Date: 12-21-90 (09:10) Number: 247 of 262 (Echo)
To: ALL Refer#: NONE
From: TIM RUE Read: (N/A)
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
Vacation Time. I won't be around for about a week, so any postings to
me will have a delayed responce. Merry Xmas to all.
Food for thought:
Consider the process of evolution from the big bang to the
consciousness of man. From all the substance produced from the big
bang, how various elements combined to produce and/or eliminate other
combinations, and how the resulting combination of elements interacted
to further produce and/or eliminate combinations. Bring resolution
into the elements and combinations of elements and resursion. Note
that out of all possibilities that all possibilities don't happen, but
rather only those that the previous interactions allow. From here
consider the path of human consciousness and how it may be represented
by an analogy of zooming into a fratical. Now, what enviromental
elements interacted that caused man to become conscious, beyond all
other known cognitive creatures which enable man to develope
technology to what it is today?
Going one more step, what elements are involved in developing/evolving
AI?
Happy Holidays. Tim
Date: 12-23-90 (03:42) Number: 253 of 262 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: TOOL Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Ted: Fess up Tim! You are someone's AI project! Is this not
> so?
>Tim: ??? Not sure what you mean. :-)
>Tim: The human brain is the ultimate unifier!
Hooray! Another Prolog user!
>Tim: I read about AI developer and realize they're doing the
> same thing (not turning the screw).
I look at the example of the telephone as positive! We are all groping
in the dark for intelligence. I am convinced that we have hardware
which is on the order of complexity of the brain. We do not have a
clear direction toward intelligence.
My work at the University of Kansas AI lab focuses on self-organization
of information. Our constants, if you will, for the Knowledge
Navigational Mapping/Expression Tool are marginalization and projection
of sets. I like to think of them as reduction and abstraction.
>Tim: There are some things I don't post (least not yet, first
> things first, one step at a time, etc..).
I can understand that. You approach AI by looking at yourself
objectively, but realize that in fact self-observation is subjetive. I
would imagine that to some of the lurkers in the conference that your
statements border more on fancy than inquiry. I have found your posts
to be interesting and worth the effort, so if you are willing to take
steps ahead, I'm with you.
>Tim: (What I call the forward loop of technology, ...
What about the forward loop of actually creating an intelligence? I
really can't imagine designing an intelligence that simply is. I
believe that any approach will involve a learning and growing process.
I don't see any reason why that proccess should stop once a human level
of intelligence is acheived. (We have a spark of intelligence, but all
in all mankind is stil fairly dim...) One of the many possibilities for
a higher intelligence would be genetics. We design a superior computer
which designs a superior human which designs a ....
Put that in your forward loop and smoke it!!
---
þ DeLuxeýáa #681s ú Time is an illusion. Lunchtime doubly so!
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 12-23-90 (07:36) Number: 255 of 262 (Echo)
To: ALL Refer#: NONE
From: ALFONSO HERMIDA Read: (N/A)
Subj: THOUGHTS ON AI Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
I think that AI has been reduced, in a programming sense, to a Tree
problem. That is, finding data which has been arranged in graph form.
What should AI really be? Humans, for one, think 24 hours a day and are
managing info/data in order to reach certain goals and to fulfill
"nature" calls. Most AI systems right now pretend to find solutions to
problem in maybe a few minutes or days, while humans have taken
hundreds of years to reach certain level of knowledge. I do believe
that AI can work...It's just a matter of redefining the goals for it.
---
þ ILink þ The Science Lab BBS þ Baltimore, MD þ 301-466-0949
Date: 12-24-90 (12:58) Number: 256 of 262 (Echo)
To: FREY WAID Refer#: NONE
From: WILLIAM WRIGHT Read: NO
Subj: NO LIMITS TO LIMITATIONS. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
FW> Humans can pick up an apple. A computer with a robotic
arm can do the same. Take this analogy to its logical
conclusion and a computer with certain physical
additions can do the same things as humans.
Possible glitch:
If a particular "human" activity requires distributed
processing, and if _every_ cell in the human body
contributes directly or indirectly, then the only "machine"
which could emulate a human would be another human.
There's lots of evidence, for example, that human moral
judgements involve the state of our digestive and
reproductive systems as well as the connections in our
brain tissue, and so on.
As I think I 've mentioned once before, our language has
phrases such as "I see what you mean" and "I feel angry"
and "I don't like the sound of this" and "My gut feeling
is", etc etc.
All of these phrases emphasize the distributed nature of
our behavior.
---
þ SLMR 1.0 þ Return with us to those thrilling gadgets of yesteryear.
þ ILink þ Console Command HQ þ Santa Barbara, CA þ 805-683-0499
Date: 12-24-90 (12:58) Number: 257 of 262 (Echo)
To: ALFONSO HERMIDA Refer#: NONE
From: WILLIAM WRIGHT Read: NO
Subj: THOUGHTS ON AI Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
AH> I think that AI has been reduced, in a programming
sense, to a Tree problem. That is, finding data which
has been arranged in graph form.
Not completely true. Have you studied any of the neural
net, and/or parallel processor, concepts which are being
developed now? You certainly can't "graph" them or reduce
them to a tree.
They are non-linear by definition (where "linear" does not
refer to the exponents in an equation), and usually
recursive.
---
þ SLMR 1.0 þ Just because I'm typing, don't take me literally.
þ ILink þ Console Command HQ þ Santa Barbara, CA þ 805-683-0499 [R106R]
Date: 12-27-90 (03:39) Number: 259 of 262 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim: Consider the process of evolution from the big bang to the
> consciousness of man.
> Note that out of all possibilities that all possibilities don't
> happen, but rather only those that the previous interactions
> allow.
No limitations! We perceive sequentially, but that is merely our
crutch! What happened before the Big Bang? Is anything orthogonal to
reality? Film at 11:00.
---
þ DeLuxeýáa #681s ú We need a Bill Of Electronic Rights!!
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 12-28-90 (15:01) Number: 261 of 262 (Echo)
To: TEAGUE SHERIDAN Refer#: NONE
From: FREY WAID Read: NO
Subj: What are limitations? Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
TS> Do you think that a computer could actually WANT to pick up say, an
TS>apple?
The wanting to pick up an apple comes from evolutionary/biological needs
of the uniquely human condition, of course. This was just an example.
Perhaps, a computer would WANT to recharge its power cells (however,
this is done). Obviously, if a functionary has no need to do
something, then the functionary will not do it. (Could the WANT be to
give the apple to the human beside it?)
TS>FW>Even current technology would make a case that a computer can generate.
TS>FW>For example, the generation of random numbers.
TS>
TS> Do you program? Computers can't truly generate random numbers, but have
TS>complex calculations to do so. They may seem random if they use the clock
TS>as a seed, but then if you use the computer the same exact time each day,
TS>the numbers would be the same each day.
