Are you attempting to imply that a neuron (and a neuron is a biological thing)
is anything at all like any artificial device?

Would you be kind enough to indicate exactly what your teacher considers
intelligent to be?

I am because that is what I have.<br><br>Spider_plant9

Do you think that it is impossible to simulate a biological process?

I can send you an e-mail with the document of the
assignment. Generaly I have to create a program in Prolog
that helps the user choose a PC. I also have to make a
representation in Semantic networks, or Frames, or Propositional
logic etc and explain why I chose this representation.
<br>also analyse the advantages and limitations of the
chosen knowledge representation(s) using the following
criteria:<br> expressive adequacy,<br> reasoning
efficiency<br><br>I hope it is not much ....<br>My problem is which
representation to choose and why and how to make the program
intelligent.<br>Thank you

The teacher sais that it is inteligent if it the
program does not just query a knowledge base of prolog. I
actually dont understant much of it but I can say it means
that I have to make some rules in prolog to provide
the user the right choice of PC. <br>I suppose I can
do that but I dont really know what will be the
rules.

The absolute limit on such a program would be an
interactive iterative resolution of the details of the
machine. It would be intelligent in that the course of the
interaction would be dynamic utilizing the information in the
database and prior responses to decide what to ask next...
<br><br>After that you should tell your teacher to shut up
untill Turing is passed... =P

Me personally actually thinks that it is not only
possible to simulate but also repicate a biological
process.<br><br>Study lots of Grays Anatomy, general psychology, and
neurophysiology and tell me what you think about it. From
studying all that I'd say you'd know just like I do that
it isn't complicated to replicate things just hard
to make the technology to do
so.<br><br>Spider_plant9

What your teacher means is that if you just have
it looking up parts and what not it is just an
average database but if it looks up parts and suggests
cheaper alternative and is also able to have a teacher
tell it through the program of new features that it
will be able to include these new features in later
queries.<br><br>That's just my guess though.<br><br>Spider_plant9

There is a simple algorithm for backpropagation<br>in C in the book _Exploring
Parallel Processing_<br>by Edward Rietman.  See page 188.<br><br>trg

Has anyone ever heard of using Genetic Algorithms or related programming styles
as units (Neurons) in an artificial net?<br><br>If so, how succesful was it and
what does it's potential seem to be?

I wonder if any of the existing NLP systems are
able to handle resolution of abbreviations more or
less properly (I mean rather general ones, to be
sure). For an example, can existing systems determine
whether MP in a given ssentence stands for one of the
following?<br><br>1) Military Police.<br>2) Member of Parliament<br>3)
Modus Ponens<br><br>Felix.

I could tell you how to build a proggy that would do that.  It is just a matter
of the proper examples and a context proggy or two.<br><br>Spider_plant9

Hi!<br><br>Does anyone here code in
Python?<br><br>I'm finishing work on an agent framework in
<br>Python that uses HTTP for the transport of<br>KQML
messages. (The KQML messages are encoded like sort of like
SOIF, so the KQML messages can <br>carry arbitrary
content.) Anyone interested?

Why, I wonder whether you could actually write a
prototype. I must state that I think the problem is hard,
since is requires much of NLP (ellipsis resolution, for
an example) - but I have a feeling that at least a
subset of this problem could be done.<br>Felix.

could you use a markov chain that represents the probabilities that a certain
abbrevaiation will be used in context with certain sets of nouns?

What, using probabilies to understand language!!
Ok serously, it is probably a good idea to use
probability in some cases of NLP, but I don't think so much
in this case. The actual expansion of the
abbreviation is dependant only upon the context it is being
used, thus probabilites leave too much room for error.
For example, NLP in this case may (not likely, but by
probability it could happen) be expanded as note proccessing
language or someother fool thing like that when we know we
are talking about natural language. Anyway that's my
view on it.

Does a computer require the ability to speak, or
just the ability to understand speech to be considered
intelligent? My view is that it need simply be able to
understand what is spoken to it, for if that's not true,
then the question arises are mute people intelligent?
We know they are, so therefor, computers need only
understand some form of communication to be intelligent.
What's everyone's view on this?

Depends on your definition of Intelligence.<br>I
think a lot of more "primitive" animals are<br>actually
quite intelligent -- when compared<br>to our computer
systems. Instead of understand<br>speech, I would like to
say "unsterand <br>environment," instead. If a
computer can <br>understand its environment, learn from
it, and<br>reason about it -- then I would say it's
<br>intelligent. My $0.02<br><br>Nate:SCHCATS!

I don't see what the problem is. Is there not a
high probablity that when I use the abbreviation /NLP/
in a sentence containing the words /software/ or
/AI/, then NLP is equivalent to Natural Language
Processing? I am not sure that a Markov chain would work
dynamically ( as in adding new abbreviations as the software
runs across them or updating the probababilty vectors
to reflect erronious decisions ) but you can at the
very least model the problem that way. I am just
talking about an easy solution that any CS student can do
- know what I mean?

