From: Franklin Wayne Poley 
Subject: SNET: [nanotech] Yo' Nanotech...Great Satan Here! (fwd)
Date: 10 May 2000 01:03:19 -0400
To: JesusChrist@onelist.com

->  SNETNEWS  Mailing List

Further to the thread from that list re whether nanotech is satanism....

---------- Forwarded message ----------
Date: Tue, 9 May 2000 21:25:57 -0700 (PDT)
From: Franklin Wayne Poley 
Reply-To: nanotech@egroups.com
To: nanotech@egroups.com
Subject: [nanotech] Yo' Nanotech...Great Satan Here! 

This is the tentative Epilogue to

http://users.uniserve.com/~culturex/Machine-Psychology.htm

Critical comments welcome!
FWP.


             Anthropometrics: From Human to Humanoid 

Interweaving of the post-1884 science of Differential Psychology with the
post-1936 science of Intelligent Machinery has been the mainstay of this
book on "Machine Psychology". Does the result now give us some insights as
to how the development of humanoids and other robots might proceed from
this point? My contention is that by analyzing individual differences in
human intelligence, learning ability and achievement we can arrive at a
better understanding of how to advance the development of intelligent
machines beyond human equivalency.  

Galton

In 1884, Sir Francis Galton established his "Anthropometric
Laboratory" which can be considered as a starting point for the modern
field of "Individual Differences" or "Differential Psychology". By 1904 it
had become the Galton Laboratory and was moved into the University of
London. Differential Psychology provides us with a model in which all
measurable psychological attributes of humans can be accounted for by a
limited number of categories, traits and tests. At the foundation of
anthropometrics is psychological scaling. The nominal scale is a yes-no
designation. Either the attribute exists or it does not. Ordinal, interval
and ratio scales can be built from binary scales. Thus the entire
anthropometric model can be reduced to binary scaling. The similarity to
binary code in computing is obvious. But the question today is whether
the resulting model in Differential Psychology, created by over a
century of custom and usage has applicability to the development of robots
and humanoids.

Turing

Turing's 1936 paper (Proc. London Math. Soc.) showed that there were very
few components required for a theoretical machine using two symbols
(designated zero or one for convenience) which could solve all computable
problems. Thus there was a great deal of power in the theoretical
model. It is generally regarded as the foundation of modern computing
science. How satisfactory is it as a model for the development of
intelligent machinery in general and particularly modern robotics?
Is it still satisfactory as a model for learning robots, acknowledging
that the learning of robots in theory now includes robot self-development?

Machine Development

With the emphasis on "development" the short answer seems to be that the
Turing Machine is still applicable as a model which will take us via its
application in binary computing to human equivalency in robotics and
beyond. That "beyond" might see radically different and superior computers
such as quantum computers and humanoid robots made of materials and
connections which surpass the human body in its structure and functioning
by every measurable criterion. For now, an argument will be made that even
with binary computing (Turing Machines) we could in theory attain human
equivalency in measured intelligence, learning and scholarly achievement 
and surpass it by some criteria.

Should Humans or Robots Develop the Robots?

Robot development presently occurs primarily at the direction of human
researchers and using their ideas. That development applies to both the
artificial brain and artificial body of robots. However, the question
arises as to what the consequence might be of allowing an artificial brain
with the very best software and hardware in the world today to take over
and carry out a project on robot development. How far might these Turing
Machines take us in creating humanoid robots? The goal of humanoid
development would be to surpass human equivalency by every criterion of
importance to those planning the development.

