Maybe so, but we do know that having gained spatial knowledge of a visual
field, this is pattern matched against internal templates of a more complex
nature. As in, hey, that is not just an oval, that is a face. Hey, that is
not just any face, that is Bob's face. How this pattern matching takes
place seems to include general descriptive sort of symbolic information in
combination with visual/spatial snipets that uniquely identify something.

I think that an "understanding engine" is a faulty idea. In so much that I
view understanding and knowing to be subtle form of emotion. For example, I
identified the face of Bob. That face means (identifies) Bob. That I
understand and know this is an emotional sense of comfort and confidence
tied to a lack of contra indicators to this "truth". I know Bob well,
nobody else looks like him, I am close enough for a good look, Bob's hair
grew, but that is definitely his hair. Is that my understanding engine at
work? Perhaps, but there must also be very similar engines none of which
are fundamentally different from the other processes of refined
identification and differentiation. One could just as easily argue that the
understanding occurs when there is no longer a state of anxiety assosciated
with trying to figure out if that is a face and whose face it is and are
they a threat, allie, or unknown person. This might however lead to a
higher state of anxiety in two out of the three cases, but there was a
fleeting "satisfaction" in having understood whose face that was.

Any comments.

Dave Horsman

Hi Lester,

It seems to me that at the point where something is visually percieved, it
starts a process where the summarised and symbolic information is on a
seperate "channel" from the pure sensory data. I would argue that cognitive
inference is taking place from that point forward. This information is
closely linked, just as it is when we ?relive? and think about a visual
memory or fantasy, or to plan jump across a puddle. I understand that the
internal modelling of self and the external world are generally accepted as
something that minds do. Could we also say that there are close links
between percieving someone throwing a ball, remembering having done so, and
the trial runs that precede the immenent intent to throw a ball yourself. I
think these processes are pertenant to the issue.

My point is that cognitive inference would not have to generate diffenences
amoung the differences. That may well be the case, and is a good example of
cognitive strategy, but I don't quite follow why that must be the case. At
the point where the seeing becomes identification I think more robust
cognitive processes kick in, some of which may trigger changes in the focus
of visual perception. In general terms, most mental processes are
overlapped and feed back to varring degrees, and very few exist in an
entirely isolated or linear form.

Please excuse if my confusion stems from a lack of understanding of your
jargon. I doubt I have required knowledge to comment, but is perceptual
inference defined by differences? It seems more likely that a process of
similarity and differences are involved. Neurologically, I think both are
involved. Perceptually, a concurrent cognitive process of catagorisation
and differentiation seem to be in effect and closely linked to the
perceptual tasks being performed. In short, I doubt there is a clear line
between the two, and that in particular, perception includes a great deal of
cognition.

On another vaguely related matter, I believe that your viewpoint on machine
intelligence threads being frustrated with philosophy issues is likely true.
However, keeping in mind I am neither a cognitive scientist or philosopher,
I have a few comments. Specifically, philosophical truths should not bog
down cognitive scientists too much. In my own work,

1) I allow that most of what we know is accepted knowledge via authority.
Be that parental, expert opinion, popular media, whatever.
2) Other knowledge is very similar in nature, but can be categorised as "by
definition".
3) Another huge source of data, memories, are loosely accepted, based on the
authority of their being "my" memories.
4) That direct knowledge, through the senses are accepted, after validation,
as "true", without great difficulty.

Now re item 4, there is of course a difference between the reality of an
object and our perception of it. Further, I accept that there is no
utlimate proof of knowing, no objective truth, and that objective reality is
seperate and unprovable. That truth it is ultimately an act of "faith".
This is no different than the basic truths assumed in scientific fields and
mathematics. Even then, such systems are inherantly flawed, yet I doubt
that Godel convinced anyone to stop using a calculator.

In machine intelligence, the nature of how we "know" something is a reality
and must be carefully defined and accomodated in storing data to arrive at
meaning. To continue, perception followed by understanding are dependent as
much on finding a good match as finding very few differences, or that this
is basically a symantical argument or more accurately one of point of
reference. The glass is half full, not half empty. Those objects have few
diffeneces or a lot of similarity. Basically, these are opposite poles of
the same strategy and as such not an important distiction, the important
issue being that a cognitive scientist ignore them at their peril.

Now before you jump on "meaning"...I would simply state that understanding
and meaning are "feelings" we get when an item is represented by a rich
internal database of examples, generalizations and associations in
combination with an acceptance of the source authority or definition.

A few other important points come to mind. One, this email may be a flawed
message in that I lack the proper "jargon" associated with your field of
interest. In order to communicate effectively, cumulative experience in
philosophy and cognitive science lead to a "jargon" where words and ideas
can be discussed. There may be important diffenences between the formal
defenitions of cognitive and perceptual inference that could be at odds with
a layman's interpretation of the terms; this of course aside from any
theoretical debate on the finer points. On the one hand, I don't mean to be
an apologist for my own posting, on the other hand even if misguided, my
posting raises a few issues that touch not only on language but that our
internal symbology may also have an objective truth distinct from either
common usage or varrying usage within a tightly constrained jargon.