What is that we do to generate "truly" random numbers? Couldn't it
be boiled down into a set of complex calculations, also? But, let's
not quibble over whether or not a computer or human can "truly"
generate. I don't know what "truly" means. The fact is, is that a
computer can GENERATE. And that's the problem with the given
definition of a cognitive process if the definer is trying to show how a
computer does not fit in.
TS> Yes, but outside of neural nets, what Tim said is true.
Well, your talkin' to a guy who thinks that neural networking is the
only technology that's going to make any serious headway into AI.
TS> But still, a computer can't actually think that something is beautiful,
TS>except by patterns or a set of rules like if nose size < 300 then person =
TS>cute. In those cases it could only come to a conclusion that it is
TS>beautiful, no?
First, the point was to say that we, as humans, anthropomorphize our
subjects (make them man-like). Nature does not think itself beautiful.
We think nature is beautiful. Nature does not care. Nature is not
human. The tendency to interpret our subjects' actions in a human way
is perfectly natural (pardon the pun). But, we must be cautious when
we define "thought" and "intelligence" in a human way. For these are
supposed to be universally applicable concepts.
Second, how do we, as humans, define anything except by patterns or a
set of rules? Sometimes our criteria are not innumerated at the
conscious level, but they are met in some way.
TS> We must define "consciousness" some way, and if the only way is
TS>subjectively, then that it must be until we find a better way.
Fine....Then, the only way you may appropriately use your definition is
on yourself.
TS>Intelligence is defined as thought or reason;
Here we go....Webster again.
TS>so we go to thought: the power to reason or imagine, consideration,
TS>concern (not at all what a computer could do)
Again anthropomorphism. Okay, by this definition an alien who came and
landed on the planet in a spaceship would probably not have thought.
What is concern or consideration, but human-linked emotive states to
reason? And imagination, well, that word is thrown around alot. Can a
computer imagine now? Maybe in a limited sense. Will it imagine later
on? I imagine so....
TS>and reason: to think logically (what you are saying). So basically it comes
TS>down to a choice, and what YOU believe, as there is no way to really tell.
I think you've hit the crux of the problem. Indeed, there is no way to
"really" tell. Turing has pointed this out. If someone is able to
function well in our society/environment, then this someone is
intelligent. That's the only criteria most people need to determine
whether that someone is intelligent. Why is it any different for
anything else?
You are throwing around words like "truly generate" or "actually think."
What is the difference between thought and actual thought? Would true,
actual thought be better? I guess real, true, actual thought would be
even better. Anyway....
Frey
---
þ EZ 1.33 þ Stop that...stop that! It's silly.
þ ILink: -=The MBT=- þ Blacksburg, Va þ (703)552-8767 HST/DS
Date: 01-04-91 (09:23) Number: 270 of 272 (Echo)
To: WILLIAM WRIGHT Refer#: 248
From: TIM RUE Read: NO
Subj: AI Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
TR> ...I mentioned internal representation of knowledge
[by means of] the Mandelbrot fractal ... allowing an
infinite number of points to be used for internal
representation ... calculate various levels of
resolution ... application of resolution in a neural
net ... identify "rules" [as constants] ...
WW>It's an interesting thought. There are some new chips &
software which (somehow) compress images by analyzing them
into fractals. I haven't the faintest idea how they do
the analysis. Does anybody know more about this ??
Mandelbrot's classic book about fractals states that (as of
1983) nobody has proven or disproven that neural
connections follow a fractal geometry. Mandelbrot's Set
is only one of many fractals. And strictly speaking,
there are other types of recursion besides fractals.
What you're really talking about is using a single recursive
rule as a descriptor for a dataset (or subset).
=======================================================
My use of the Mandelbrot fractal is that of an analogy (although it is
possible to actually use it as a maping tool). I'm well aware of other
fractal geometry. The reason I chose the Mandelbrot is 1) It's well
known, 2) It has an unlimited depth, and shows the need for resolution
to make possible any reference point, 3) It has three deminsions (not
counting resolution), 4) Every point is unique, allowing one to relate
an infinite number of things to unique points. 5) The fractal geometry
is preceivable through sight (using visual to help me communicate
something that is not visual).
*** The concept I was communicating has nothing directly to do with
images or compression of images. But rather the concept of a non-
visual tool for Knowledge Navigational Mapping and internal
representation of knowledge. My reference to neural nets was inqury as
to weither or not resolution in neural nets is possible and if so
would constants be exposed (I was asking a question). *** I WOULD NOT
say I'm talking about using a single recursive rule as a descriptor
for a dataset (or subset) although a single recursive rule might be
used to access a dataset (or subset/finer resolution) within a system
of infinite (possible) locations/rules and combinations of.
In reguards to other types of recursion besides fractals: This is
all well and fine, perhaps there is something that would work
better than the Mandelbrot fractal, however, when you want to
accomplish something you should set out a plan of action because
having a plan of action (good, fair, or poor) is better than none.
In this case it's in having a tool for internal representation and
knowledge navigational mapping. Having a tool is first priority,
second might be determining what works best. But how can you
determine what works best until you get experience with a tool(s)?
William, And any of you all following my postings, may want to capture
them for later reference. We humans tend to forget or at least get
distracted/ disconnected from a line of thought. I myself have done
this as a result of my vacation and have taken a day or so to allow
myself to regain my focus. At times it is good to get away from a
subject matter and let things settle, then look at it again in a
fresh mind, perhaps see something you didn't see before.
Date: 01-04-91 (09:24) Number: 271 of 272 (Echo)
To: FREY WAID Refer#: 251
From: TIM RUE Read: NO
Subj: NO LIMITS TO LIMITATIONS. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
FW> Sorry to step in but I read this and had to take issue:
.
.
* refer to message form Frey *
.
.
FW> Please reply,
Frey
*****************************
Frey,
All that man creates, he creates as an extension of man.
Allowing man to go beyond his physical and mental limitations. Man
creates such things to serve man with some useful function(s). Even in
creating something just for the satisfaction of doing it serves a
purpose.
Anthropomorphizing is very important for humans, if man did not
do this he would not exist. Perhaps it is this anthropomorphizing that
is giving man the drive to create a machine that will mimic man? Silly
perhaps to create such a machine which has such limitations. Consider
this statement: "Computers can calculate complex equasions very fast,
equasions which would take man lifetimes to calculate without his
invention/creation of computers". Understand the inherent priority.
I believe neural net have an important usefulness (perhaps
input/output filters), but why not have a system which contains the
ability that neural nets give along with the calculation ability? In
doing so such a system would not have something that a human has, the
limitation of calculating ability.
Sure we can build/create a machines which can interact with it's
environment, and we can call it cognition. But it's really just the
appearance of cognition. Consider a person in a coma, they appear to
have no cognition, yet for those who come out of a comma, there must
have been some cognition (something to use as a reference/feedback in
direction to go). Without cognition a human will cease to exist.
