the predicate module, verbPhrase was
reporting detour flag settings up to the SVO module, so that
SVO could abort and call Ask for a question instead of stating
a transitive verb and its direct object. If Predicate is called
from ENGLISH, it no longer makes sense to deal with the
positive detour flags in the SVO module.

For both pathways, transitive verb or verb of being, failure
to find either a direct object or a predicate nominative should
result in a detour flag setting.

Actually, we can now dispense with the "What do [blanks] do?"
question that we have been using up until now in the case of
a detour away from a missing verb, and we can switch to asking
questions with a verb of being, such as "What is a [blank]?" or
"What are [blanks]?"

We may be ready now to keep a verb active even when it is missing
a direct object. For instance, if we are discussing what various
animals eat, the AI may start thinking, "Unicorns eat...", and
not be able to finish the answer. Then we would like the AI to
detour into a question of, "What do unicorns eat?"

Our basic issue right now with the predicate module is, what
will prompt the AI to use a verb of being in its thinking?
If a human user has entered a "What is...?" question,
then obviously the AI is predisposed to pick up the use
of the same verb of being.

Because our earliest AI Minds have used only plural nouns
and plural verb-forms to demonstrate proof-of-concept thinking,
we have perhaps conditioned ourselves to think too much in
terms of plural forms and transitive verbs. If we were to
reformulate our (still) primitive cognitive architecture
in terms of singular forms, we might have a better approach
to the use of verbs of being. For instance, if we made the
detection of knowledge-gaps detour not into questions like
"What do [blanks] do?" but rather "What is a [blank]?" or
"What are [blanks]?", then the internal use of a verb of being
would be just as conducive to further usage as the input of
a verb of being from a human user. [We might even have an
"Essence" module for verbs of being.]

After some vacillation, we have converted the whatAuxSDo
module into a whatIs module, to ask "What is...?" questions
rather than "What do... do?" and we have modified the ASK
module to call whatIs rather than whatAuxSDo.

The code that we upload today is greatly improved over
the confusion of our last upload. Our latest code has a
more stable functionality and shows the approach that we
are adopting to the issue of how our AI Minds shall be nudged
into using intransitive verbs of being instead of transitive verbs
of action. Since human input of "What is...?" questions can
prompt an AI Mind to use a verb of being in response, we get
the AI to pose its own internal questions in the same format.
A third source of instigation for verbs of being would be
episodes of wonder and astonishment on the part of sentient
robots outfitted with MindForth AI software.

For discussion of MindForth, visit the Usenet newsgroups
comp.lang.forth
comp.robotics.misc