A second point, sense data is constantly suspect and under both review and
influence of cognitive processes. We frequent "see" errors and distortions
in our visual field and know it. We often identify an object, "see" the
thing we identified, and realise it is something else through some
exception, and then "see" what it really is. Case in point, how many
hunters are shot every year by other hunters vs how many only come close to
being shot as a buck? I don't know but I bet very few are tied to a roof
rack before the idiot realises, "hey....thats not a deer..." Another case in
point, as far as I know, many schizophenics are quite acutely aware that
they are having a halucination at the time it is occuring. Another, "am I
really seeing that....I can't believe it!" and the many subtle mental
acitivities that occur with that reaction. Yes perception is often
different from an existant undeniable objective reality, atleast undeniable
from the point where we begin to accept our perceptions as real.

Anyway, enough of this foolish posturing old sweat, let us cut to the chase.
I note you have made a number of postings to other AI threads, attempting to
draw attention to, and to clarify, various issues of philosphy without
coming right out and saying so. A noble cause indeed, but it seems your
efforts are often futile, and result in only more soap box dialogue, rather
than any true insight on the part of your readers. When you are forced to
defend yourself from personal attack, you retreat into dry wit, and yet I
ask myself....why....

Further, many of these philosophical issues you toss out as obstacles or
counter points, where I doubt you view them as such yourself, but are merely
baiting the cognitive processes of those reading the thread. Why, to be
perfectly frank old man, I suspect that this very posting has some alterior
motive behind it...that you don't quite believe this contention
yourself...but are up to something. The question is not only what, but how
will we know that we know that what we know is knowable, with regards to
your intentions... And how good sir do you explain the prolithic volume of
your correspondance...suspicious indeed.

I believe that cognition, and in particular understanding, are the product
of biological processes of perception and neurological processes related to
but partially independent of perception. Further, I would assert that all
knowledge is a by product of biological brain as much as mind, and therefore
carries all the associated biological baggage. Spefically, that most of
what we know has an emotional component as much as it has sensory (direct or
recalled) and coginitive components. Does knowing something not involve
such emotional components (and their polar extremes) such as confidence,
trust (in authorities) and faith?

I believe that many of these issues that are contentious, are only so
because we attempt to view thinking as somehow linier and issolated, rather
than the product of the state of multi-purpose state machine such as the
mind, where no one broad area is completed divorced from another (ie
perception and truth, sujective and objective). Certainly, it is well known
that the history of science is a history of fashions, trends, social and
cultural dynamics, and so theories certainly have a similar content. You
should reconsider the usefulness of an interest in the history of science
with an eye to how this relates to what we "know" and just how emotionally
attached we can become to a system of beliefs, without having even looked at
the competing view points. This goes on at both a cultural and personal
level and I believe is reflected in many of the threads you have contributed
to.

Why I could easily prove the point, were I willing to resort to such gutter
tactics as calling you a "closet behavioralist". Could this news group be
pre-occupied with the differences rather than similarities between the
various competing view points and their emotial attachment to their own
school of belief... or did a behavioralist run over your cat and explain it
away as an inevitability due to previous behavior and circumstantial
environmental contingencies. Almost all of our knowledge has an
emotional component, and that "knowing" and "understanding" are in
themselves, emotive in nature rather than subjective...and curse any half
wit that dares to disagree with me!!!!

OK...I'll stop now, but I doubt you believe your comment:

The suggestion then is that perceptual inference is drawn among
properties in the visual field but that cognitive inference is drawn
between properties in different perceptual fields.

It seems to me much more likely that perceptual inference is drawn amoung
properties in the visual field and that cognitive inference is drawn between
the symbolic identifiable results of perceptive inference, in the same and
different perceptual fields, cross referenced against our existing symbolic
knowledge and stored perceptive knowledge (templates and examples) given a
series of perceptual events in an identified context. Further, I would
suggest that the brain uses this symbolic information as feed back to direct
the focus of the perceptive process, including both zooming in on details
and zooming out to broader components of the perceived field. Although the
differences between the differences would seem to count as cognitive, I
would suspect that direct perceptual referrents would still be involved
frequently, or atleast not a defining attribute.

Now would you mind starting by defining what a direct perceptual referrent
is...

Probably just whatever lies in a given perceptual field. There might

I think both induction and deduction could be involved in the visual
engine - the induction could be used in designing such a visual engine
while deductive logic could be used by this visual engine at runtime
to generate information.

But this is only for the lower-level spatial knowledge. For
higher-level knowledge such as face recognition mentioned by David,
there must be some more complicated logic involved than just the
straigtforward induction and deduction implemented at the hardware of
the visual engine.