Without cognition a machine simply continues to run via how it is
designed. There is no internal will/motivation to do so. To give a
machine such will/motivation requires man to give it total
independance from man, and in effect risk losing it's usefulness to
man. Also in doing such would also require the machine to have the
ability of self-creation or some form of self preservation (a matter
of independance) and certainly anthropomorphizing (cognition -
relating enviroment to oneself).
Emotion are a result of ones internal value system applied to
information (even incomplete information). Emotions are not to be
judged or used to judge a person or their values. Actions however can
be judged, reguardless of the logical or emotional base of the action
(thought does not call for action, action is a choice). Such a
statement as "nature is beautiful" is emotion based and I'm not sure
if the word anthropomotphize would be correct to use here.
The complex cognitive process of consciousness results in or allows
abstraction (in the purest sence). Such pure abstraction cannot exist
without effort and disconnection form physical reality. In turn such
abstractions can be put into application in reality, such as taxes
(which have no real connection to reality). Some confussion can
result in viewing the application in reality, but clarity can be
found in recognizing the requirements needed for the application to
work (with taxes it's a matter of human decission and sometimes
application of force or fear). However, taxes are an application of a
pure abstraction. So in a very strong way, consciousness is not a
matter of interacting with ones environment but rather creating ones
environment. Yet, without cognition there can be no awareness of what
to create and from what. Taxes are one of the creations of conscious
man.
No reason to say your sorry for stepping in. That's what this
conference is for (interaction). How else can we improve ourselves and
our thoughts. I have found you message quite interesting and certainly
worth responding to. Thanks!
Date: 01-04-91 (09:26) Number: 272 of 272 (Echo)
To: TED KIHM Refer#: 253
From: TIM RUE Read: NO
Subj: TOOL Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
Ted> Hooray! Another Prolog user!
Sorry to disappoint you but I'm not a prolog user. Actually such a
responce does not supprise me in the least. When I became interested
in AI, I did alot of pattern matching in an effort to find something
that would do what I wanted. I still keep an eye out but have not
found it ( although I'm sure there are many languages that can
accomplish what I want, none allow the simplicity and versatility). At
times I even thought there must be some inside pact that has kept such
simplicity from being put into application. At any rate I had thought
several times I had found it, but on closer inspection I found the
opposite. So I went to school and learned "C" so as I may create the
gears and bearings I want. These "gears and bearings" are very capable
of doing all that the various and numerious languages can do and
perhaps much more (including perhaps some things we would consider
rather foolish). Note: Man has the ability to take something very
simple and make it so complex that it is almost impossible to
understand. Man can make something as simple or complex as he choses,
and in looking for something he believes is complex, he can miss the
simple (turning the screw a quarter turn).
Ted> You approach AI by looking at yourself objectively, but realize
Ted> that in fact self-observation is subjetive.
Yes, I do realize self-observation always has some element of being
subjective. There is also a little trick one can use in being
objective about self-observation. By realizing one cannot change the
past, the use of hind-sight is rather helpful.
Ted> What about the forward loop of actually creating an intelligence?
Ted> I really can't imagine designing an intelligence that simply is.
Ted> I believe that any approach will involve a learning and growing
Ted> process. I don't see any reason why that proccess should stop
Ted> once a human level of intelligence is acheived.
As defined in our common reference book of knowledge expression tools,
intelligence is "1) capacity for reasoning, understanding, and similar
forms of mental activity. 2) manifestation of such capacity." The key
word here is "Capacity". This capacity is in the form of constants
upon which the variables pass (variables being knowledge, information,
data, etc.). If such constants were not (change), then our capacity
would also change and in effect there would be constant change in our
degree of mental capacity or ability. However the human mind has the
capacity/ ability to understand anything (thanks to the abstraction
ability of consciousness), so the constants are constant and
intelligence simply is --- a matter of constants that give way to
capacity for....
Sure learning and growing are related to intelligence in that learning
and growing cannot happen without "capacity to...". I too see no
reason why such a process should stop when a given level of
intelligence is acheived, however without the capacity to create and
use abstractions, memory and mental activity finds a limit.
Ted> I am convinced that we have hardware which is on the order of
Ted> complexity of the brain. We do not have a clear direction
Ted> toward intelligence.
I am convinced we have the software to enable machine intelligence,
only it has not yet been put into the proper configuration.
Date: 01-05-91 (10:31) Number: 276 of 287 (Echo)
To: TED KIHM Refer#: 259
From: TIM RUE Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL
Tim> Consider the process of evolution from the big bang to the
consciousness of man.
Note that out of all possibilities that all possibilities don't
happen, but rather only those that the previous interactions
allow.
Ted> No limitations! We perceive sequentially, but that is merely
our crutch! What happened before the Big Bang? Is anything
orthogonal to reality? Film at 11:00.
Ted, would you get the lights please? Thanks.
Although our individual sences need to reference time or movement
in order to be functional, we do indeed have and use parrallel
perception, accomplished by using a combination of our sences.
Determining the speed of sound may be done using sight and sound.
Sequentially the human eye is constantly moving (no matter how still
one trys to keep them) and there are simple experiments one can do to
determine that time/movement is indeed a constant upon which our
perceptions travel (touch: place your hand on an object and hold it
very very still, given enough time your sence of touch seems to stop,
your nerves may cause you to move however).
Through the use of parallel perception such as the eyes, we are
able to determine depth of field (Please Do Not Try to drive a car
with only one eye, any optomitrist will tell you this). By using our
sences in various combinations we are able to preceive a great deal
more than only being able to preceive sequentially.
From "Surpassing Einstein's Ultimate Goal":
"All events of the universe fall within nature's mighty Grand
Cycle, the dominating, all-inclusive energy wave involving the entire
universe. That cycle consist of nature's longest energy wave exactly
counterpoised with nature's shortest energy wave. All other cycles,
waves, or forces of nature, ranging from cosmic and gamma rays to
radio waves fall within the Grand Cycle. ...The Grand Cycle is
described in Table 2 Below:
----------------------------------------------------
| Table 2 |
| |
| The Total History Of The Universe |
| (omitting the unifying element of consciousness) |
| is contained in THE GRAND CYCLE which consists of|
| |
| The Googol-Year Explosion = Half-Cycle, Long Wave|
| |
| The Googol-Year Implosion = Half-Cycle, Long Wave|
| |
| The Googolth-of-a-Second = Full-Cycle, Short Wave|
| |
|(a googol equal 10E100 or 1 followed by 100 zeros)|
----------------------------------------------------
The following capsulized account of the Grand Cycle starts with the
birth of the universe in a big-bang explosion as illustrated in Table
3 Below: (Sorry folks, no graphics, read the book).
All existance in the new-born universe pushes outward in nature's
longest energy wave lasting trillions of years until all energy is
spent. Then gravity takes over. All existence in the universe then
pulls back together in an equally long wave lasting trillions of years
untill all energy is restored. Approaching the end of that long wave,
the universe becomes so densely packed that gravity collapses the
universe into a crushing black hole. An then, the Cycle
cataclysmically ends in nature's shortest energy wave as the universe
turns inside out in an instant of effect another big-bang explosion
that rebirths another universe in a new Grand Cycle as shown in Table
3 above. (No graphics, read the book)
The actual time frames of the cycles above are not yet known. But
with increased knowledge, they will someday be measured directly and
accurately. Until then, the googol unit of time serves to express the
unimaginably long and short time periods between the longest and
shortest energy waves in our universe.
