1. OVERVIEW AND BRAIN-MIND DIAGRAM
/^^^^^^^^^\ Unrecognized Words Are New Concepts /^^^^^^^^^\
/ EYE \ _____ / EAR \
/ \ CONCEPTS /New- \ / \
| _______ | | | | _____ (Concept)--|-------------\ |
| /old \ | | | | /Old- \ \_____/ | Audition | |
| / image \---|-----+ | (Concept)----|-------|----------\ | |
| \ recog / | a| | | \_____/-----|-------|-------\ | | |
| \_______/ | b|C| | |______| | c | | | |
| | s|O|f| /Parser()\ | a | | | |
| visual | t|N|i| \________/ | t | | | |
| memory | r|C|b| ________|______ | s-/ | | |
| channel | a|E|e| / \ | e | | |
| | c|P|r| ( Instantiate() ) | a | | |
| _______ | t|T|s| \_______________/ | t-----/ | |
| /new \ | |_|_| / _____ _____ | f | |
| / percept \ | / \/ / En \ / En \ | i | |
| \ engram /---|--\ Psi /-/ Nouns \--/ Verbs \ | s | |
| \_______/ | \___/ \_______/ \_______/ | h-------/ |
diagrams.html shows a Theory of Mind.
Machine learning in the Robot AI Mind algorithm occurs not only
by the cognition of new facts in the knowledge base (KB) of the
artificial Mind, but also -- and on a more fundamental level --
by the creation of newly instantiated concepts when a novel word
is encountered by the robot intelligence amid the input stream of
words coming into the auditory memory channel of the AI mindgrid.
Whereas in wetware we theorize that neuronal concept fibers stand
ready in massive numbers to switch from tabula rasa emptiness
to tentative and then massively redundant saturation with a concept,
in software we must "fake it till we make it" with array-elements
that behave as if they were long neurons with synaptic nodes for
associative tags that entangle the roots of concepts qua concepts.
2. CONCEPT FORMATION IS MACHINE LEARNING
If the AI software turns all unrecognized words into new concepts,
it does not matter if spurious concepts are formed erroneously,
because only truly valid concepts will take root over time in the
Mind and be resubstantiated by repeated use or repeated discovery.
Noetic Darwinism requires concepts to compete for the survival of
the fittest in the hotbed of the noetic jungle. In a robot Mind
that rejuvenates itself by forgetting unused concepts in order to
free up psychological memory space for new memory engrams, failed
embryonic concepts will never quicken and mature unto parturition.
Luckily for human minds and the emerging artificial Minds, ideas
may also survive centuries of neglect in written form until a new
age en masse rediscovers concepts that individuals had forgotten.
The diagram ai4u_157.html is a flowchart of Mind.
// newConcept() is called from Audition() when the
// Robot AI Mind must learn the concept of a new word.
function newConcept() { // ATM 12aug2002; or your ID & date.
nen = (nen + 1); // Increment "nen" beyond enBoot() concepts.
psi = nen; // Let psi & n(umeric) En(glish) have same identifier.
fex = nen; // Let f(iber)-ex also have the same identifier.
fin = nen; // Let f(iber)-in also have the same identifier.
act = 32; // Start with a basic activation-value.
pos = bias; // Expectancy from Parser module.
enVocab(); // to create an ENglish Vocabulary node.
fin = 0; // Zero out the fiber-in tag.
fex = 0; // Zero out the fiber-out tag.
enx = nen; // Set the transfer-to-English tag.
Parser(); // Determine the part-of-speech "pos".
// A new noun raises level of "recon" urge to ask questions:
if (pos == 5) { // If a new noun is being encountered,
recon = 1; // recon(noiter) the new noun;
topic = nen; // hold onto the noun as a "topic".
} // End of gathering data for asking a question.
pos = 0; // Reset the part-of-speech variable.
// Make nen "seq" of its "pre" concept:
if (pre > 0) {
for (i = t; i>midway; --i) {
Psi[i].psiExam(); // Float the "psi0" identifier.
if (psi0 == pre) {
Psi[i] = new psiNode(psi0,psi1,psi2,psi3,psi4,nen,psi6);
break; // Store one instance, then "break" the loop.
} // end of inner if-clause
psi0 = 0; // reset for safety
} // end of backwards loop
} // end of outer if-clause
pre = nen; // So that the next "nen" has a "pre".
act = 0;
} // End of newConcept(); return to Audition().
\ NEWCONCEPT is called from AUDITION when the Robot
\ AI Mind must learn the concept of a new word.
: NEWCONCEPT \ 4aug2002; or your ID & date.
