Re: Strong AI Thesis (No Chinese room, I promise)


Don Geddis wrote:
I wrote:
In other words, there's tons of evidence that the education of young
human minds is highly planned at a genetic basis, and far from just a
random self-organization due to interaction with the environment.

curt@xxxxxxxx (Curt Welch) wrote on 21 Mar 2006 07:4:
Well, all the network ideas I've played with have to build layers of
abstraction, on layer at a time. The second layer can't form until the
first one has formed. The third can't form until the second has formed,
etc.
This type of slow growth in the network can easily explain why child
development moves along in very predictable phases.

You've missed part of the key. In humans, many of these stages are
age-dependent. Whereas if your theory were correct, it would depend instead
on how much exposure the child had to the relevant topic area.

You also haven't explained the remarkably similar brain structures in each
human. Why always two hemispheres, not sometimes just one and sometimes four?
Why the visual center ALWAYS at the back? Why the speech center always in the
same place? Why the primitive brain (emotions, desires) always in the same
place, at the bottom/base?

There's a large amount of evidence that suggests much about adult brain
structure is already encoded in our DNA, and only a portion of it is learned
through interaction with the environment.

It's just the beginning version of what happens all through our lives. We
can't learn calculus until you have first leaned algebra. You can't learn
algebra until you learn to add and subtract. You can't learn to add and
subtract, until you learn to count.

But you can do this at any time. You can learn algebra(->calculus) starting
at 14, or 18, or 25, and it takes about the same amount of time and results in
roughly the same outcome. So sure, there are certainly examples like you
suggest, of learning random new information. The brain definitely can handle
such training.

But you've failed to appreciate how much of what we learn is NOT like this,
cannot be done (equally well) at any time.

So, the idea that learning proceeds in stages is not at all inconsistent
with the idea of a generic learning system at work.

Not by itself, correct. But the details of how humans actually do learn _is_
at odds with (only) a "generic learning system".

There's no doubt that the brain has a lot of specialized adaptations to
make vision work better, and language work better, etc. But what I believe
is that the cortex is first and foremost, a very generic, data independent,
reinforcement learning network.

OK, possible.

This would imply we need to first solve the data independent problem of
learning, and once we have a strong data independent learning system, we
can make adaptations and modifications, to maximize the results for
different aspects of intelligence, from language, to vision, to motion
control, to spatial awareness, to whatever.

This doesn't follow at all. A few posts ago I agreed that a general learning
system was necessary -- but not sufficient -- for intelligence. Surely it's
a required component. That doesn't mean AI scientists will necessarily make
the most progress working on that part first (especially in isolation from
all the other components).

-- Don
_______________________________________________________________________________
Don Geddis http://don.geddis.org/ don@xxxxxxxxxx
A fool does not learn from his mistakes. A normal man does learn from his
mistakes. But the exceptional man learns from the mistakes of others.

Please excuse me if I tack this on, I'm reading
by date, and I was interested in Curts views,

Theres an article in this weeks New Scientist about
the Piraha (I was trying to remember their name)
Its called 'Lost for words'
The Piraha have no myths, no numbers or
colors and few words for past and present. They
use seven consonants and three vowels.Each sentence
refers to one event, and the language has the simplest
known set of pronouns with no plurals, not even a
distinction between 'I' and 'We'. Theres no 'all'
'each','every','most' and 'some'. Apparently the
language is fiendishly difficult to learn.

(Now this is what I call a 'natural language')
....'The patterns of stress and intonation are highly
complex'...'many words have a variety of meaning
depending on the inflection'.

Although the article states quite clearly that they
don't have myths, they do believe in a spirit world,
although a spirit may for example be a panther.

I don't really understand why people separate meaning
from myth, it sometimes seems like a peculiarly elderly
trait!. To me, that which generates realism and life to
the object is encompassed in both its representational
quality/value + a potentiality to act autonomously

If you use a pictographic system for instance,
a kettle is an object + 'breaths' steam. it is
'spirited' somehow, either autonomously or by forces
yet to be discovered.

The article continued that the language lacks words
for numbers'-and indeed any abstract concept that
involves quantification, such as colour and time-
because of the cultures emphasis on referring only
to immediate, personal experience'...'not one of
them had learned to count to 10 or even to add 1+1'
(sounds like my kinda peoples!)

I thought that they would have a rudimentary class
for colours, red, green, beige or yellows, but
maybe even these are subjective.

-----------

I found a post by David Longley regarding verbal behaviours,

http://groups.google.com/group/comp.ai.philosophy/msg/6414459bf6b604dc?dmode=source&hl=en


'Traditional Formulations' an excerpt from B.F.Skinner

A science of behavior does not arrive at this special field to
find it unoccupied. Elaborate systems of terms describing verbal
behavior have been developed. The lay vocabulary abounds in them.
Clas sical rhetoric, grammar, logic, scientific methodology,
linguistics, literary criticism, speech pathology, semantics, and
many other dis ciplines have contributed technical terms and
principles. In general, however, the subject here at issue has not
been clearly identified, nor have appropriate methods for studying
it been devised. I,inguistics, for example, has recorded and
analyzed speech sounds and semantic and syntactical practices, but
comparisons of different languages and the tracing of historical
changes have taken precedence over the study of the individual
speaker. Logic, mathematics, and scientific method ology have
recognized the limitations which linguistic practices im pose on
human thought,

I repeat the last sentence;
Logic, mathematics, and scientific method ology have
recognized the limitations which linguistic practices im pose on
human thought,

a little later Ogden'n'Richards;
the development of the practical art of
writing has provided a ready-made system of notation for reporting
verbal behavior which is more convenient and precise than any
available in the nonverbal field.
(...)
One unfortunate consequence is the belief that speech has an in
dependent existence apart from the behavior of the speaker. Words
are regarded as tools or instruments, analogous to the tokens,
counters, or signal flags sometimes employed for verbal purposes.

Neither adequately separate the difference between the written word
in culture from natural language.


Just a thought,

N.

.

Relevant Pages

  • Re: Gradual Learning, not Reinforcement Learning
    ... You seem to believe that "learning" means acquiring behaviours the system has never exhibited before. ... You use learning HORSE as an example, and posit HOXEL as an intermediate stage, with HORSE as an "ideal" to be achieved. ... Firstly, the system must be capable of "producing language", else it cannot learn the semantic difference between HORSE and HOTEL, both of which are easy consequences of HOXEL. ...
    (comp.ai.philosophy)
  • The Gaelic Society of Moscow speaks out!
    ... TEN REASONS FOR LEARNING IT ... ©Gaelic Society of Moscow,2003 ... language if you have few opportunities to use it in addition, ...
    (soc.culture.scottish)
  • Re: The Gaelic Society of Moscow speaks out!
    ... TEN REASONS FOR LEARNING IT ... ©Gaelic Society of Moscow,2003 ... language if you have few opportunities to use it in addition, ...
    (soc.culture.scottish)
  • Re: Opinions on intro lisp books
    ... But Lisp is a little different, ... Some languages support one style of programming better than they ... Even if that weren't the case, I'm not sure that a language being a ... I don't believe that learning to program in CL requires more theory ...
    (comp.lang.lisp)
  • RE: C# book for 16 yo?
    ... First I would suggest first learning Visual Studio.NET. ... Then I suggest 2 books to actually learn C# - Microsoft has a book called ... important to quality programming. ... language that combines all the best features of previous languages and is ...
    (microsoft.public.dotnet.languages.csharp)