Previous Table 3 shows that all activity during nature's longest
wave, the googol-year exploding/imploding cycles, exactly equals all
activity occuring during nature's shortest wave, the
googol-of-a-second cycle. A full understanding of that seeming paradox
will evolve over the next few pages." (Read the book)...
An interesting observation about Time is that in a black hole time
becomes distorted and in infinity Time has no reference point.
Flap Flap Flap.....
Would someone get the lights please?
BTW, Table 3 is labeled with "Einstein's Theories Demonstrated"
Date: 01-07-91 (03:22) Number: 281 of 287 (Echo)
To: TEAGUE SHERIDAN Refer#: NONE
From: RANDY BENNETT Read: NO
Subj: What are limitations? Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
=> Who or what is a turing?
Alan Turing was a genius who lived in England aound and during WWII.
He proposed a test, called "The Turing Test", to determine if a machine
was truely artifically intelligent. If a person, communicating only
through a terminal, is unable to tell if the thing they are
communicating with is a machine or a person, then the machine they are
communicating with is artifically intelligent. That's the short form.
And yes, Bill Gibson did borrow the name for "Neuromancer".
---
þ QNet 2.04: ILink: HyperLinc East þ Newtown Square, PA þ 215-356-1630
Date: 01-09-91 (11:00) Number: 283 of 287 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim: Would someone get the lights please?
Good stuff as long as the topic cops don't get us! Here's a cosmology
for you, and I invented it to boot! Suppose the universe is radically
simpler than we now posit. In my supposition, the universe is all
possible configurations of matter in space. No laws of motion,
relativity, etc., simply existence. One way I've thought of this is as
a universe with an algorithm which randomly distributes matter
throughout the universe every goololth of a second. Keep doing that for
a few googolplexes and you'd have a whole lot of configurations of the
universe. Undoubtedly, one of those slices of the universe would be
remarkably similar to your current perception of the universe.
Thus, I propose that rather than existing as beings in a 3-D universe,
we are more like threads of universal instants which we string together
in order to create reality.
---
þ DeLuxeýáa #681s ú If AI is so dern smart, why ain't it rich?
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 01-09-91 (04:49) Number: 285 of 287 (Echo)
To: DARRYL BASNER Refer#: NONE
From: RANDY BENNETT Read: NO
Subj: ARTIFICIAL INTELLIGENCE Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
=> Are neural networks the same as fuzzy logic chips?
No, they are quite different. First, when software people mention
"neural networks" the word "simulations" should be mentally added. All
we can do in software is simulate a neural network. But, even the
hardware neural networks (of which there are few) are different from
fuzzy logic. The strength of fuzzy logic seems to be its ability to
accept "maybe" as an answer, right along with "yes" and "no". So, maybe
certainty factors used in expert systems are closer.
BTW: I have already said publically that by the mid-90's neural nets,
fuzzy logic, object-oriented programming, and case-based reasoning will
be duking it out for dominance, since they will all do very similar
things very well.
---
þ QNet 2.04: ILink: HyperLinc East þ Newtown Square, PA þ 215-356-1630
Date: 01-09-91 (04:49) Number: 286 of 287 (Echo)
To: TIM RUE Refer#: NONE
From: FREY WAID Read: NO
Subj: And such is thought... Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
TR> Anthropomorphizing is very important for humans, if man did not
TR>do this he would not exist.
Anthropomorphizing is important in an evolutionary sense. In other
words, we act/react the way we do because we evolved the need. However,
this anthropomorphic way of viewing things can be confusing, inaccurate
and even wrong when it comes to mental abstraction. I'm not going to argue
whether or not man still needs to be so anthropocentric. But, when it
comes to definitions of such things as consciousness, cognition and
intelligence, we need to be very careful about how we apply those terms.
Do you truly believe that intelligence can only be found in man or in
creatures like man? If so, then AI would be an oxymoron.
TR>Perhaps it is this anthropomorphizing that is giving man the drive to
TR>create a machine that will mimic man? Silly perhaps to create such a
TR>machine which has such limitations.
One part of AI is just that: Machines that mimic man. But, one part.
AI is much, much larger than just that narrow vision. It is interesting
though that you're willing to admit limitations due to man's
anthropomorphic nature <grin>.
TR>Consider this statement: "Computers can calculate complex equasions
TR>very fast, equasions which would take man lifetimes to calculate
TR>without his invention/creation of computers". Understand the inherent
TR>priority.
Okay, so what does this have to do with AI? And to what kind of
computers are you referring?
TR> I believe neural net have an important usefulness (perhaps
TR>input/output filters), but why not have a system which contains the
TR>ability that neural nets give along with the calculation ability? In
TR>doing so such a system would not have something that a human has, the
TR>limitation of calculating ability.
No, reason. In fact, you've brought up a good point. I'm sure when the
technology avails itself, people will augment their brains with just
that. I probably would (if I could afford it).
TR>Sure we can build/create a machines which can interact with it's
TR>environment, and we can call it cognition. But it's really just the
TR>appearance of cognition. Consider a person in a coma, they appear to
TR>have no cognition, yet for those who come out of a comma, there must
TR>have been some cognition (something to use as a reference/feedback in
TR>direction to go).
I'm a bit confused here. You defined the appearance of cognition as the
ability to intereact with the environment. And then you defined
cognition as something to use as a reference in a direction to go. I'm
not sure what the last means. Are you saying that a cognition is
essentially a memory that one thinks about and then acts on? I think
I'm still unclear as to your definition of cognition. So, I'll let it
rest until I understand what your point is.
TR>Without cognition a human will cease to exist.
Well, the human could be a vegetable, hmm?
TR>Without cognition a machine simply continues to run via how it is
TR>designed.
Which is what we do -- run according to our evolutionary design.
TR>There is no internal will/motivation to do so. To give a
TR>machine such will/motivation requires man to give it total
TR>independance from man, and in effect risk losing it's usefulness to
TR>man.
First, I don't think you can separate will/motivation from intelligence.
