1.
Theory of
AI4U
Algorithm Steps: Code the wtAuxSDo Mind-Module
/^^^^^^^^^\ Generation of "what do fish do?" /^^^^^^^^^\ / EYE \ Pronoun +AuxVerb +Subject +Verb / EAR \ / \ | | | ___________ / \ | _______ | | | | / Sentence \ | | | /"fish" \ | | | | \___________/-------|-------------\ | | / percept \---|---|-+ | | \ | | | | \ engram / | d|f|w| | _\____ | | | | \_______/ | o|i|h| | / Verb \ | | | | | |s|a| | ( Phrase ) | | | | visual | |h|t| | /\______/ | auditory | | | | | | | | / / | | | | | memory | | | | ___|__ / / | | memory | | | | | | | / Noun \/ _/_ | | | | | channel | | | |( Phrase ) /Aux\ | | channel | | | | | | | \______/ \___/ | | ____ | | | | | | | | \ | | / \ | | | | | | | | \ | | /"what"\ | | | _______ | | | | | _\_|_ | / "do" \ | | | /fresh \ | |_|_| _|___ / En \ | / "fish" \| | | / image \ | / \ / En \ / Verbs \-|-\ "do" / | | \ engram /---|--\ Psi /-/ Nouns \ \_______/ | \ / | | \_______/ | \___/ \_______/------------|---\______/ |http://mind.sourceforge.net/diagrams.html shows a Theory of Mind.
The what-do-Subjects-do (whatAuxSDo) module of
the AI Mind has evolved adaptively since its introduction
in 2002 as a subroutine of the Ask module for asking questions.
As of 2008, the rigid structure of whatAuxSDo has not changed,
but the initial functionality of asking a question about any
new noun has become less rigid. Since the operation of
whatAuxSDo is to ask a question about any "topic" fed to it
by the thinking mechanism of the AI Mind, it was a natural
progression from making any new noun the topic of questions
to the more sophisticated functionality of using whatAuxSDo
as the first default option in serious of default options
when the more evolved AI Mind has insufficient knowledge
to think a thought about any noun in the knowledge base (KB).
2. WHAT DO QUESTIONS DO?
Human: Questions generate problems.
Robot: What do problems do?
Human: Problems invite solutions.
Robot: What do solutions do?
Human: Solutions cause questions.
In the more advanced AI Minds of 2008 and beyond,
questions give the human attendant of an AI the chance
to fill in gaps in the knowledge base (KB) of the AI.
As a proof-of-concept artificial intelligence, the AI
is designed to display meandering chains of thought
by thinking up sentence after sentence in a chain of
thought -- what we humans might call "musing" or
"reflection" or "daydreaming." We humans easily lapse
from one topic of thought to another, but an AI Mind
composed of software needs some initially rigid rules
for the selection of topics of thought. Later, as AI
Minds evolve even further, the initally rigid rules
may relax and give way to more humanlike modes of thought,
but as of 2008 the current mind-design ordains what the AI
is supposed to do if it tries to generate a thought
about a topic as the subject of a sentence for which
no verb springs to mind. Rather than shanghai some
low-activation verb which would express nonsense,
the AI rejects the incipient thought, backs out of it,
and follows a "detour" pathway into a series of default
options -- of which the first default is the whatAuxSDo
mind-module, because whatAuxSDo gives the machine
intelligence a chance to learn something, thereby
increasing its knoweldge base (KB). Other defaults
may include self-referential egocentric thinking
just to keep the meandering chain of thought going,
or a more sophisticated KB-traversal of considering
all available tidbits of thought that can be generated by
means of a systematic traversal of the knowledge-base.
3. ALL YOUR SUPERCOMPUTER ARE BELONG TO US
The wtAuxSDo module, if resident on a sufficiently powerful super-
computer, can already -- on an as-is basis -- with no further evo-
lution and no further speciation of the Seed AI into Sons of AI --
fill a stupendously enormous memory space with stupidly phrased
and maddeningly limited ideas on the order of the simple inter-
change presented as a quasi-transcript above. Now, no cyborg in
its right Mind would want to have such a mentally deficient AI,
but we humans tend to get excited and yell "Eureka!" when such a
possibility has been demonstrated, because there are hordes of
brainy programmers out there who may even now be souping up the
jalopy engines of Mind-1.1 computers into know-all supercomputers.
http://mind.sourceforge.net/ai4u_157.html
is an overview of Mind.