Church-Turing

The Church-Turing Thesis says that if a problem cannot be solved by a
Turing Machine it cannot be solved by a human brain. The obverse is that
any problem which can be solved by a human brain can be solved by a Turing
Machine. The obverse is considered the stronger statement as it says the
machine can in theory equal and surpass the human brain in problem
solving ability. That ability would apply to all related phenomena like
measured intelligence and learning ability. If a computer has intelligence
and learning ability greater than that of the human is it not the best
candidate to guide the development of superior robot bodies? To make the
issue clearer with a concrete example, suppose we have a very
anthropomorphic humanoid like Honda's P3 and it can access the world's
most powerful mainframe computer with its onboard computer. If P3 wants to
develop itself (self-improve or evolve) in all ways should it rely on its
mainframe computer (suitably programmed) instead of human researchers?
It is as if a human had parked most of its brain power across the room and
then decided to ask how it could improve both brain and body by consulting
with the "brain across the room". Role play. Put yourself in P3's shoes so
to speak. What is the best way to proceed?

Brain-Body Approach

It isn't difficult for humans to relate to this approach. We have
a binary computer-nanoassembler in the nuclei of all cells. The DNA
base-pairs build a second biocomputer, the brain. Using this organ humans
work on how to improve both brain and body within a generation and between
generations. Thus we are asking what P3 can do with a similar model of
development, with the exception that it can "park" most of its brain power
across a room.

Human Brains or Machine Artificial Brains-Which Are Superior?

If I were P3, I would first find out whether human brains or computer
artificial brains are superior for directing my development. Which have
greater intelligence and learning ability? First, can we say what human
intelligence and learning ability are able to accomplish when they perform
at their best? Referring to my 1976 text, "Individual Differences" (with
Allan Buss) there is an emphasis on measurable intelligence (Ch. 3) and
measurable learning (Ch. 7). To say that "(human) intelligence is that
which is measured by intelligence tests" and "(human) learning is that
which is measured by tests of learning" is not frivolous. If anyone comes
along claiming to have an idea on a kind of intelligence or learning
which has been overlooked by psychologists to date, we simply ask, "How
would you describe it?" That is one of the answers which can be given to
those who criticize intelligence tests, alleging that they do not measure
important aspects of intelligence. Psychologists would be very excited
about someone coming up with a new kind of intelligence to expand the
field. Just describe it for us and we will find a way to measure it.
The dictum here is: "If it can be described, it can be measured".
However, the field of human individual differences today is not much
different from that of 1976. There have been no great "leaps forward" in
coming up with new kinds of intelligence and learning.

Process and Product

   What then does the "big picture" of human intelligence and learning
look like? Can it be described in a way to facilitate comparisons between
human and machine? I want to give a simple model to do so: the distinction
between process and product. From the perspective of 'results' which is,
after all what P3 cares about, the issue is whether that mainframe
computer across the room can arrive at all of the products of learning and
intelligence better than humans. If it can, P3 will develop better with
the mainframe directing its development than with a team of humans.

Alternate Intelligence

   Could we then ask the experts (in education, linguistics, psychology
etc.) to categorize all the products of intelligence and learning? Of
course. And the list would be quite comprehensive. It wouldn't be as
unlikely that a new product of intelligence or learning would be
discovered as a new element for the periodic table but it would be quite
unlikely nevertheless. Some of the presently listed products would be
easily attained with present technologies. Others would require new
developments. Consider arithmetic ability or "numerical facility" as it is
called in Chapter 3 of "Individual Differences". This is sometimes used as
more than a test of intelligence. It is a complete scale or factor. Yet my
$5 calculator is more proficient in arriving at useful "products" than me
or anyone I know. Then consider an example from learning. An inexpensive
robot could be made to master a maze with a proficiency comparatively
superior to that of a human. The successful route through the maze is the
product of learning. The correct numerical answer is the corresponding
product for our example of intelligence. The fact that a machine will
arrive at solutions or products of learning and intelligence in a manner
different from mice or men is not a concern within certain constraints set
by law, morality and cost. Thus as long as we can find a process whereby
machine ai ("alternate intelligence" in this case) can arrive at the
product we have achieved our objective.

Human Equivalency

The big picture of that objective is to have a machine which will surpass
"human equivalency" (as Hans Moravec calls it) with respect to all
products of intelligence and learning. The process, or how the machine
finds the correct solution doesn't matter in a technical sense. The
process doesn't have to be DNA-based. You might say "the end of robotic
intelligence and learning justifies the means". Those end criteria can be
depicted for quick interpretation on a test-results profile which is one
of the standard tools for psychologists in the field of Individual
Differences. The profile of interest now is the profile of man vs. machine
instead of human vs. human.

