Prev: Call for papers (Deadline Extended): SETP-10, USA, July 2010
Next: Real Time Data Feed Design Questions
From: Mentifex on 6 Apr 2010 19:18
The Amazon sales rank for the renegade AI4U
textbook of artificial intelligence recently
showed a slight uptick indicating that some
genius AI devotee has prudently purchased the
"leading" (idea-wise, not sales-wise :-) AI
textbook in order to study open-source AI For You.

Each such instance of incremental AI funding
prompts the author of AI4U here to work
incrementally harder on advancing the state
of the art of the AI in AI4U.

http://www.scn.org/~mentifex/AiMind.html for
MSIE has yesterday been updated with a significant
change to the basic underlying mindgrid. The
enLexicon array now contains a male-female-neuter
"mfn" gender flag so that the EnPronoun mind-module
may (hopefully soon) begin to substitute the proper
choice of "he", "she" or "it" when referring to an
English noun.

This minor change has a major significance,
because non-English European languages (German,
Russian, French, etc.) require mechanisms for
thinking in terms of the gender of nouns as a
show-stopper item.

http://AIMind-i.com is potentially only the first
offshoot of the Mentifex open-source AI initiating
AI evolution along hereditary bloodlines and leading
who-can-tell to the triggering of the Singularity.
Hence this AI4U Singularity Alert is issued by

Very truly yours,

Mentifex
--
http://mind.sourceforge.net/ruby.html
http://www.scn.org/~mentifex/mindforth.txt
http://doi.acm.org/10.1145/307824.307853
http://doi.acm.org/10.1145/1052883.1052885

--- EXCERPTS FROM THE MINDFORTH PROGRAMMING JOURNAL ---

Tues.6.APR.2010 -- Updating the basic mindgrid structure

Yesterday in the AiMind.html JavaScript AI (JSAI) we
added the male-female-neuter "mfn" variable to the
enLexicon array so that the JSAI would be on a par
with the basic mindgrid structure of the MindForth AI.
Although we were eager to commence programming some
primitive code involving grammatical gender in MindForth,
we first installed the "mfn" flag in the JSAI structure
so that the JSAI would not lag too far behind the
cognitive architecture of MindForth.

Since we may be showing the way towards successful AI
implementations, we regard it as important to set the basic
structure of the evolving AI minds as early as possible.

Tues.6.APR.2010 -- Considerations of the use of "mfn"

When we introduced the EnPronoun mind-module in
the fp091229.html MFPJ entry of 29 December 2009,
we attained the immediate goal of having the AI
replace a plural English noun with the word "they"
in response to a "what-do-X-verb?" query. In our
history of AI development, it was a natural point
to begin substituting pronouns for nouns, because
question-and-answer dialogs would sound unnatural
without the use of pronouns. However, we were not
yet prepared to substitute "he", "she" or "it" for
singular English nouns, because we needed some kind
of gender flag to govern the proper selection of a
pronoun. Now with the insertion of the "mfn" flag
into both MindForth and the JSAI, we are ready to
implement the mechanisms of thinking with gender.
Various questions and concerns automatically arise.

The basic questions are how will the AI keep track
of gender, and how will the AI replace a singular
noun with its appropriate pronoun? Since we have
already demonstrated how to replace a plural noun
with "they" in our AI Mind code, the main question
is how to keep track of gender.

The cues for selecting a gender-pronoun must likely
come from two origins: the previously "stamped" or
"recorded" gender of a noun, and from hints carried
into the AI mind along with usages of the singular
noun in question. For instance, if a human user says that
"Mark is a boy", and the AI already knows that all boys
are masculine, then the AI should be able to assign a
masculine "mfn" tag to "Mark". However, such an inference
is actually getting ahead of the game, because there are
some much more primitive mechanisms available for
assigning "mfn" tags.

If we cause the AI to check for a pre-existing "mfn"
tag and to simply re-use it when a new instance of a
singular noun comes in, we ought to get the noun-gender
right for most instances of a noun, especially for
proper names. Of course, there are exceptions, as
for example when a name like "Pat" can be masculine
or feminine. But if a word like "the man" is stored
as masculine and the "mfn" tag keeps getting passed
forward, we should there have a pretty safe mechanism
for getting the gender right by using "he" as a
pronoun to refer to "the man".

More complex situations will arise in cases such as
when the AI must assign a gender flag to a noun because
a human user has used "he", "she" or "it" to talk about
the noun. If an AI in a science museum asks a human user,
"Where is your friend?", the answer could be "He is here"
or "She is here". We do not want the AI to assign a
long-term gender flag to the word "friend". Rather,
we would like there to be some sort of temporary
mechanism that assigns a gender flag only to a
temporary antecedent -- which may require the creation
of a new variable such as "tempant" to hold such data
as the temporary "mfn" value. Then there might be some
kind of override, so that the AI thinks in terms of the
temporary antecedent instead of in terms of the
gender-ambiguous noun "friend".

As we think about these cues and hints for selecting
a gender-prounoun, today we suddenly see a special
reason why a language like German tends to use a
definite article in talking about a person. For instance,
a German-speaker might say, "Der Ben ist hier, aber
die Natascha ist nicht hier." Students learning German
may think that such a custom is merely quaint, but AI
coders faced with implementing gender-cues may suddenly
see a special reason for the quaint custom. The colloquial
use of "der Ben" or "die Natascha" serves as an aide-memoire
to jolt the German-thinking mind into the immediate
assignment of the proper gender-flag by the other party
in a conversation. If one person asks, "Wo ist der Ben?",
it is easy for a responding person to answer, "Er ist
nicht hier," because the gender-cue is contained in the
very question. If you ask a German-speaker, "Wo ist Ben?",
the answer might have to be "Ben ist nicht hier", because
the respondent is rushing to provide information and does
not want to go the extra distance of letting a latent
gender-flag influence the response.

--- END OF EXCERPTS FROM THE MFPJ ----
 | 
Pages: 1
Prev: Call for papers (Deadline Extended): SETP-10, USA, July 2010
Next: Real Time Data Feed Design Questions