Mind.forth Programming Journal: 13 DEC 1998

jfox at ricochet.net jfox at ricochet.net
Fri Dec 18 10:01:00 EST 1998


In article <3676C2E8.3F62C72B at nccn.net>,
  The Name <adoyson at nccn.net> wrote:
> I want to add A.I. to my Robot.
>
> What exectly is MIND.FORTH? The only page I can find about it is the
> SOURCE file. What INPUT do you give the program?  Only TYPED in from a
> keyboard? When You give it input, what does it output?  Is it a
> language/sentence processor?  More?

Mind.Forth is the second implementation of Mr. Murray's theory of
artificial mind.  I was impressed by the first implementation of the
theory in Mind.Rexx.  I spoke to the author who interestingly enough
had just extended Rexx to a parallel language with the Linda parallel
programming extension.  I encouraged the people working on the project
to port the mind to Forth because the Forth version is so much smaller
and simpler and it opens the door to porting to Forth hardware designed
for simulation of neural systems.

I am not quite as wordy as Mr. Murray and rather than go into explainations
of all the theory when asked the question what is Mind.Forth I would like
to say that it demonstrates one of my theories that what looks like a
mind can be built with a few reflexes.  The idea is a bit disturbing
and few people make it past the verbose language and grandious claims
to get to the simple but profound concepts behind the Mentifex artificial
mind model.

> I don't use Forth, never have.  I do have access to a Forth program for
> the computer which my robot has ONBOARD, but before I go to the trouble
> of preparing the program, I want to know what Mind.Forth does!

I doesn't have everything that was in Mind.Rexx, but it has some
new stuff.  What exactly does a human mind do?

> Will any Forth (?compiler?) run the mind.forth program correctly?

No.  To begin with it uses Blocks for source.  Secondly it is an older
Forth dialect and although an effort was made to write it portably
in Forth it is still in the very early stages of development and is
primarily meant to demonstate that it is not just a theory.  Porting
might take a little effort at this time.

Because the Mentifex consciousness is like simple reflex it responds
to any input the same way.  The first version takes typed English
and responds to it almost as if it understands English! It is much
easier to deal with balance, touch, hearing, and sight this way
than English.  It trivializes construction of the self awareness down
to reflexes in simple easy to understand Forth hierarchy.

Building artificial insects that have enough reflexes to crawl around
on the floor was a divergence from the traditional AI approach of
trying to model a God like consciousness first and then implement it
on a machine.  Mind.Forth demonstates that you can build something
that tackles a much more difficult problem in AI, understanding and
dialog in a language, that is as simple conceputally as other people's
implementations of artificial bug minds.

Despite the difficult to understand documentation on the Mentifex
project I saw a striking similarity to the AI programs that I had
written with a simple English Expert system sitting on top of
a bunch of small Forth reflex routines could fake intelligence
very well.  Using this approach I was able to get tiny machines to
recognize me when I walked into the room and carry on a dialog in
spoken English.  Many years later we are still teaching people
to write long programs to get their rug-bot to reverse direction
when it runs into something.  In fact using learning expert system in
my bot could recognize the difference between a person and a cat
and between me and another person.  It learned how to dial into
my account at the bank and negociate changes in the bank computer
systems and adapt to them. It would dial in and send me pictures
on my PC if it ever saw anything suspicous in my house when I
was not there.  It also did sound a speech recognition and response.
It could answer the question, "why do you think that?" or "why did
you do that?" by tracing back trough the logic trees and printing
out the associated strings.

That system was also setup to answer all the mail from my boss at
Bank of America.  He never knew the difference.  My coworkers
were always sending me puzzles in the mail to try to see if the
response was coming from me or a machine but my boss was the only
person who was always carrying on a dialog with a simple program
in a simple machine.  Because I let the learning expert figure
out how to get into the banks systems, adapt to changes, and
answer my bosses mail from watching me do it.  Because I had written
the thing in Forth I was many months ahead of all of the people
dealing with other languages and departments.  I didn't even
really program it.  I just let it learn what I had to do to keep
my boss happy and let it do it for me.

My design had a less generalized consciousness and more specialed
reflexes than the Mentifex model.  Mind.Forth combines the high level
perception and language processing with the same mechanism as reflex.
Mentifex is really quite an interesting theory of mind operation.

--
Jeff Fox   Ultra Technology
www.UltraTechnology.com

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own    



More information about the Neur-sci mailing list