If you let the robot program itself [ self
organizing system ] it will learn these things without us
having to puzzle over what the "rules" are. It will
learn by example.<br><br>When I read a magazine article
the first time a term comes up it is usually spelled
out and the abreviation that they plan to use for the
rest of the article is next to it. <br><br>While it is
nice to apply statistical models to language ... with
the idea of devining its meaning ... I suspect that
when you open the lid and look in on a system that has
taught itself you may find that it has made up its own
rules, aproximations, some of which may be
startling.<br><br>This is especially true for genetically evolved
programs. <br><br>Paul F. Grayson - Chief
Engineer<br>AMERICAN INDUSTRIAL MAGIC<br>1892 Pinewood
Ave.<br>Traverse City, MI 49684-9022<br>(231) 946-0187, FAX (231)
946-1122<br>pgrayson@...

Well,<br><br>Think this way how did you know ESP
and esp. ?<br><br>ESP extrasensory perseption<br>esp.
espescially<br><br>You would only know if I used it in a sentense or I
used the common ways of denoting the difference such
as caps, puncs. , and the ever present
cont.<br><br>My abrv. words are pretty easy to fig. out if you
know cont. ques.<br><br>Cont (text) (tent) (inue)
which you fig. it out.<br><br>how do you know, why do
you know this and how do you add knew ones with
similar to same spelling and extrapolate
meaning.<br><br>Con(text) is most important as well as how common the the
abv. is. If you know how it is commonly used and
confused you know a few of them. Then after that you note
which ones go with which subject and what they are
based on Pb is latin so might expect Hg or Au would
also be latin. This would then tell you that it would
relate to science and that they could have historic
refference also. Pb or Plumbum (guessing what is used to
make a plum) is what Lead is, Hg or Hydrogyrium or
Mercury (also latin), and last on the list is Au or
Aurium (splendor metal or sun metal) that would be gold.
Humans are noted for redundancy in their contexts so a
computer really should have a problem picking it
out.<br><br>now you understand what I mean?<br>s it gtng clr
enuf?<br><br>ICU laffen @ me<br><br>Spider_plant9

Of course I thought it meant Neural Lingustic Processing / Programming but from
my interpretation you can see how I view AI and how it would work in my
eyes.<br><br>Spider_plant9

Okay that says the deaf aren't intellects oh and the blind aren't either.  What
would miss keller say about that?<br><br>Spider_plant9

Of course you don't see the problem. The thing is
what would say choas theory, astrology, tarot, runes,
and the number 12 and 15 have to do with
AI.<br><br>The thing is context is not just based on words and
probable meaning it is based on probable meaning to a
specified group within a specified group of
people.<br><br>complete the sequences<br>r,g, ? (only a three
letter)<br><br>m,t,w,t,?,?,s, ...<br>j,f,m,a,?,j,?,?,?,o,n,?,
...<br>m,n,o,?,q,?,s,t,?,?,w,?,?,z,?,?,c, ...<br>0,7,0,7,1,?,?,?,
...<br>0,86,602,54,0,?,
...<br><br>My point is context and probabily are just the
edge.<br><br>Oh and before you scroll at least
guess<br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br\
><br><br><br><br><br><br><br><br><br><br><br><br><br><br>First: days of the
week<br>Second: months of the
year<br>Third: alphabetic sequece starting with m and
cycling<br>Fourth: sin 45<br>Fifth: sin 60<br><br>Spider_plant9 - no
hard or easy just is

I've done that and learned a lot of how to
program an AI better. I just program little blipps and
watch them build bigger programs. Stop them from
growing and disect them to make better ones and to learn
how to learn better myself.<br><br>I figured you for
a real techy BTW had any luck with the 8051's and
AI apps. I'm just getting into the 8051 and it looks
ideal for the next real live AI child I'm going to
build although PIC is nice I like the 8051's
architechture it seems easier to understand. Still, waiting to
get a PIC book though.<br><br>Spider_plant9 -
Aspiring Engineer (love psychology though)

I am asuming that since you're asking in this
group, you're interested in neurons with regard to
artificial neural networks. If so, see the FAQ for
comp.ai.neural-nets over on Usenet for a start. If you are interested
in biological neurons, I'm sure any decent search
engine will turn up a good
introduction.<br><br>Predictor

Genetic algorithms are optimizers: given a
process (real or simulated) and a set of tuneable
parameters, they find an optimal (near-optimal,
techncially)output by tweaking the process parameters. I'm not sure
how one would use GAs as "units" within a neural
network, although GAs have been used both to train neural
networks (with poor results) and optimize neural network
parameters (with some success).<br><br>Predictor

This Years Christmas Card to you is posted in
PHOTOS at AMERICAN INDUSTRIAL MAGIC.<br><br>Merry
Christmas from the people and animatrons at AMERICAN
INDUSTRIAL MAGIC.<br><br>Paul F. Grayson - Chief
Engineer<br>AMERICAN INDUSTRIAL MAGIC<br>1892 Pinewood
Ave.<br>Traverse City, MI 49684-9022<br>(231) 946-0187, FAX (231)
946-1122<br>pgrayson@...

I don't think that one could do that in a
technical way, after all, wouldn't that be out of the
definition of neuaral net? Where would the parameters come
from? Given these problems, I don't think that GA's
could be units in a neural net. Infact, from my
perspective, the two technologies are mutally exculsive if one
want's good results (one can't combine the two into one
algorithm is what I mean here, not that they can't both be
used in the same problem, I hope I am being clear on
this)