| | My understanding of Rand's position is that perception indicates "that" something exists, not "what" that thing is. "What" it is, is a conceptual issue, which is only handled implicitly on the perceptual level. Perceptually, we might recognize that two things are different, and, based on prior experience, implicitly treat them as entities of a particular sort.
A camera using a frozen neural net processor for auto-focus is a perceptual system of that sort. If you only wanted to focus on red objects, then you could use a filter to force the auto-focus to do just that. Or, perhaps you might want to focus only on a particular object moving against a background. Many newer cameras have that option built in as well.
The camera, however, has not a clue as to the actual nature of what it is perceiving. It is only organizing the data in such a way as to yield a particular, pre-specified result, based upon the sensory input. This is why no paint program, with its own library of techniques, frozen from experience in the form of program functions - e.g., directional sharpen - can reliably and precisely separate an object from its background.
The areas of ambiguity, where by chance adjacent pixels will be too close in color value, will always defeat it. (I hate to think of how many precious hours I've spent just cutting images from the backgrounds in COREL Draw.)
One of the areas that Rand did not discuss in writing, so far as I can recall, is just how it is that even on the perceptual level, living creatures manage this trick of picking objects from backgrounds. I brought this issue up in the early '70's among my objectivist friends of the time, some of whom were quite brilliant in terms of Rand's epistemology, and was told that it was a non-issue, but never with a satisfactory explanation.
In fact, for a living creature objects are an integration of multiple sensations based upon the identity of the actual object being perceived. This integration is not something built into the system from scratch, genetically programmed, but follows from an internal integration of normative processing - eg., value processing (pain or the expectation of pain, or pleasure, etc.) - with both individual objects and classes of objects. A large part of the identity of an object is its value relationship to the perceiver, something that I think Rand barely touches upon.
In theory, we could use all kinds of perceptual methods of classification - as in all red objects, or all square objects. In fact, we use those that allow us to function effectively as living things. We see that which we have pre-judged as important, based on our history, and ignore the infinity of other ways to see it and the universe of objects that are irrelevant to our values.
We can pick the object from the background because we already know what the object probably is, or at least what its general characteristics likely are. Thus, we know that regardless of color similarities in two overlapping faces in a crowd photo, that they are two people, not one, and that one person's face ends here and not on the far side of the other's nose. No mere software program could do that reliably across contexts.
A crow can recognize a particular person, even at a distance, even when they are wearing completely different clothes, a hat, a beard, etc. It can't abstract the process by which it knows that you are the crow molester who rescued the baby crow that decided to play in the traffic circle, but it will never ever forget you, because the memory of you is integrated with the normative evaluation "CROW MOLESTOR!!!" (which it will broadcast to the entire world every time you step outside for the next decade or so).
At the same time, the crow is capable of formulating implicit rules based on experience with regard to humans in general. As in, any human is to be treated as dangerous and suspect. However, with time and patience on the part of the human, it is possible for the crow to make exceptions to the rule in that regard, as well.
My understanding is that Rand included the concept of "Ist level concepts" to bridge from perceptions to explicit concepts. However, my impression is that she was using a passive model for the formation of percepts. I'm reasonably convinced that this is impossible. As in the camera model, you can only push a passive, non-normative model so far. It requires pre-judgment, and a limiting function based on values (to whom and for what) to reduce a buzzing, whirling set of sensations to a set of finite, concrete entities to be dealt with.
If you try to define perception as just the passive organization of sensations - an integration "based upon" the spacial temporal location of the source of the sensation - then you are not merely forced to bring in a new level - first level concepts - to deal with issues such as those I bring up above, but also you are faced with the fact that the "based on" function has been left undefined, leaving perception as a floating concept.
I think that this area of Rand's epistemology needs further explication and refinement at minimum.
(Edited by Phil Osborn on 8/16, 3:32pm)
|
|
