Artificial Intelligence AskUser Mind-Module
Mentifex
the WhatAuxSDo module, the EnCog module,
and the KbRetro module. There needs to be
some sort of grammatical or even philosophical
trigger for the AI Mind to start asking questions
that are really a departure from whatever conversation
is already occurring. Currently two conditions must
be fulfilled before the high-level EnCog module
abandons normal thought and veers off into
http://code.google.com/p/mindforth/wiki/AskUser
to pose a question. The mind-designer or AI coder
arbitrarily establishes these conditions and may
therefore change these conditions. First the
http://www.scn.org/~mentifex/AiMind.html
must encounter a plural noun (e.g., "unicorns")
about which it can not find knowledge in the
knowledge base (KB). So the VerbPhrase module
calls WhatAuxSDo to ask, "What do unicorns do?"
Whatever the human user tells the AI Mind
about unicorns goes into the KB. Perhaps the
human user says, "Unicorns fight battles".
The AskUser module then takes part of the
response-idea and switches to a different
subject in order to form a query-idea such as
"Do robots fight battles" or "Do you fight battles?"
or "Do I fight battles?" The query-idea needs
only a "Yes" or "No" response from the human
user to impart knowledge to the AI Mind.
The KbRetro module acts upon the terse
yes-or-no response to adjust the knowledge
base retroactively. If the answer is "Yes",
http://code.google.com/p/mindforth/wiki/KbRetro
leaves the query-idea alone as a confirmed idea.
If the answer is "No", KbRetro keeps the
associations linking the concepts in the
query-idea but inserts an adverbial "NOT
modifier into the verb-engram to negate
the original query idea. If the original question
was "Do robots need food?", now the
knowledge stored is, "Robots do not need food".
If, on the other hand, the human user answers
"Maybe" or anything else different from "Yes"
or "No", the KbRetro module goes into the
original query-idea and zeroes out the
associative tags from concept to concept,
so that the AI Mind has gained neither a
positive affirmation of the idea nor a negative
denial of the idea.
Predecessor versions of the AskUser -- see
http://mind.sourceforge.net/ask.html --
module were in charge of coordinating the
various questions that the AI might ask.
More recently, concrete triggers call the
modules that ask the more concrete questions.
The AskUser module has taken on the role of
asking speculative questions. Right now the
http://code.google.com/p/mindforth/wiki/EnCog
module for English cognition calls the AskUser
module, but eventually it may be the prerogative
of the FreeWill volition module to embark upon
episodes of asking many questions occurring
to the emergent ConSciousness of the AI Mind.
The AskUser module enables the AI Mind
to conduct a search for information. Instead
of asking questions of a nearby or remote
human user, an inquiring AI Mind could just
as well conduct an Internet search for what
it wants to know. The Win32Forth
http://aimind-i.com program spawned from
MindForth along a different branch of AiEvolution
already has the ability to search the Web.
A mentifex-class AI translated (ported) into
Perl could reside on the Web and conduct
searches on behalf of other persons,
reasoning with Netizens to refine the search
and communicating the results within a
context of natural human language.
Mind-designers and AI programmers should
take note of a relatively new DeFault methodology
employed in the AskUser module. First the word
"ROBOTS" is a DeFault subject for any question
to be asked. Then the subjects rotate among the
concepts activated by the KbTraversal module,
to provide some variety and even some surprise
in the output of the AI Mind. Another DeFault is
that the module will use the word "anything" as
the direct object if the query-idea has failed to
make the transformation from user input to
speculative question. So instead of saying,
"Do you need food?" the AI might ask,
"Do you need anything?"
In our mind-design it is an arbitrary decision
to have the lead-up to AskUser involve first
the posing of a WhatAuxSDo question about
a plural subject, then waiting for a response
from the human user, and finally converting
the human response into a query-idea for the
asking of a question expecting a yes-or-no
answer. We use a plural mystery subject
like "unicorns" as the basis for asking
questions (e.g., "What do unicorns do?")
when we could just as well have seized upon
any singular subject to ask a similar question,
but we reserve the calling of WhoBe and WhatBe
for cases involving a singular subject for good
reasons. We want the AI Mind to start thinking
in terms of asking "What is..." about a singular
subject, and not plural subjects, so as to make
progress in the area of classifying concepts
into an ontology of things that exist in the world.
If a human user makes the first mention of a
unicorn in the singular, we want the AI to ask
the WhatBe question, "What is a unicorn?"
and not the WhatAuxSDo question, "What
does a unicorn do?" For purposes of machine
learning and machine reasoning, it is more
valuable to ask what something is rather than
what something does. Indeed, what something
does is a sub-element under the topic of what
something is.
Mentifex (Arthur)
--
http://www.scn.org/~mentifex/AiMind.html
http://cyborg.blogspot.com/2011/01/aiapp.html
http://www.chatbots.org/ai_zone/viewthread/240/
http://code.google.com/p/mindforth/wiki/AskUser