// wtAuxSDo() is a question of "What do (nouns) do?".
function wtAuxSDo() { // ATM 12aug2002; or your ID & date.
tov = t; // time-of-vice for echoing input.
for (i = t; i>midway; i--) { // Look backwards for 54="what".
enLexicon[i].enExam(); // Inspect the English nodes.
if (en0 == 54) { // If the #54 concept "what" is found...
motjuste = 54; // "nen" concept #54 for "what";
aud = en5; // Auditory recall-vector for "what".
// break;
} // End of search for #54 "what".
} // End of loop finding the word "what".
Speech(); // Speak or display the word "what".
// Call form of auxiliary verb "do":
auxVerb(); // Any of several auxiliary verbs.
for (i = t; i>midway; i--) { // Look backwards for "topic".
enLexicon[i].enExam(); // Inspect the English nodes.
if (en0 == topic) { // If the "topic" concept is found...
motjuste = topic; // "nen" concept #"topic";
aud = en5; // Auditory recall-vector for "topic".
// break;
} // End of search for #"topic".
} // End of loop finding the lexical "topic" item.
Speech(); // Speak or display the lexical "topic".
for (i = t; i>midway; i--) { // Look backwards for 59="do".
enLexicon[i].enExam(); // Inspect the English nodes.
if (en0 == 59) { // If the #59 concept "do" is found...
motjuste = 59; // "nen" concept #59 for "do";
aud = en5; // Auditory recall-vector for "do".
// break;
} // End of search for #59 "do".
} // End of loop finding the verb "do".
Speech(); // Speak or display the verb "do".
topic = ""; // Reset "topic" after using it.
} // End of wtAuxSDo(); return to the Ask() module.
\ whatAuxSDo is: what + Auxiliary + Subject + "Do":
: whatAuxSDo \ ATM 4aug2002; or your ID & date.
\ Call interrogative pronoun "what":
midway @ t @ DO \ Look backwards for 54=what.
I 0 en{ @ 54 = IF \ If #54 "what" is found,
54 motjuste ! \ "nen" concept #54 for "what".
I 5 en{ @ aud ! \ Recall-vector for "what".
LEAVE \ Use the most recent engram of "what".
THEN \ End of search for #54 "what".
-1 +LOOP \ End of loop finding the word "what".
SPEECH \ Speak or display the word "what".
\ Call form of auxiliary verb "do":
auxVerb \ 4aug2002 Any of several auxiliary verbs.
midway @ t @ DO \ Look backwards for "topic".
I 0 en{ @ topic @ = IF \ If "topic" is found,
motjuste @ topic ! \ "nen" concept #"topic";
I 5 en{ @ aud ! \ Auditory recall-vector for "topic".
LEAVE
THEN \ End of search for #"topic".
-1 +LOOP \ End of loop finding the lexical "topic" item.
SPEECH \ Speak or display the lexical "topic".
midway @ t @ DO \ Look backwards for 59=do.
I 0 en{ @ 59 = IF \ If #59 "do" is found,
59 motjuste ! \ "nen" concept #59 for "do".
I 5 en{ @ aud ! \ Recall-vector for "do".
LEAVE \ Use the most recent engram of "do".
THEN \ End of search for #59 "do".
-1 +LOOP \ End of loop finding auxiliary verb "do".
SPEECH \ Speak or display the auxiliary verb "do".
; \ End of whatAuxSDo; return to ASK.
6. Analysis of the Modus Operandi
The wtAuxSDo module is not a permanent feature of the artificial
mind, but is rather a contrived mechanism to show how an AI mind
may ask questions and receive answers in a process of machine learning.
Operation of the module is triggered by the input of an unknown noun,
and therefore the module operates outside of the normal rules of
conceptual activation.
The "what-do-Subjects-Do?" (wtAuxSDo) code is carefully organized
so as to be as normative as possible and as permissive of further
tweaking or development as possible. The helping-word "do" is
not directly hard-coded into the algorithm but is instead treated
as the results of a search for an auxiliary verb in the "auxVerb"
module -- so that "may" or "must" could just as easily be chosen.
As we sail into uncharted territory in the development of Seed AI,
we are well-advised to provide an AI scaffolding upon which other
voyagers may build in the future.
You may pour your heart and soul into devising intricate AI code,
only to monitor the Web and watch "AI: The Next Generation" rip
your treasured source code to shreds, but think:
Robot: What do pioneers do?
Human: Pioneers blaze trails.
7. Troubleshooting and Robotic Psychosurgery
In Mind.Forth (as opposed to the JavaScript Mind.html AI),
it is possible just to run whatAuxSDO by itself after
a run of the AI Mind has shown a need to troubleshoot
the AI. Entering whatAuxSDo and [RETURN] will show the
basic format of a question as generated by the module.
8. wtAuxSDo Module Resources for
Seed AI Germination and Evolution