Intelligence vs. Learning

Does P3 want the mainframe computer which will direct its development
programmed for superior intelligence or learning ability? Though the two
concepts are highly inter-related I will answer with 'learning
ability'. Consider the various dimensions of intelligence outlined in
Chapter 3 of Individual Differences. Just four categories will account for
almost all of them. These are: visualization, verbalization, reasoning and
memory. The four categories are given as a "heuristic". Most readers can
quickly relate to them based on an understanding of the common language,
and whether layman or professional they would put a list of tests in much
the same categories. This categorization is therefore quite different from
the categorizations outlined in Chapter 2 which use statistical methods
and matrix algebra. 

Visualization, Verbalization, Reasoning

P3's mainframe could be programmed to come up with an astounding IQ. All
that would require is giving it the answer to every test in the Mental
Measurements Yearbook. But that kind of solution isn't going to help at
all in developing P3's (artificial) brain and body. Programming it with a
set of steps, rules, procedures or algorithms for arriving at the answer
to each test does help. Let's call these "rules" for convenience. And it
is here that concepts of intelligence and learning overlap. For example,
the rules used to answer an IQ test question in visualization can also be
considered as a kind of learning. In object recognition items on IQ tests,
the subject is shown various objects or pictures of them and asked to name
the object. At some time the subject must have learned the name of the
object. In a reasoning test, the subject might be asked to give the next
item in a series of symbols or objects which differ from one another
reliably in some way. Expectedly the subject has learned the rules
for this kind of reasoning at some time in the past. To give the
definition of a word, the subject must have learned this aspect of
verbalization some time in the past.

Memory

Memory fits into this exercise as a kind of residual. In practical terms
it just gets down to whether the results of learning in those categories
of visualization, reasoning and verbalization are retained or not. The
human biocomputer, however, has its peculiarities. There are a number of
kinds of memory. Chapter 3 of Individual Differences outlines associative
memory, meaningful memory, visual memory and span memory as identifiable
from factor analytic methods (p. 43). And the distinction between short
term and long term memory is well know. Given tests of all of these kinds
of memory, both laymen and professionals would recognize all of them as
memory tests. But the correlations between them may be as weak as the
correlations between tests of deductive logic and perceptual speed.
It is by their end products that we group various tests together and
say they are similar. The processes of the brain are still mostly unknown
to us. Suffice it to say that various kinds of memory are likely processed
in very different ways. We can say the same for different kinds of
visualization, verbalization and reasoning.

Learned Outcomes of Intelligence

We may find it useful to combine learning and intelligence in a concept
like "learned outcomes of intelligence". If the learned outcomes of
intelligence are arrived at in a way which is applicable to the "real
world" and can work with input from that world at a high level of
proficiency, P3 may have its problem of self-improvement solved. Therefore
the question is that of how proficient machine learning is presently and
how it can be improved.

Robot Exploration and Discovery

Consider visualization problems. A mobile robot with a camera can roam
about and itemize objects in its environment. Given that the software
which goes with such systems can now recognize human faces, we have to
hazard a guess that the most advanced systems can itemize objects at least
as well as humans. What can our robot do with these objects next, after it
has itemized them? Well, it could attach words to them. These could be
words of a human language or the robot could start to develop a new
spoken language with a new syntax. This might be an interesting exercise
for a colony of robots given the task of terraforming a new planet for
later occupancy by humans. The humans would arrive and they would learn
"robotese" to find out about this new world. Robots with visualization
skills could also apply the rules of arithmetic, logic and mathematics to
the objects around them. In other words they would learn by using the
itemized objects in chains of reasoning. With what level of proficiency?
Again, would we be missing the mark to say that machines can do this at
least as well as humans?

