The first question seems to be: Is context necessary to establish meaning?
I think it is. It then seems reasonable (if not demonstrated) that the
"level" of intelligence (i.e., meaningful thought related to some arbitrary
hierarchy?) exists as a function of the context, which obviously involves
sensory interactions and experiences with that context.
The second question deals with the nature of the context. I speculated on
this idea earlier with regard to the experience of animals, e.g., the
"cleaver Hans" phenomena (horse that supposedly could add). While the horse
was obviously "intelligent," the context of its intelligence was its
relationship and experience with the owner and his moods; not as it turned
out, with mathematics. The intelligent horse was not experiencing the
problem on that level (the arbitrary hierarchy here being the emotional
relationship with the owner vs. the abstract realm of mathematics).
Walter's question is closely related I think. By suggesting that an AI
system must have similar senses to what we wish it to emulate, he is
suggesting that it must be able to experience the same context, which is
necessary to establish similar meaning. Even a human with normally clear
thoughts can become very confused and irrational (unintelligent?) if placed
in a very different and unfamiliar context, i.e., one to which his/her
ideas have not been related. Similarly, historical events we may read about
seem empty until someone artfully fills in the context, through a vivid
imagination or perhaps a well done movie.
This raises a third question: How much of the context is necessary to make
the information meaningful? The example above refers to filling in the
context through imagination or a movie. These are vicarious experiences
that do not require limbs. However does one need to have had prior
experience with limbs to understand the vicarious experiences? I think we
may be dealing with a contextual gradient here, where it is a matter of
degree how well the context is established by various means. Could, for
example, simulated or virtual limbs be adequate to emulate the necessary
experience?
So, as Cliff's question implies, exploring this with artificial systems may
teach us about humans. I think it will teach us about the importance of
context (what I think of as experience, in the case of real organisms).
At 11:12 AM 6/2/98 -0600, you wrote:
>Classically, the two camps of AI are (A) those wishing to make
>machines which do smart things, and (B) those wishing to make machines
>which are smart like people are smart.
>
>It's an interesting suggestion that the latter goal cannot be
>accomplished without making a WHOLE articificial human, or at least a
>system whose inputs/output sensor/affector sub-systems are similar or
>equivalent (how? structurally? functionally? informationally?
>semiotically?) to a real human.
>
>Personally, I don't think that's where the interesting issues
>lie. Presumably the goal of (B) is to understand better how humans
>ACTUALLY think. If not, there's a lot cheaper and easier (and more
>fun!) way of making all the REAL humans you could ever want.
>
>Then you could argue that the path to making an artificial human
>should naturally go back through making an artificial dog, insect,
>worm, plant, protozoa. In principle, each of these would also require
>system-appropriate sensor/affector grounding. And, in fact, there's a
>large research program going exactly in that direction (see, at least,
>www.c3.lanl.gov/~joslyn/ISAS98).
>
>O--------------------------------------------------------------------------
-->
>| Cliff Joslyn, Member of the Technical Staff (Cybernetician at Large)
>| Computer Research Group (CIC-3), Los Alamos National Laboratory
>| Mail Stop B265, Los Alamos National Laboratory, Los Alamos, NM 87545 USA
>| joslyn@lanl.gov www.c3.lanl.gov/~joslyn (505) 667-9096
>V All the world is biscuit-shaped. . .
>
>
-----------------------------------------------
John J. Kineman, Physical Scientist/Ecologist
National Geophysical Data Center
325 Broadway E/GC1 (3100 Marine St. Rm: A-152)
Boulder, Colorado 80303 USA
(303) 497-6900 (phone)
(303) 497-6513 (fax)
jjk@ngdc.noaa.gov (email)
(303) 497-6513 (fax)