YKY:
MT:> Jeez, there is NO concept that is not dependent on context. There is NO
concept that is not infinitely fuzzy and open-ended in itself, period - which
is the principal reason why language is and has to be grounded (although that
needs demonstration).
YKY:My current approach is to use fuzzy rules to model these concepts. In
some cases it seems to work but in other cases it seems problematic...
For example I can give a definition of the concept "chair":
chair(X) :-
X has leg #1,
X has leg #2,
X has leg #3,
X has leg #4,
X has a horizontal seat area,
X has a vertical back area,
leg #1 is connected to seat at position #1,
etc,
etc....
But what if a chair has one leg missing? Using fuzzy logic (fuzzy AND), the
missing leg will result in a fuzzy value close to 0, which is not quite right.
Probabilistic logic is also inappropriate. I know *every* time that a chair
missing a leg is "somewhat" a chair -- there is no probability involved here.
REPLY:
YKY,
We can kill a whole flock with one stone here - both the infinite
open-endedness of concepts AND, as a consequence, why any
General-Intelligence-level reasoning *must* be grounded in sensory images and
imagination.
You couldn't have picked a more archetypal concept. "Chair", from my indirect
reading, is the concept Plato picked to illustrate his idea that eternal forms
must underlie concepts and words and the objects to which they refer. He
suffered from the illusion that all AI/AGI-ers suffer from - & that literate
culture - which dates from the alphabetic Greeks to approx. 2008, now that
multimediate culture is replacing it - also suffers from. Namely that words and
other symbols refer to real structures or forms of objects. We can look at
millions of different chairs, and yet instantly recognize that they all fit the
word "chair" Therefore there is a) an "essence of chair" and b) the word
"c-ha-i-r" somehow captures that essence. and c) that essence can be defined
with more words. ( I doubt that this illusion would have been possible in
pre-alphabetic culture, when words were rendered in more or less *pictographic*
rather than alphabetic form and therefore did not have a standardised, uniform
form).
AGI-ers certainly believe that words - and those essences of objects - can be
successfully defined with other words, and therefore that purely verbal/
symbolic, or ungrounded reasoning is possible. Of course, they are all aware,
YKY and esp. Ben included, that this may be a very complex business. In Ben's
case, that it may take a massive CYC-scale operation to fully define any word.
But it can be done. Oh yes, it can be done. He has no doubt of that.
Well, let's see. ("See" being the operative word). Let's look at some chairs.
You will note in the following two chair picture sets - which BTW I consider (&
please disagree) an awesome a) set of pictures b) set of human creativity and
c) examples of the mind's powers of categorization, that you can recognize
almost every example as a *chair* immediately - although you will probably
question a few.
But I defy you to define any single attribute of "chair" that they all share,
(they certainly don't all have "legs").
1. CHAIR SLIDES
http://www.mediafire.com/?mwm5ivjmmcd
2. SET OF CHAIRS
http://www.mediafire.com/imageview.php?quickkey=vmj2jkptlcn&thumb=4
I further defy you to define any attribute of chair, period, including "seat"
or "something to sit on" that an inventor has not already, or could not,
circumvent - and still produce a recognizable "chair."
By extension, you will not be able to definitively define any concept, period
- "table," "cow". "human," let alone prepositions like "in" (Ben's word in
his essay), "through", or "over." - let alone mildly to massively complex
processes like "push,"handle","conversation," "sex", "evolution." Nor will you
be able to definitively define any *individual* - *Ben Goertzel," "Pei Wang,"
"Madonna."
All these concepts can be defined in infinitely open-ended ways because the
classes of object, both artificial and natural, that they refer to are
themselves infinitely *open-form* and, usually, evolving - constrained by some
parameters, perhaps, but not limited. .
How then did you come to form the concept of "chair", "table," "cat", "dog"
etc. or "Ben G" from these open sets of forms? Not verbally or symbolically.
You did it the same way your evolutionary parents did - the way all those apes,
bears, snakes, birds etc recognized each other - who despite having no words
have plenty of general intelligence. You did it by visual and other image
processing. Processes which do not and cannot result in a single coherent image
or template, but result rather in a *flexible set* of images. What is your
imagistic concept of "amoeba"? There is not and cannot be a single one for such
an open-form, continuously form-changing object. It has to be rendered -
ideally/normatively as well as psychologically - by a set of flexible images.
Just as continuous form-changing processes like "sex" and "foreplay" have to be
rendered and "defined" by flexible sets of images.
Words are just useful labels stuck on top of this flexible image processing -
useful *precisely* because they serve to encapsulate *several* images with
*one,* fixed, "magical" label.
So if you are going to be able to handle concepts and recognize whether
either an actual object or a verbal description of an object constitutes a
proper example of that class of object - if you're say going to respond to
"well, a Hockney chair is too fragmented in my mind to be a true chair," - or
"Clintonian sex is not really sex," you will have to consult the image(s) of
that object, either in actuality or your mind's eye.
You will have to *ground* your thought processing - no verbal set of
definitions can satisfactorily encompass those infinite sets of forms.
Yes,sometimes, purely verbal processing will work. Computer language
translation IS useful. But it will always be far too flawed as a whole to
constitute general intelligence - because it will keep running into exceptions
to its verbal definitions.
[And bear in mind that we have only talked so far atomistically about
interpreting *individual* concepts. When we start to talk "molecularly" about
understanding combinations of two or more concepts - deciding whether a "cow"
or "cat" or "tractor" can or can't "sit" "on" or "have sex with" a "table,"
"chair".etc .. the idea of producing any kind of symbolic formula that could
encompass the "infinitely squared" possibilities is stratospherically absurd.)
-------------------------------------------
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244&id_secret=108809214-a0d121
Powered by Listbox: http://www.listbox.com
