I want to create an android

The AI and robotics community reccomends donating your own
brain for the project.  You are such an intelligent person, if you
transplant your brain into the robot, you will have a superior
product.

That's old Star Trek, i.e., "Spock's Brain" episode. I'm talking Star
Trek the Next Generation with the android Data.

Craig

Actually, if you get a chance, rent 'Saturn 3" with Kirk Douglas and Farrah
Fawcett.

Very cool

good movie

Just had to inject my two cents in here...
If he donated his brain, it would no longer be an android.  It would become
a cyborg.

An android is a humanoid robot.  (entirely artificial, non-biological)
A cyborg is a biological/mechanical hybrid, not necessarily humanoid.

Let's keep our terminology straight here!

Well who knows? If your level of knowledge in robotics enables you to think
that a "Star Trek - Data" type android is possible with current technology,
perhaps you should tell us.

It is not possible with the current state of human knowledge.  If it
will ever be possible, it will be far in the future.  It is not even
known for sure if a digital computer can have a mind like a human,
although many members of the AI community believe this as an act of
faith.  For an excellent presentation of a different view see Roger
Penrose's book "The Emperors New Mind".

Mitchell Timin

--
"In theory, there is no difference between theory and practice.  In practice,
there is."

http://annevolve.sourceforge.net  is for those seeking something to do with their
wetware.

Humans may know that my email address is zenguy at shaw dot ca.

Why wouldn't it be possible? Admittedly, there isn't much in the way of
AI, but programming a mimic of human behavor is not all that difficult.
Humans highly overrate the complexity of the behavior of the average
human. Seems to me, to make it credible, you'd have to make it act sort
of stupid, like human behavior. :-) So, the only problem is the software.

As for neural nets - I expect that is just a passing fad, like bubble
memory. Neural nets was an exercise in the study of how neurons work;
not a very practical solution to AI, imho.

well, AI is not able to create anything like that now but when we give
birth to AI, i think the first machine compareble to "startreks data"
will be created with techniques heavely inspired on the 'process of
creation' of ourselves. So my guess it would be something like an EANN
(evolutionary artificiall neural network) based on principles from
molecular cell biology. eggenberger-hotz is developing EANN's.

if its possible to create 'data' on a symbolic level i think it would
be much later.

and dont pay any attention to these spare-time philosophers just
remember:
they laughed at the wright brother but the also laughed at bozo the
clown

Since you bring up ANNs and Evolution, let me call attention to the URL
below, which is devoted to that.  There is software there, written in C,
that you can download at no charge.

Mitchell Timin

--
"In theory, there is no difference between theory and practice.  In practice,
there is."

http://annevolve.sourceforge.net  is for those seeking something to do with their
wetware.

Humans may know that my email address is zenguy at shaw dot ca.

Thanks I. Myself,

I look forward to trying out your url.

Craig

Hello Craig, here is an interesting site that you might enjoy:

http://www.singinst.org/

Cheers,

Dave

cafeinst@msn.com wrote:

Check to see if the positronic brain has been invented yet.

The solution will not be as simple as a neural network. Even if it has
the same number of neuron as the brain.

Come back in 50 years time and ask again.

--
Steve Wolstenholme     Neural Planner Software

EasyNN-plus. The easy way to build neural networks.
http://www.easynn.com

***********************
C

you question is naive and unsophisticated.

you pose a question that is on the edge or beyond our presence science
(at least in the open literature domain)

for clues as others have said study Penrose and wait about 50 years

P

The answer is obvious, IMO. Since the only intelligence we know of is
biological intelligence and since biological intelligence is made of a
number of integregated cell assemblies (neural subnetworks), it
follows that future androids will use integrated artificial neural
subnetworks as well. However, these subnetworks will bear little
ressemblance with current ANNs. For one, biological networks are
discrete signal processors, i.e., they process neural spikes. It is an
inherently temporal phenomenon.

Why neural networks, you ask? Because it is the only paradigm that can
handle the astronomical connectedness of human-level intelligence.
Nothing else can come close.

Louis Savain

Why Software Is Bad and What We Can Do to Fix It:
http://www.rebelscience.org/Cosas/Reliability.htm

Before the Wright Brothers, the only flying things
we knew of had flapping wings.

Biological systems have limitations that engineered
systems do not.  The wheel never evolved in a
biological system because there is no way to get
blood and nerves to a spinning wheel.  Birds and
insects flap their wings because, lacking propellers
or turbines, they use their wings to generate both
lift and thrust.  But airplanes don't have that
constraint, and no airplane was successful till
the "flapping" was abandoned.


The human brain uses astronomical connectedness,
but that doesn't mean that is the only way to
acheive human-level intelligence.  That may be
the only way to do it within the limitations of
biological neurons, but semiconductors have very
different limitations.  Neurons only switch 10
to 100 times per second.  Transistors can switch
billions of times per second, so maybe far, far
fewer of them are necessary.  A biological brain
has to be fault tolerant, since thousands of
neurons die everyday.  Transistors are vastly more
reliable than individual neurons.  A biological
brain has to self-assemble, heal itself, and do a
lot of other basic biological activities that a
computer does not have to worry about.

Also, biological brains are limited to what
evolution can produce.  Which means it can creep
from one local maxima design to the next, but
cannot make dramatic changes.  That may work okay
for a neural network, but you can't produce a
good symbolic algorithm by random evolution
"jiggling".  Engineers can produce things
that would never emerge from evolution.


To the contrary, I don't know of anything that
an artificial neural network can do more efficiently
than a symbolic algorithm.

It seems to me that the main argument for neural
networks is "AI is really hard, so rather than
actually trying to understand intelligence, let's
just throw a bunch of artifical neurons into a
network, and see if it just magically emerges."

I dont think there is a symbolic solution to the problem. Yes you can
add endless pieces of information to your system, yes you can see
excactly how its work. BUT when you add line after line to it, you
expect the system to "magically" take over and become concious? Because
that is what it takes for a symbolic system to become 'mister data'. I
think neural networks are the answer, but not the traditional kind we
are all familiar of.

Any neural network can be simulated using a
symbolic algorithm.

Any symbolic algorithm can be simulated by
arranging artifical neurons into NAND gates
and building a turing machine out of them.

So anything a symbolic algorithm can do, a
neural network can also do, and vice versa.

It is simply a question of which approach
is more computationally efficient, and which
is easier to implement.  My opinion is that
a symbolic algorithm is always more efficient,
and for most problems, is easier to implement
as well.


"Conciousness" is ill-defined and subjective.
Besides, the goal is "intelligence", not
conciousness.  I do not expect it to magically
emerge.  I expect it to incremently improve.