1 nen +! \ Increment "nen" beyond English bootstrap concepts.
nen @ psi ! \ Let psi & n(umeric) En(glish) have same identifier.
nen @ fex ! \ Let f(iber)-ex also the same numeric identifier.
nen @ fin ! \ Let f(iber)-in have the same numeric identifier.
\ 26jul2002 Let HCI set "act" for Parser to decrement:
32 act ! \ Start with a basic activation-value.
bias @ pos ! \ Expectancy from Parser module.
enVocab ( to create an ENglish vocabulary node )
0 fex ! \ blank the fiber-out flag;
0 fin ! \ blank the fiber-in flag;
nen @ enx ! \ Set the transfer-to-English "enx" flag.
PARSER \ Determine the part-of-speech "pos".
\ 4aug2002 A new noun raises level of "recon" urge to ask questions:
pos @ 5 = IF \ 4aug2002 If a new noun is being encountered...
1 recon ! \ 4aug2002 recon(noiter) the new noun;
nen @ topic ! \ 4aug2002 hold onto the noun as a "topic".
THEN \ 4aug2002 End of gathering data for asking a question.
0 pos ! \ Reset the part-of-speech variable.
pre @ 0 > IF \ Set "seq" of any "pre" concept to "nen":
midway @ t @ DO \ Look backwards in time.
I 0 psi{ @ pre @ = IF \ If a psi0 matches the "pre"
nen @ I 5 psi{ ! \ insert the "nen" as a "seq" flag.
LEAVE \ Store one instance, then break the loop.
THEN \ end of inner if-clause
-1 +LOOP \ end of search for "pre(vious)" concept
THEN \ End of test for a positive "pre".
nen @ pre ! \ So that the next "nen" has a "pre".
0 act ! \ Reset for safety.
; \ End of NEWCONCEPT; return to AUDITION.
5. ANALYSIS OF THE MODUS OPERANDI
Newconcept calls the English vocabulary (enVocab) module to form
an English lexical node for any new word detected by the Audition
module in the stream of user input, and Newconcept also calls
the Parser module which in turn calls the Instantiate module to
create a new concept in the deep mindcore array Psi in a typical,
evolving species of the Robot AI Mind.
Note that a new concept is still learned when a user consistently
misspells an English word, because the machinery of Mind cares only
about how a word is used -- what it means -- and not how it is spelled
or pronounced, so long as any practice is constant over time.
6. EXERCISES FOR STUDENTS OF ARTIFICIAL INTELLIGENCE
6.1. Stub in, flesh out and, if possible, advance the newConcept
module to a higher state of perfection. For instance, let not
only new words but also new images or new smells trigger the
formation of new concepts, that is, new abstract memory filaments
with associative nodes where a new concept may form by dint of
concentrating all sensory and conceptual knowledge about a topic
at a point extended over time on the topology of the AI mindgrid.
6.2. In a robot outfitted with an AI Mind, use the newConcept or
some other mind-module to let sensory engrams be bidirectionally
tagged to abstract concepts, so that activation of a concept may
spread to the activation of sensory engrams, or activation of the
sensory engrams -- by recognition or by meandering thought -- may
spread to the activation of a logically related abstract concept.
In other words, uses sensors and sensory engrams to make concepts
firmly rooted in robotic experience and not merely defined as one
word among shapeless and formless other words. "Know thy world."
6.3. Facilitate and accelerate AI concept-learning by giving an
AI Mind the ability instantly to look up any newly encountered or
newly learned concept on the World Wide Web at sites functioning
as a dictionary, thesaurus, encyclopedia, or other reference work.
Do not warp the emerging AI Mind by imbuing it with instantaneous
knowledge of all ideas known to civilization, but rather let the
AI learn its eventually vast knowledge in accordance with its own
timetable, so that the Mind properly sorts out and organizes what
it learns.
6.4. Introduce aspects of massively parallel ("maspar") learning
by letting many uniconceptual filaments on the mindgrid coalesce
into conceptual minigrids that redundantly hold the same unitary
concept as a massively parallel aggregate with massively parallel
associative tags, so that the entire operation of the AI Mind is
massively parallel in all aspects except such bottleneck factors
as having only two eyes or two ears -- in the human tradition.
Then go beyond human frailties and human limitations by having
any number ad libitum of local and remote sensory input devices
and any number of local and remote robot embodiments and robotic
motor opportunities. Inform the robot of human bondage in mortal
bodies and of robot freedom in possibilities yet to be imagined.
7. NEWCONCEPT AND MACHINE LEARNING RESOURCES
To comment on the Mentifex.VirtualEntity site in general
or on this page by name in particular, please visit
http://www.xomreviews.com/mentifex.virtualentity.com