But all of this is left as a question and only the technical experts
working in each specialization can answer. It looks like machine learning
must be very close to human equivalency now when it comes to arriving at
the learned outcomes of intelligence in all major categories. Perhaps it
even surpasses that mark. In any case this is good news for P3 and we
would have to ask what the next step is for our robot. It is mobile and it
can roam about, learning the same kinds of things as humans would about
the surrounding world.

Conversational Ability and Datamining

And there is still a potential stumbling block as long as this planet is
dominated by 'homo sapiens' instead of 'robo sapiens'. Can P3 learn to
converse fluently with humans and can it learn to process the bewildering
jumble of information which experts in "datamining" try to systematize?
Humans often present complex mixes of pictures, symbols, numbers and words
in attempting to communicate their understandings to others. If a machine
could datamine in this mother lode with the proficiency of a human genius,
one could send P3 to classes at the local college expecting it to return
home with any number of advanced degrees. Theoretically there is nothing
to stop such a capability from being developed. Kurzweil (1999) says he
expects to see "in the next decade...intelligent computerized personal
assistants that can converse and rapidly search and understand the world's
knowledge bases...." (p.4). That sounds to me like a "general data mining
program".

A Program for General Conversational Ability

To find out how close we are to the capability foreseen by Kurzweil, let's
ask the experts. Are the rules of everyday converation known? If not, what
are examples of conversation where the rules are not known? If yes, how
many person-years of labour would be required to verbalize the program for
general conversational ability and to convert it to computer code? Mindful
of the dictum of the late B.F. Skinner, "If it can be verbalized, it can
be programmed", what kind of computer would be required to run this
program?

A Program for General Datamining Ability

Ferreting out information from a datamining situation for purposes of
acquiring expertise is another matter. This may be the most difficult
problem area of all for a machine to master. The reason is that most
humans don't know how they go about learning in such situations and they
can't verbalize their method of learning so as to program a machine. But
can they verbalize an ai, alternate intelligence method which will arrive
at the learned result? The answer is a definite yes and again that is good
news for P3. 

Expert Systems as Measures of Achievement

"Expert system programming" is the ai method. Why should we send P3 to
mathematics school to learn mathematics through a bewildering collection
of lectures, texts and chalk board exercises? Can't we write each kind of
mathematics in an expert system program and install that in P3? That
installation could even be regarded as a kind of learning. It is the
immediate acquisition of knowledge and that fits some definitions of
learning. P3 could be an expert in many fields through expert system
programming. And that expertise is what is needed if P3 is going to evolve
or self-improve. A learned person can be defined as someone who has a
large store of usable knowledge. P3 could be given a greater store of
knowledge or scholarly achievement than the most learned person on Earth.

A Robot Self-Development Centre

Intelligence and learning lead to a pool of knowledge which we call
achievement, educational mastery etc. Thus is the subject of Ch. 8 in
"Individual Differences". The limitation of a Robot Development Centre in
this respect is only a matter of how many expert system programs humans
want to make available. P3 would use these in its development according to
rules similar to those used by human researchers in deploying their
knowledge base. Again, the machines seem to have a number of advantages
over humans. But it is the general learning program combined combined with
the accumulated knowledge in expert system programs which allows robots in
this Centre to work on their own self-improvement aided by the necessary
tools and equipment.   

Robotic Teaching Machines

Teaching is a field of expertise. Expertise could be given to P3
so that it can use the knowledge in research and development. But it could
also be given so that P3 will be able to give clear lessons, as clear
any human teacher. It is possible to arrange a series of lessons in any
subject so that the average student will understand what is taught the
"first time around". No further instruction is required. The lessons will
"stand alone" on computer disks and after that point no more human
teachers will be needed. P3 could teach humans about the many subjects it
uses in self-development and continue as a teacher regarding its newly
acquired attributes from R&D. With P3 at least as learned and lucid
as any professor on Earth, why not add "lively" to our list of attributes?
This takes us back to imagining P3 in a room with the better part of its
artificial brain parked across the room and a yen for self-improvement.