You can't take away self-preservation/procreation/... from an
intelligent being. So second, it would not be up to "man" to give or
take will/motivation from an intelligence that man may or may not
rightfully claim to be creator. Third, usefulness, shmoofulness <grin>.
TR>Also in doing such would also require the machine to have the
TR>ability of self-creation or some form of self preservation (a matter
TR>of independance)...
Okay, but remember don't think of those terms in the human sense. These
terms are much larger than that.
TR>...and certainly anthropomorphizing (cognition -relating enviroment to
TR>oneself).
Are you now defining coginition as relating the environment to oneself?
And why is that anthropomorphic? The environment (whatever that may be)
is important to interact with at least in terms of self-preservation.
TR>Emotion are a result of ones internal value system applied to
TR>information (even incomplete information).
Hoo, boy, now you've opened a can of worms. Are you saying that babies
don't have emotions because that haven't lived long enough to establish
an "internal value system?" Are you saying emotions must have
information to set them off and so manic depressives don't actually have
that emotive state?
TR>Emotions are not to be judged or used to judge a person or their values.
TR>Actions however can be judged, reguardless of the logical or emotional
TR>base of the action (thought does not call for action, action is a choice).
TR>Such a statement as "nature is beautiful" is emotion based and I'm not sure
TR>if the word anthropomotphize would be correct to use here.
I think though you understood my point that our view of nature and
nature's view (yes, it seems silly) of nature do not converge. But, so
what if it's a statement based on emotion (although I could argue that it
isn't). It doesn't change our anthropomorphic view of things. It
supports it and my point.
So, here's another one. The sky is blue. Well, a dog does not see in
color so I guess a dog would probably say it's gray. The grass probably
wouldn't care. A bat doesn't see well at all but echo-locates, so the
bat would probably call it bottomless. I think you probably get it.
TR>The complex cognitive process of consciousness results in or allows
TR>abstraction (in the purest sence). Such pure abstraction cannot exist
TR>without effort and disconnection form physical reality. In turn such
TR>abstractions can be put into application in reality, such as taxes
TR>(which have no real connection to reality). Some confussion can
TR>result in viewing the application in reality, but clarity can be
TR>found in recognizing the requirements needed for the application to
TR>work (with taxes it's a matter of human decission and sometimes
TR>application of force or fear). However, taxes are an application of a
TR>pure abstraction. So in a very strong way, consciousness is not a
TR>matter of interacting with ones environment but rather creating ones
TR>environment. Yet, without cognition there can be no awareness of what
TR>to create and from what. Taxes are one of the creations of conscious
TR>man.
This may be the best thing you've said yet. But, you still have not
encompassed the whole of consciousness. Abstraction comes from
consciousness. This I will not quibble with because I still don't know
what consciousness is (or at least how you're using it). If you are
saying that if one cannot abstract, one is not conscious (since it is a
logical fallacy to say if one can abstract, one is conscious), then perhaps
we have some criteria for determining what is not conscious. But, since
abstraction is not a constant or easy activity, determining what is not
conscious will be very difficult. And who's to say that this thing is
not conscious just because we haven't observed any abstraction going on.
So, I guess we're back where we started, and your definition of where
abstraction comes from hasn't much helped.
Frey
---
þ EZ 1.33 þ Roight!
þ ILink: -=The MBT=- þ Blacksburg, Va þ (703)552-8767 HST/DS
Date: 01-11-91 (11:03) Number: 289 of 292 (Echo)
To: ALL Refer#: NONE
From: TIM RUE Read: (N/A)
Subj: FULL CIRCLE Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
Phase 1:
The Big Bang.
Atomic and sub-atomic particles combine to form new atomic structures
which further combine to form other structures, etc. During this
contining process there is also a process of elimination of some of
the structures via conflict between structures. Also during this on
going process there are structures which become relativly constant.
Three events - creation, elimination, and stablization of constants.
As the process continues more and more constants develope, which in
turn allows matter of larger size to build on the smaller structures.
A variety of structures develope into what we know as the universe.
Out of this variety is a portion which allows the formation of life
sustaining conditions and upon further evolution of some of these
portions comes the single cell life form.
Phase 2:
Life.
The evolution of life is much like that of Phase 1, having the three
events. As a variety of life forms develope there is a small portion
which evolves in the direction of early man. At this stage man
functions in his enviroment much like other life forms, but overall
mans evolution does not stablize to fill a nich in the enviroment. As
man evolves his ability to organize improves, he discovers/creates
some primitive tools which allow him to better survive in his
environment. One such tool is that of working together. But as man
evolves, and his society become more and more complicated, his
bicameral mind begins to over-load and break down.
Phase 3:
Consciousness.
Also containing the three events of Phase 1. In order to survive, man
creates consciousness which allows him to evolve further (create).
However, in the variety, some humans also finds they can use deception
to control other humans (the elimination/limitation tool of mysticism).
As evolution continues mans consciousness allows him to win out over
mystism a peice at a time (constants are realized). Man gains more and
more control over his environment as his technology increases, but
mystism also continues to parrallel mans technology. Mans population
grows as a result of his increased control over his environment but
society again is becomming more complicated. In order to continue his
advancement/evolution man creates tools to handle the increasing
complexity of his society. Amoung the variety of tools comes the tools
of Knowledge and Intelligence (AI and it's many fields). But mystism
is still present so the evolution of this technology stumbles again
and again, yet out of all the research, a variety of application are
being developed which fit into the society man has created. However,
one area in the variety of AI research fields seems to never advance
(always having unfilled promises), the area of Actual Machine
Intelligence which is capable of fulfilling all the unfulfilled
promises of AI. ******* YOU ARE HERE *******
Phase 4:
The absolute destructive nature of mystism and deception realized.
Three elements of Phase 1 here also. Mans society had grown so complex
in Phase 3 that it broke down, deception and destruction became so
wide spread man was forced to realize the constants amoung all humans
and to respect it. The one area of AI which had not advanced in Phase
3, now advances via discovering the constants of Consciousness, the
mystism and deception which held it back are gone. Man again continues
to evolve and with his new found technology becomes more and more in
control of his enviroment and life. Man discovers the Unified Field
(application, not theory) and is in total control of his environment,
for he knows he will live forever (infinity) because it is being a
conscious being that allows him to create a universe via causing a
black hole. Which leads to a big bang (going Full Circle). Man also
accepts the company of other conscious life forms far more advanced,
and conscious life forms he has created. Man finds his nich in the
environment of existance.
Consciousness - the missing element of the Unified Field Theory. The
Unified Field equasion - the missing algorithym for machine
intelligence (Knowledge Navigational Mapping and Internal
Representation - Input/ Translation/ Output).
