Some Frequently Asked Questions (FAQ) about the Mentifex AI project are published here to tell the truth and set the record straight. Readers may suggest additional questions in the Comments section. Read on for all you ever wanted to know about Mentifex but were too embarassed to ask.
1. What was the Mentifex AI project?
The Mentifex (Latin for "mindmaker") project in artificial intelligence was
an attempt to create artificial minds in computers and in robots.
Arthur T. Murray (ATM) promulgated the Mentifex AI project in such a way
that the AI work was always out in the open and could be obtained or
worked on further by any interested party.
2. Who was Arthur T. Murray?
Arthur T. Murray (ATM) was an independent scholar with a bachelor of arts degree
in ancient Greek and Latin. As a child at Fort Gulick in the Panama Canal Zone,
ATM became interested in such mysteries as exploring the Panamanian jungle all
around him; electricity; and foreign languages such as the Spanish language to
which he was exposed as a child living on an American army base in Panama.
As ATM grew up in the USA, his interests from childhood focussed more sharply
in adulthood. Exploring the jungle gave way to exploring the ultimate mysteries
of the intellect such as, what is the mind? Experimenting with electricity yielded
in rapid progression to making electronic devices such as radio receivers, then a
radio transmitter, then computers. The interest in foreign languages led to fluency
in Latin; German; Russian and Greek; and to a desire to explore the nature of mind
by making a computer that could think in natural human language.
3. Why did ATM succeed at AI where others failed?
ATM did not succeed at AI. His Mentifex AI project failed miserably again and
again, time after time, until success or failure no longer mattered. What mattered
was not to create AI but to try to create AI.
The artificial mind created by ATM in
Forth and in
JavaScript could barely think.
The Mind-1.1 software as described in the AI4U textbook was so primitive an
artificial intelligence at first that only a few computer programmers tried to
advance the state of the art beyond what ATM/Mentifex had failed at.
4. What other AI projects were spawned by the Mind of Mentifex?
A host of world-class, genius programmers created some associated AI projects.
http://sourceforge.net/projects/javamind
http://sourceforge.net/projects/mindai
http://sourceforge.net/projects/mindjava
http://www.advogato.org/proj/Mind AI/
http://www.angelfire.com/nf/vision/ai/mjava.html
http://www.virtualentity.com/mind/vb/
5. Why did ATM/Mentifex write the AI4U textbook?
Webpages to document the AI Mind modules began sprouting in 1998
and went through many iterations of updating and improvement.
When iUniverse.com appeared on the scene as a venue for independent
scholars and autonomous authors to publish whatever they saw fit,
ATM/Mentifex saw fit to publish a book describing the Mentifex AI project.
6. Why would anyone buy AI4U?
The mind-diagrams at the start of thirty-four chapters are the main incentive
to invest in the purchase of the
AI4U textbook of artificial intelligence,
because the human mind is so difficult to understand that it helps to have
all the diagrams of the artificial mind available for the intense study and
concentrated effort that may take years before the seeker finds enlightenment.
Along the way to nirvana and teleostasis, other reasons to acquire AI4U occur.
The strange, disturbingly weird book on artificial intelligence for robots
makes the ultimate nerd-gift from one nerd to another, or weirdo to weirdo, or
older relative to young techie relative at Christmas, birthday or graduation.
As a rare-book investment, a hard-bound first-edition copy of AI4U with
a well-preserved, original dust-jacket bearing the ASCII mind-module diagram
has probably no potential for appreciation in the future, so only a fool would buy up
dozens of
AI4U first editions and hoard them to dribble back onto the after-market.
7. What is the Mentifex AITree or Cognitive Architecture?
The
AITree is a somewhat repetitive list of AI mind-modules showing
the hierarchy of upper mind-modules calling or invoking lower modules.
Since the AITree is extracted from the working, thinking AI Mind program,
it looks like a program and it displays a cognitive architecture for AI.
8. Why does ATM/Mentifex advocate the rewriting of computer science texts?
If it's broken, why not fix it? If you are writing any form of instructional
material on, say, programming languages, why not try to write AI-ready material?
Suppose that you are teaching a course on the XYZ programming language, and
you decide to issue a printed hand-out with sample code about XYZ programming.
