VisRecog is the most primitive and yet the most sophisticated of
all the mind-modules in the MindForth AI. It is primitive because
it is only a stub. It is sophisticated because it is the first
mind-module to bridge the gap between pure reason and physical
embodiment of the AI Mind in a corporeal robot. The Mentifex AI
project does not yet have a physical robot or even a robotic
camera (botcam) for the AI Mind to look through and for VisRecog
to recognize objects and report their names to the conscious AI.
Therefore, for lack of physical implementation, VisRecog exists
just barely as a software implementation and as a Grand Challenge
for the best and the brightest among AI coders and roboticists to
flesh out the VisRecog design with a real robot and real robotic
eyes.
Computer vision is regarded as an intelligence in its own right,
comparable in complexity and in algorithmic difficulty to the
cognitive mechanisms of thinking with natural human language.
The stubbing in of VisRecog suddenly became possible on
21 September 2011 when it was necessary to work on the problem
of English verbs and how they might be linked dynamically to
direct objects not previously available in the knowledge base (KB)
of the AI Mind. Previously for a dozen years MindForth had found
the direct object of a transitive verb by searching backwards in
the knowledge base. Now suddenly it was necessary and urgent to
implement the possibility of an AI Mind thinking in real time
and describing dynamically what it sees through its robot eyes.
Therefore the primitive but extremely sophisticated VisRecog
gets its foot in the sensorimotor door by offering a profound
choice to the linguistic VerbPhrase module trying to answer
the simple but AI-coder-terrifying question of "What do you see?"
As of 22 September 2011 you may start the AI MindForth program
and ask it what it sees. Since the AI Mind has no robot eyes
and therefore cannot see, the VerbPhrase module branches off
not into NounPhrase for a knowledge-base answer but into
VisRecog for a DeFault (wiki-page spelling) answer of "NOTHING".
The simple but sophisticated answer will be, "I SEE NOTHING".
But because there is now a VisRecog module waiting for the
next Donald Knuth or Jorn Barger to take up the problem and
to apply the art of computer programming to the grand challenge,
and maybe to write a weblog post about it, now that weblogs have
been invented and have multiplied by the millions, there may
someday be an AI Mind that identifies for you any object that
you hold in front of its Cyclopsean eye. But first there had to be
the stub of VisRecog on the AI MindGrid with a DeFault of NOTHING
for what it sees in the outside world.
=== Code ===
\ The visual recognition module of MindForth AI for robots
\ when fully implemented will serve the purpose of letting
\ AI Minds dynamically describe what they see in real time
\ instead of fetching knowledge from the AI knowledge base.
: VisRecog ( identification of objects seen by a robot )
svo3 @ 0 = IF \ if no direct object is available;
midway @ t @ DO \ search for an automatic default
I 0 en{ @ 51 = IF \ 51=NOTHING 22sep2011
I 7 en{ @ aud ! \ hold address for SpeechAct
LEAVE \ search no further after finding one engram
THEN ( http://aimind-i.com )
-1 +LOOP \ end of looping through English lexical array
THEN ( http://opencv.willowgarage.com )
; ( http://code.google.com/p/mindforth/wiki/VisRecog )
Mentifex (Arthur)
--
http://www.scn.org/~mentifex/AiMind.html
http://www.chatbots.org/ai_zone/viewthread/240/
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com