===================================================================
This posting consist of a sequence of Knowledge Expression Tools which
(due to mystism and deception) have the ability to be in conflict with
ideas based on mystism and deception. It is not, in any way, a
detailed description of atomic, subatomic, structures, life,
consciousness, etc. But rather a very brief summary and
projection of conscious evolution going full circle.
Date: 01-15-91 (09:28) Number: 295 of 297 (Echo)
To: TED KIHM Refer#: 283
From: TIM RUE Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
-> >Tim: Would someone get the lights please?
->
-> Good stuff as long as the topic cops don't get us! Here's a
-> cosmology for you, and I invented it to boot! Suppose the universe
-> is radically
There may very well be different realities (your's sound like
superstring theory), however, in order to function within the realities,
man needs to (in working together) agree on the tools he uses. This
applies to Language and the fact that it's the agreed use of language
and words that make these tools useful. As humans working together there
are the tools he uses in order to advance, if man does not yet know of
some unseen reality then he does not use it, but continues to function
with what he does know. As to the comos and mans understanding of it
there are the tools he uses to build his knowledge of it. Untill man
becomes aware of other realities he has only available the tools which
have been verified to work (perhaps not 100% but well enough to
accomplish something).
An interesting thought is the question of - what tools did
DaVinci use that enabled him to invent the things he did (going beyond
what his enviroment contained). He may not have wrote backwards in
effort to hide his knowledge from others whom may have thought him
insane, but rather a side effect of the tools he used enabled him to do
so naturally.
Mans lack of awareness of other realities does not stop him from
advancement, yet these other realities must (if they exist) have some
effect on what man does achieve.
The subject mater of this conference is defined, but it's
subtopic is knowledge processing (any knowledge) and as a result the
decussion of other fields of knowledge are not prohibited, as long as
there is a worthwhile connection to the topic of this conference.
Beside, this conference is quite slow compared to some of the other
conferences. To draw a clear connection of this message to the
conference topic -- Knowledge Expression Tools do not define absolute
knowledge but rather parts, and it is these parts which are processed by
AI.
Date: 01-14-91 (09:13) Number: 296 of 297 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: 01-16-91 (08:03) HAS REPLIES
Subj: FULL CIRCLE Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim Phase 4:
> The absolute destructive nature of mystism and deception
> realized.
You mean god is not going to stop the Persian war???? Isn't it fun to
have a mature mind amongst the primitives!??!
---
þ DeLuxeýáa #681s ú If you meet the Buddha on the road, Kill him.
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 01-16-91 (08:03) Number: 297 of 297 (Echo)
To: TED KIHM Refer#: 296
From: TIM RUE Read: NO
Subj: FULL CIRCLE Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
-> >Tim Phase 4:
-> > The absolute destructive nature of mystism and deception
-> You mean god is not going to stop the Persian war???? Isn't it fun
-> to have a mature mind amongst the primitives!??!
It may not be a good Idea to respond but I will say this. I and the odds
of cosmic sicence realize the probability of other intelligent and
conscious life in the universe (and there are those of us humans who
know it's not a probability but a fact). Suppose religion has a
foundation connected to alien interaction and human mis-interpretation.
In this case the religions could be considered mystism (the actual
realities not realized). How does this relate to AI? Simple, the same
way the idea of the world being flat and the center of all things, it
leads man down the wrong road. AI development to really evolve must
overcome the mystism in society (anyone who's had to bring an AI package
into application has had to deal with resistance somewhere).
As to the Persian war, I find it worth noting where christianity came
from and the location on the earth where there seems to be constant war
and conflict. Perhaps this says something about the validity of this
religion and the depth of mystism in society and the wall AI will need
to break through in order to advance.
Q: How is mankind to evolve AI if he's looking through distorted
glasses in reguards to himself?
Thanks Ted for your responce. I needed that...
Date: 01-17-91 (04:36) Number: 302 of 302 (Echo)
To: TIM RUE Refer#: NONE
From: FREY WAID Read: 01-18-91 (08:39)
Subj: AND SUCH IS THOUGHT... Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
TR³Frey,
³ How do you define the following:
³ Intelligence?
³ Cognition?
³ Emotion?
³ Consciousness?
Well, let's see. In order for me to be comfortable with any definition
I must incorporate certain empirical evidence with a healthy bit of
philosophical skepticism. This does not make me a Skeptic or an
Empiricist because my intent is to adopt a proposition until it can be
proven false. Similar to "innocent until proven guilty." So, onward....
I believe intelligence must have a broad (almost generic) definition.
The only non-subjective, mostly non-anthropomorphic definition I can
think of is:
Intelligence is the total number neurons related to the rate of
neural interconnectivity in a given system.
I'm using neurons and neural interconnectivity in the most generic
sense. Anything which can be remotely related to these two functions
does by definition have intelligence. Cockroaches, slugs, plants,
people, rocks, planets, universes. Sort of like the Gaia principle.
But, what we're mostly concerned about is the "higher" intelligence or
more appropriately the intelligence like ours.
Let me jump down to consciousness here. As I said before, I think
consciousness is usually defined subjectively. The only way I can think
of consciousness non-subjectively is by saying that it's some "critical
mass of intelligence." I know this isn't very satisfactory, but it's
all I can think of.
But, let me harp on consciousness here. I really don't think it matters
whether or not we can call something conscious. We, as humans, have a
strong tendency to call things similar to ourselves conscious and those
not....well. So, an alien would be thought more conscious if it had
similarities to us. But, it would be hard to deny that the alien is
intelligent.
Emotions in humans and animals have been evolved to help personal and
species survival. I'm not going to give examples because I'm sure this
is all very obvious. But, we know that emotions are related to
hormones, pheromones, heredity and environment. Do we know that if
there is intelligence, then there must be emotion? No. But, if there
is emotion, then there will probably be "higher" intelligence.
In my definition paradigm, cognition is not very useful. I would assume
that it is a type of learning. But, the moment it's linked to
consciousness, I hesitate to use it for fear of inconsistencies.
Inconsistenphobia <grimace>.
Please punch holes,
Frey
---
þ SLMR 1.05 þ Strange women lying in ponds...
þ ILink: -=The MBT=- þ Blacksburg, Va þ (703)552-8767 HST/DS
Date: 01-17-91 (18:09) Number: 307 of 316 (Echo)
To: DARRYL BASNER Refer#: NONE
From: WILLIAM WRIGHT Read: NO
Subj: neural net vs fuzzy logic Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
RB> Are neural networks the same as fuzzy logic chips?
It might be useful to add a word of explanation (for those
who aren't familiar with these terms):
"Neural net" and "fuzzy logic" are two different methods of
processing data. Each of them can be implemented on
paper, or in software, or in hardware.