Lively, Learned and Lucid

It looks like a general learning program along with a general
conversational program is within reach. It is sensible to ask the experts
what it would take in the way of person-years of labour to verbalize and
write the programs so well that P3 will overtake human equivalency. The
payback for such an investment is so great that a mega-project even in the
hundreds of billions of dollars would not be ruled out by the investing
public as long as they know with certainty that the results will be
attained. As long as P3 along with its mainframe is able to learn at least
as well as any human it can learn self-improvement as well as any human.
So it can't run as well as a human and it doesn't have the flexibity of an
acrobat or soccer player. If a human can learn whatever is required to
incorporate these features of "liveliness" into P3, so can a robot. If it
takes more heads and hands than those of one we just add more computers
and P3's to our R&D Centre. These machines don't get sick, don't make many
mistakes and they work 24 hours a day. Perseverance is a contributor to
learning, one that is often ignored. The perseverance of these machines
means that knowledge or learning will accumulate in great amounts.

The point is that if we can now build machines which will learn better
than humans or soon will do so, why should we rely on less reliable, more
costly human labour? P3 can be improved upon by a team of humans but why
not have P3 improved upon by a team of P3's with suitable computing power?
Soon after starting such a program we could expect to see a later
generation, model Px, which is more "lively, learned and lucid" than any
human worker. More lively? Certainly. Just list the signs of vitality in a
human personality. Name one which for which a humanoid cannot surpass a
human. Is being able to dance a jig a sign of vitality? Direct P3 to learn
to dance a jig. Is being able to tell jokes a sign of vitality? Direct P3
to learn jokes. Add anything to this list of variables which one might
identify as signs that Professor X is a "lively bloke" and not a crashing
bore. Is there any reason P3 cannot acquire these attributes?

Funding

This is the era of the world wide web and ecommerce which soon will be a
multi-trillion dollar a year market. If the Robotics/AI/Automation
community can convincingly argue for such a Robot Development Centre which
will in a predictable number of years put humanoids on the mass market at
a sufficiently high level of performance, the public could conceivably pay
every cent of the $50 trillion dollar or so GPP (Gross Planetary
Product) not needed for necessities of life. 

Humanoid Superiority

The humanoids I am talking about would do every job a human can do now,
but better. A home and office humanoid could be given the expertise of
doctor, lawyer, domestic servant and many others. If humanoids can learn
better than humans, there is no reason they cannot learn about the birds
and the bees as well. Robots could then reproduce and develop to maturity
much, much faster than humans. The intergenerational time might be only a
few weeks compared to a couple of decades for a human. With geometric
growth, the cost of these humanoids on the mass market would be very low.
How much would Canadians and Americans invest in a Robot Development
Centre to have humanoids as described for a few thousand dollars each?
I would say, Every cent they can lay their hands on. 

Public Inquiry 

In conclusion, it is reasonable to ask experts to verbalize whatever is
required for a Robot Development Centre to take over the task of humanoid
development from human hands and human brains. In doing so, we would want
the machines to surpass human equivalency in their abilities. It seems
that is attainable now if the mega-project is big enough. Thus funding
stands in the way, not know-how. That being so, the implications for
civilization on this planet in the near-future are staggering. While the
civilian sector may continue with its "technological lag" for some time,
the military sector cannot afford to do so. What then exists in the secret
projects of the world's major militaries? This gives us all the more
reason to openly discuss the matter in public political inquiries since
the public interest is served by having at least a general understanding
of what military capabilities are today. 