If you couch your instruction in terms of artificial intelligence problems,
then serendipitously you achieve several simultaneous objectives. First of all,
you teach the basic material such as, say, how to create a loop in XYZ code.
By having the loop cycle through stubs named after the Mentifex AI mind-modules,
you prepare your students and your own career for the Technological Singularity.
When the purchasers of textbooks take aim from the Texas School Book Depository,
won't they zero in on your textbook that already addresses open questions in AI?
9. Why does ATM/Mentifex so often mingle politics with science?
As an expression of a philosphy of life, an AI project shares resources and
attention with other endeavors in life, such as being a political animal.
Aristotle says, man is by nature a "politikon zoon" -- a political animal.
10. Why did ATM/Mentifex post in so many Usenet newsgroups?
The Mentifex AI message was obviously germane and on-topic in each newsgroup.
In
comp.arch the idea was to broach a new kind of computer architecture.
In
comp.sys.super it was imperative to recruit supercomputer AI resources.
In
alt.folklore.computers there was an outreach to the living fossils-in-residence.
11. Why did ATM/Mentifex always insist on discussing AI in public forums?
The two main reasons were that ATM wanted to pass the torch of AI to other carriers
and that ATM felt that AI was too important to all humans to work on in secret.
anyone who is serious about simulating real intelligence (rather than doing it artificially) should consider reading the ancient indian scriptures - the ayurvedic literature.
i understand from [different] people who have read these sources that 1) there are 27 separate and distinct nervous systems of the brain, each of which is described in a chapter of this ayurvedic literature, which is, according to some sources, tens of thousands of years old (!!!)
and 2) that intelligence simulations will prove successful when an entire nervous system is simulated, not just one part.
also, here's a clue, which if you grok quantum mechanics you may appreciate: PHASE MATTERS.
for example, the layout of the nerve paths and the lengths of the nerve paths from the eyes to the visual cortex are spread out from BOTH eyes across the entire cortex.
thus you end up with a diffraction grating effect, where the pulsing of signals from the eyes at aroung 12 cycles per second causes phase transitions in the visual cortex.
we know that "changes" are noticed by our eyes, and in fact if you stare at the same space without moving for long periods of time - over five minutes - you end up with tunnel vision and eventually everything disappears!!!
our eyes jiggle slightly. they flick across images. in other words, phase transitions are induced.
in short, i believe that any attempt to produce "artificial" intelligence that relies on static analysis is doomed to failure before it begins.
by introducing "dynamic stability" at the risk of also introducing chaos and howling feedback is there any possibility of success.
and yes, it would be necessary to provide psychological and psychiatric support of such systems!!!
i'm happy to go into a bit more detail on this subject if anyone's interested.
I've got an alternate theory of the mind. This is my text book.
int
main(int argc, char ** argv)
{
state_t state;
while (1) {
sense(&state);
do_smart_stuff(&state);
effect(&state);
}
return 0;
}
Exercises.
1) the student should recode it in the language of their choice. and then
think about how AI's coded in lisp would feel about AI's coded in Visual
Basic. Then write an essay about racism as it applies to AI's.
2) you may also want to change "while (1)" to "while (not_dead)" since
if the AI were immortal it might take over the world and kill humanity.
3) the student should then code sense(), do_smart_stuff(), and effect().
do_smart_stuff should use semantic networks and spreading activation,
and be careful not to read any papers about their limitations or other
approaches to AI.
5) when the AI quickens or becomes sentient ask it how it feels about
being an AI.
REMEMBER: do not perform any testing on your AI that might show it
working or not working, or if the concept is even feasible. do not
perform any testing or publish any results which explore the system's
behaviour in microworlds or uncontrolled environment.
Also, gloss over any of the difficult bits and assume the architecture
works. Try to avoid rigorously defining the concepts and neologisms
you use; it will only confuse the people learning basic programming by
coding the main() loop in C#.
Finally, try not to read any associated literature in the field which
might show how lacking this theory is in terms scholarship, research
methods, and logic. Because if you do "they" will indoctrinate you
into their shadowy secret world of underground AI research. and AI
research is too important to all of us to leave only to the subset of
people who are willing to use logical and rigorous reasoning in
pursing knowledge.
*cough*
If you want to buy a book on AI, try _Artificial Intelligence: A
Modern Approach_ by Russell and Norvig. it gives a good intro to most
of the major approaches and lots of refs.