***NEURAL NET***
You can think of a "neural net" as a network of
interconnected nodes, usually arranged in layers:
node-1 node-2
/ \ / \
/ \ / \
node-3 node-4 node-5
/ | \ / | \ | \
/ | \ / | \ | \
node-6 node-7 node-8 node-9 node-10
I didn't waste time drawing a fancy diagram. Networks can
be as many layers deep as you wish, with whatever
connection pattern you wish, including connections both up
and down, skipping over layers, etc etc. Of course, the
simplest patterns have been studied the most.
Each node has a very simple job: it adds up all the inputs
which it receives. If the summed input is large enough,
the node sends an output to other nodes.
In other words, you feed a certain input pattern into one
layer (usually the top layer), and you wait until some
other layer (usually the bottom layer) develops a stable
output.
The key ingredient of a neural network is interaction
between nodes. An input pattern may bounce around the
network many millions of times before the output layer
finally stabilizes. It's sort of like a human being who
sees something, has an initial response, then realizes
something doesn't fit, so he thinks it over and revises his
conclusion.....
You might feed the image of a bird into the top layer, and
wait till the bottom layer settles down and outputs a
string of letters such as "B-L-U-E J-A-Y".
A neural network is a "distributed processor" because no
single node sees all of the data that bounces around the
network. In the "Blue Jay" example, you can't identify a
particular node which decided what kind of a bird it was,
or even that it was a bird instead of an elephant. Each
node contributed a little bit (no pun intended) to the
final result. Oftentimes disabling a single node does not
change the final output because the node had only a small
effect on the final output.
***FUZZY LOGIC***
"Fuzzy" logic is derived from set theory. Instead of
saying:
BOB weighs 200 pounds. Hence he's a member of the set
named HEAVY.
you might say:
BOB weighs 200 pounds. He's a member of HEAVY with
degree=0.75. If BOB weighed 100 pounds, he'd be a
member of HEAVY with degree=0.05; and if he weighed 300
pounds, then he'd be HEAVY with degree=0.99.
In other words, BOB's membership in HEAVY is "fuzzy" and is
measured by a number.
This seems like a childishly simple concept. But it turns
out that you can replace most of the operations of set
theory (such as UNION or INTERSECTION) with simple
arithmetic operations (such as MAXIMUM or MINIMUM of the
degree of membership).
A typical problem would be: "If Peter Falk is 10% virile
and 99% clever and 24% low-salaried, would he make a better
TV detective than Sean Connery who is 98% virile and 2%
clever and 11% low-salaried?"
Set theory would conclude either "yes" or "no." Fuzzy
logic might conclude "93% yes" (depending on how you
defined membership in the fuzzy sets of VIRILE, CLEVER,
LOW-SALARIED and GOOD TV DETECTIVE).
The appeal of fuzzy logic is that human thinking is seldom
absolute. If you apply for a job, the boss will decide
that you are "reasonably intelligent" or "intelligent
enough" or "off the deep end" (but exactly where is the
'deep end'?).
The newest VCR cameras use fuzzy logic to get rid of jiggles
in the picture. The camera must decide what "jiggle"
means, and this is definitely a matter of degree.
You're probably moving the camera unevenly as you pan across
the scene, and maybe a butterfly is fluttering in the
background, and a flag is flapping in the breeze. Your
kid is splashing in the swimming pool and droplets of water
are reflecting sparkles of light in random directions, and
the water ripples are changing direction as they collide
with each other. So which pixel defines "no jitter?"
What you really want from the camera's processor is: Search
for the _best_ fit to the set called UNIFORM MOTION --
even if your kid is thrashing in every direction at once
and almost everthing in the scene (including your camera)
is moving in a jerky fashion.
This is a "fuzzy" set!
---
þ SLMR 1.0 þ I BBS because nobody can read my handwriting.
þ ILink þ Console Command HQ þ Santa Barbara, CA þ 805-683-0499
Date: 01-18-91 (09:02) Number: 314 of 316 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim An interesting thought is the question of - what tools
> did DaVinci use that enabled him to invent the things he
> did (going beyond what his enviroment contained).
Excellent example, but DaVinci didn't go beyond his environment, rather
he understood that his environment contained all future inventions. His
environment became subject for study.
>Tim To draw a clear connection of this message to the
> conference topic -- Knowledge Expression Tools do not
> define absolute knowledge but rather parts, and it is
> these parts which are processed by AI.
We've all heard the parable of the blind men encountering the elephant.
I think an appropriate analogy for AI research includes the blind men
and the elephant. In this case however, its a tag team wrestling match
and the elephant has been oiled up with crisco....
I think we've been doing fairly well at trying to express some of the
emergent properties we would like to see an intelligence exhibit. But
as you are fond of pointing out, these are not fundamental enough.
Let's start back at the beginning again! What is learning? Can we come
up with an example which is "atomic" learning? What is the smallest,
most precise, simplest, singular piece of learning possible??!?
---
þ DeLuxeýáa #681s ú You can't always .GIF what you want.
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 01-18-91 (09:20) Number: 315 of 316 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: FULL CIRCLE Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim As to the Persian war, I find it worth noting where
> christianity came from and the location on the earth
> where there seems to be constant war and conflict.
Chrsitianity as well as Muslim beliefs both began in Jerusalem. I find
it quite interesting that an older name for Iraq is Mesopotamia. The
birthplace of civilization?? You bet! Looking at all that sand its
hard to imagine what was once the lushest, most vibrant part of the
world. Slash and burn agriculture absolutely DESTROYED the bounty which
allowed us to be. The sahara is not a naturally occuring phenomenon.
We did it, to our own mother to boot. Now we bomb it, soon we may nuke
it.
The one thing of which I am certain: god has nothing to do with the war.
It is ours and ours alone.
>Tim How is mankind to evolve AI if he's looking through
> distorted glasses in reguards to himself?
Intelligence always develops that way. Infants do not posses the
ability to intorspect at all and must rely upon the distorted vision of
others which is subect to miscommunication to boot. At some level of
maturity the individual becomes able to assume the responsibility of
fixing their own galsses!
---
þ DeLuxeýáa #681s ú Theory #2: DINOSAURS CAUSE EARTHQUAKES!
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 01-18-91 (10:08) Number: 316 of 316 (Echo)
To: FREY WAID Refer#: NONE
From: TED KIHM Read: NO
Subj: AND SUCH IS THOUGHT... Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Frey I believe intelligence must have a broad (almost
> generic) definition. The only non-subjective, mostly
> non-anthropomorphic definition I can think of is:
>
> Intelligence is the total number neurons related to the
> rate of neural interconnectivity in a given system.
>
> Please punch holes,
I understand the motivations which led you down this path, but are you
really comfortable with this definition? Whats the intelligence
difference between a dead brain and a functioning live one? It would
seem that your definition could be reduced to thermal-dynamics and that
anything which was far enough from equilibrium (ie complex) would be
intelligent.
Well, I don't have a good definition ready myself, so I'd best put up
what meager target I can.