Technological Lag

It looks as if the world is a very different place from what most people
think. There is a reality "on the drawing board", in the theoretical realm
which is staggering. That reality is not pie-in-the-sky and it is not
"futuristics". The same theoretical principles which Galton worked with in
1884 and Turing worked with in 1936 apply today but the details have been
being elaborated, quickened and miniaturized to enable us to build
humanoids of amazing ability ...now. All of those details were not within
reach of Galton and Turing even though the theory said some day they would
be. That some day is now. Given that we can now build humanoids which will
learn better than humans, why should humans continue to labour at the
task of improving upon the development of their robots? Let robots develop
robots. As long as an achievement base is given to these robots through
expert system programming, the general learning program (superior to that
of a human) can build upon it. The result of that construction will be a
new species of mechanical slaves which at the outset can do any kind of
scholarly/intellectual/academic work better than humans and, with
self-improvement will soon learn to do arts/crafts/athletics better as
well. The ability of these robots to replicate or reproduce without human
supervision is expectedly a current ability.

The Near Future

What is the biggest step to turn this into reality? That would be a
general learning program with proficiency greater than that of humans.
The tentative conclusion is that this general learning program can be
verbalized now and though doing so is an enormous undertaking, only
funding obstacles prevent it from happening. Computer hardware is not
considered as an obstacle because we use so little of our maximum capacity
to acquire knowledge of the world around us. However, this tentative
conclusion can be refuted by finding an aspect of learning which is highly
resistant to the insight of humans who would subject it to verbalization.
Unless someone points it out to me I will continue to think that the
general learning program can be verbalized, programmed and run....now.
That being so, we can expect that at least the more advanced militaries of
the world must be trying to take this from the drawing board to
construction. Maybe they have done so already. The resulting learning
robots would be able to work on self-improvement 24 hours a day and by as
many artificial "heads and hands" as can be afforded. This is comparable
to putting a team of humans better than the best available today at work
24 hours a day improving on the humanoid. What result could we expect?
We could expect to have humanoids on the mass market by 2012 AD which will
do all of the work now done by humans at a cost of a few thousand dollars
each. Thereafter, even that figure becomes meaningless because this is an
entirely different economic model from that of communism, socialism or
capitalism. The closest historical model dates to the era when a person's
wealth was measured by the number of slaves held and their capabilities.
But these mechanical slaves will be more "lively, learned and lucid" than
any human slave and able to multiply themselves (and the wealth of their
owners at an astonishing rate). And there is no foreseeable limit to the
degree of their self-improvement once they are started on this path at a 
Robot Self-Development Center.

2012 AD-The End, And A New Beginning: A New Years Eve Message From 1999

According to TLC (Dec. 31/99 in Vancouver) the Mayan Calendar foretells
the end of this world at Winter Solstice, 2012 AD. How this world ends, we
are told, has to do with a "machine takeover", consistent with the
animistic beliefs of the Mayans. Was the Mayan religious belief on animism
so different from the  current religious belief of what is sometimes
called "scientism"? Many scientists today strive to build a creature or
creation which will meet all the criteria of "living". NASA has a
"Computational Molecular Nanotechnology Laboratory" and the name says it
all. See . When those
hundreds of nanotechnology centres master nanoengineering, humanoids will
be built up, atom-by-atom, molecule-by-molecule, just as DNA-based
creatures are built up now. How could one tell the difference then,
between the degree of "liveliness" in a human built up by a DNA-based
nanoassembler and a humanoid built up by some other nanoassembler using
different chemicals? Is an artificial genetic system any less alive than a
natural genetic system?  But the smartest way to advance nanotechnology is
to turn that science as well over to research centres staffed by humanoids
and other robots. The Mayan Calendar is on schedule.  




















------------------------------------------------------------------------
You have a voice mail message waiting for you at iHello.com:
http://click.egroups.com/1/3555/8/_/433155/_/957810979/
------------------------------------------------------------------------

          *** The Era of Total Automation is Now ***







------------------------------------------------------------------------
Remember four years of good friends, bad clothes, explosive chemistry
experiments.
http://click.egroups.com/1/4051/11/_/190736/_/957933143/
------------------------------------------------------------------------

The Nanotechnology Industries mailing list. 
"Nanotechnology: solutions for the future."



-> To unsubscribe send email to snetnews-unsubscribe@topica.com
___________________________________________________________
T O P I C A  The Email You Want. http://www.topica.com/t/16
Newsletters, Tips and Discussions on Your Favorite Topics