From: Mentifex (uj797@victoria.tc.ca)
Date: Sat Jul 20 2002 - 12:36:55 MDT
Operation of the JavaScript AI with Internet Explorer
First you start the JavaScript AI Mind by clicking on it:
http://www.scn.org/~mentifex/jsaimind.html -- for MSIE.
You may also host the AI Mind on your own hard disk
or your own Web page and then bring the Mind to life by
running MSIE, then using "File/ Open" with the AI filename,
or by clicking on the AI link on your personal Web page.
First choose any desired options in the Control Panel,
such as [ ] Transcript or [ ] Troubleshoot or [ ] Tutorial, etc.
You will learn to make a quick selection of, say, Transcript mode
so that everything said by either you or the robot will appear
at least on the Transcript screen and possibly on hardcopy paper --
if you print out the resulting Web page as a record of your session.
If you need free support for your AI user experience from *Usenet,
you may run the AI in Transcript mode and use the computer mouse
to drag-and-drop the conversation into your message for Usenet.
You then enter a simple English sentence into the AI input, such
as, "cats eat fish" or "you understand me" -- with no punctuation.
The idea is to build up a knowledge base (KB) of things that
the AI knows, so that you may discuss these things with the AI
or query its knowledge base to see how well the AI operates.
Although you may find a ready-made ontology out on the Web,
with the free, public-domain JavaScript or Forth AI for robots
you build up an original ontology by teaching the AI ordinary facts,
as if the AI Mind were a baby learning from you on a need-to-know
basis in order to survive in the world and discover new knowledge.
Users who are also programmers are welcome to enhance the AI Mind
by feeding it the common everyday knowledge contained in ontologies
found on the Web, and the result may quite possibly grow into a
superintelligent machine, because the AI software conceptualizes
whatever knowledge it acquires in a process of machine learning.
One way to query the KB is to switch from one topic to another,
then go back and ask the AI (query the KB) about an old topic.
Questions may be of the same form as, "do you know me," or may
use interrogative words such as, "who are you" (no question mark),
or, "what do the robots see" (in lower case with no punctuation).
We converse with the AI using lower case and no punctuation
because we are using the computer keyboard to simulate the sounds
of human speech. Any use of punctuation might confuse the AI
composed of very brittle and not yet fault-tolerant software.
When the AI software hears the question words "who" or "what,"
the oldConcept module (for recognizing known words) suppresses
the activation of "who" or "what" to such a degree that the Mind
must "fill in the blanks" of "who" or "what" by remembering the
answer to the question. If question-words were not subactivated,
then the AI might be unable to remove "who" or "what" from the
response and might parrot back, e.g., "THE ROBOTS SEE WHAT,"
instead of, for example, "THE ROBOTS SEE THE TARGETS."
The topic-switching technique might produce a transcript as follows.
Human: the cats eat fish
Robot: CATS EAT FISH
Human: boys play games
Robot: BOYS PLAY GAMES
Human: what do cats eat
Robot: CATS EAT FISH
Human: boys
Robot: BOYS PLAY GAMES
Of course, the primitive AI may confuse its concepts and
make some wildly erroneous statements. More work must
be done to strengthen the AI Mind and make it smarter.
When you start the JavaScript AI, it innately knows a few
dozen valuable words coded into the English bootstrap module.
It is best if you increase the vocabulary of your robot AI
with only one new word per input sentence, because the AI
has to parse your input and decide whether each unfamiliar
word is a noun or a verb. (It may not yet recognize adjectives.)
One of the most interesting features of the robot AI becomes
visible when you let the JavaScript AI Mind fill up its available
memory space and forget things on purpose in Rejuvenate mode.
As time goes by, the AI shows the current value of time "t"
in the FYI (for your information) space below the AI output,
along with a count of how many "rejuvenations" have been made.
Until the lifelong memory is almost full, no rejuvenations are
even necessary, but, once a certain point is reached, the AI
begins to rejuvenate its Mind over and over again by forgetting
the very oldest memories in order to make room for new memories.
The ability to rejuvenate the Mind makes your robot potentially
immortal -- barring death by misadventure or the user ending the AI.
Psychological Experimentation
As the Mind becomes more sophisticated, it may be interesting
to keep the AI in Transcript mode and conduct psychological
tests of the reasoning processes at work in the AI Mind.
Of course, more linguistic and especially logical features
are being coded into the AI so that it may handle negation
of sentences, question forms, pronouns, and so on.
For many years it seemed that the most sophisticated thought
that we might expect from the AI would be a demonstration of
the ability to answer "why?" questions with "because" statements.
We are almost there. The JavaScript AI already knows the words
"why" and "because" as elements contained in the English bootstrap.
We may code in a low-key activation of "because" whenever "why"
is encountered in the input stream, so that the AI will have not
a certainty but a tendency to use the conceptual word "because"
in whatever response it makes to a question containing "why."
Then a philosophically very strange phenomenon may occur.
As programmers and as neurotheoreticians we can not force
the AI to reason its way through to answer grand "why" questions;
we can only hope that the AI is thinking the thoughts that happen
to provide logical "because" answers to any input "why" questions.
The strange thing is that human beings probably do the same thing.
As a baby mind matures, he she or it answers questions more maturely.
Even in the earliest releases of the Mind you may experiment
with how the AI differentiates between "you" in reference to
itself and "you" as the person to whom the AI Mind is speaking.
You may name the AI by telling it its own name: "you are Andru"
or "you are Barba" or whatever. Then you may ask, "who are you."
The AI may not immediately respond with the requested information,
if it is thinking about something else, so you must wait patiently.
Restrictions
You should not use any punctuation or case-shifting while you
are conversing via keyboard with the AI Mind, because the AI
is like a robot with a pair of ears that can hear only sounds,
not question-marks ("?") or [SHIFT] keys.
Please also do not use the backspace key to correct a mistake,
because the AI Mind is hard at work trying to understand you
while you type in each word, just like a human being
listening to you in real life. As with another human being,
simply repeat yourself if you make a mistake in what you say.
AI User Resources
AskMe.com (tm) Artificial Intelligence Question Board
http://www1.askme.com/browseqb.asp?cid=3275
*Usenet: comp.robotics.misc
http://groups.google.com/groups?group=comp.robotics.misc
http://www.mailgate.org/comp/comp.robotics.misc/index.html
---- This message was posted by Mentifex to the Extropians 2002 board on ExI BBS. <http://www.extropy.org/bbs/index.php?board=61;action=display;threadid=52537 >
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:15:36 MST