I would like to define intelligence in terms of stimulus/response.
Intelligence seems to be able to temper stimulus with experience in the
formation of a response.
Suppose we have a hill of ants and an anteater comes along and "sniffs"
a yummy dinner. In this particular example however, the anthill is in
very rocky soil and try as he might, the anteater cannot dig into the
hill. The straight stimulus/response model would suggest that as long
as the anteater was subjected to the stimulus of smelling the ants, he
would keep digging in the hill. The longer he dug, the hungrier he
would get, the stronger the stimulus repsonse. Eventually though, the
anteater incorporates the experience of frustration and toddles off in
search of better anthills.
Now a chimp with a taste for ants comes along and smells the same hill.
This clever primate however exhibits much more intelligence. Knowing
that the hill is too hard to dig, he selects a slender branch and strips
it of leaves. Slinding the branch down the ant hole allows ants to
stumble onto the stick and the chimp has a delicious treat.
Please punch holes!
---
þ DeLuxeýáa #681s ú If AI is so dern smart, why ain't it rich?
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Date: 01-21-91 (11:51) Number: 317 of 320 (Echo)
To: TED KIHM Refer#: 314
From: TIM RUE Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
Ted> I think we've been doing fairly well at trying to express some of
Ted> the emergent properties we would like to see an intelligence
Ted> exhibit. But as you are fond of pointing out, these are not
Ted> fundamental enough. Let's start back at the beginning again!
Ted> What is learning? Can we come up with an example which is
Ted> "atomic" learning? What is the smallest, most precise, simplest,
Ted> singular piece of learning possible??!?
Learning is not fundamental enough. What is learning to be set upon?
Where is the "Capacity" and what is the "Capacity" enabled by?
The more we understand the subatomic structures, the more we are able
to control the properties of the larger (the products we make and
improvements to them). By looking at the development/evolution of
electronics we can see at different stages of the technology what the
"Capacity" was for producing products. Untill the Transistor was
developed, tubes were used and due to the limitations of tubes (size,
heat, lifetimes, etc..) our product potential was limited to what was
produced. The Intergrated Circuit allowed better control over the
short commings of tubes and thus better products. To jump ahead to the
present, the hardware of computers is for the most part
stable/constant/standardized enough to allow a good foundation for
productivity, but this is not the software that interfaces the
computer to the user (speaking of OS's which are software even though
they may be put into ROM chips). The hardware has the "Capacity" to do
much more than the software allows it to, but without the software
user productivity is nill. At this point it may be recognized that the
software (OS) gives a "Capacity to" do what all we currently do with
computers. New OS's have better "Capacity to".
The Hardware has it's "Capacity" which gives the software a foundation
on which to build upon. The software, in turn, has it's "Capacity"
which gives the user a foundation which to build upon.
Learning must have a foundation upon which it sets. A foundation which
has the "Capacity" needed to allow learning. Learning is a product of?
So what must be, that will give the "Capacity to" learn?
With these questions in mind, the answer(s) to your questions are a
matter of recognizing the subatomic elements needed to enable a
foundation for learning to set upon. One element is memory, another is
of course input & output, feedback is another (verification), and
there are the other tools of processing needed (perhaps a second level
of sub-atomic elements that will enable or give a foundation upon
which these tools set). By determining the very basic elements needed
to create a foundation for learning, your question "What is the
smallest,
most precise, simplest, singular piece of learning possible?" will be
answered.
When does a view of a single "atom" not become a view of a single
"atom"? When you go to a different resolution. Learning does not exist
when looking at a part of learning. Reflecting back on the earlier
postings, it is somewhat clear, when looking at the limitations of
man, it is difficult to see (if not impossible) mans unlimited
abilities. A part has limitations, the whole does not. In view of the
"Full Circle" of mans evolution (my presentation of it) it is clear
that I appeared to contridict myself in statements about limitations
of man. However, clairity is found in looking at the whole that there
is no contridiction, only resolution.
A given computer has it's hardware as a constant and it's variable the
OS. Stabilizing the OS, using it as a constant, it's variable is the
application software. Stabilzing the software (making constant), the
variable is the user. At this point the constantS consist of
hardware/OS/software and the variable the user. The user has the
foundation of the constants to rely and build upon.
Ted, what I have been expressing (or trying to) is that there is a
somewhat recursive relationship between the "Constants" and
"Variables" reguarding "Capacity to". Capacity to allow AI.
I do believe we have stabilized enough (made constants) to enable AI
to evolve more rapidly. Only we have not yet put these constants in
the proper confiuration (and leave as a constant it self) to enable AI
to evolve more rapidly. There is that missing link somewhere between
the hardware and knowledge/data/information we what to process via AI,
regardless of the "knowledge/data/information" type. The standard
foundation for AI is missing, the field of AI lacks standardization
BECAUSE it has NOT FOUND this missing link/foundation/CONSTANT.
Date: 01-23-91 (11:09) Number: 319 of 320 (Echo)
To: TED KIHM Refer#: 315
From: TIM RUE Read: NO
Subj: ESTABLISHED CONSTANTS. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
Ted, lets start from the beginning in reguards to the tools (constants)
we have to build with.
The first constant which all else comes from (the subject is of
course computers) is how data is represented, reguardless of what the
data is. The constant is a two position switch which is either on or off
or in computer terminology 1 or 0.
I know this really is basic, but humor me. We're recognizing constants
starting with the first and most basic element and in doing this we'll
develop a process which will help us recognize other constants (and
eventually the constant upon which AI needs).
What is the next established constant we have, which uses the
first as a constant?
Date: 01-22-91 (08:01) Number: 321 of 322 (Echo)
To: TIM RUE Refer#: NONE
From: TED KIHM Read: NO
Subj: FOOD FOR THOUGHT. Status: PUBLIC MESSAGE
Conf: AI (64) Read Type: GENERAL (+)
>Tim Learning is not fundamental enough. What is learning to
> be set upon? Where is the "Capacity" and what is the
> "Capacity" enabled by?
Am I being thrown off by your usage of the term "capacity" or what?
Your discussion of memory, feedback loops, etc., seems unecessarily
predetermined. I am more interested in defining what learning is rather
than how it is implemented. As I see it, it is pointless to design the
capacities you mention unless one can develop a criterion of behaviour
which the capacity will enable.
One could cast this discussion into terms of the Turing test. What is
the simplest Turing test possible? Theorizing about the Chinese room in
our universe is pointless. Simplify the universe to the point where
omnipotence is almost the same thing as ingnorance. Simplify the Turing
test until it is merely a test to determine omnipotence or ignorance....
---
þ DeLuxeýáa #681s ú Young pigs grunt as old pigs grunted before them
þ QNet 2.04: ILink: Sound Advice BBS þ Gladstone þ MO
Timothy Rue (AAi member)