rlw@wxlvax.UUCP (Richard L. Wexelblat) (05/22/84)
It has been my experience that whenever many people misinterpret me, it is due to my unclarity (if that's a word) in making my statement. This appears to be what happened with my original posting on human perception vs computer or robotic perception. Therefore, rather than trying to reply to all the messages that appeared on the net and in my mailbox, let me try a new, longer posting that will hopefully clarify the question that I have. "Let us consider some cases of misperception... Take for example a "mild" commonplace case of misperception. Suppose that I see a certain object as having a smooth surface, and I proceed to walk toward it. As I approach it, I come to realize visually (and it is, in fact, true) that its surface is actually pitted and rough rather than smooth. A more "severe" case of misperception is the following. Suppose that, while touring through the grounds of a Hollywood movie studio, I approach what, at first, I take to be a tree. As I come near to it, I suddenly realize that what I have been approaching is, in fact, not a tree at all but a cleverly constructed stage prop. In each case I have a perceptual experience of an object at the end of which I "go back" on an earlier attribution. Of present significance is the fact that in each case, although I do "go back" on an earlier attribution, I continually *experience* it "as" one and the same. For, I would not have experienced myself now as having made a perceptual *mistake about an object* unless I experience the object now as being THE VERY SAME object I experienced earlier." [This passage is from Dr. Miller's recent book: Miller, Izchak. "Husserl: Perception and Temporal Awareness" MIT Press, c. 1984. It is quoted from page 64, by permission of the author.] So, let me re-pose my original question: As I understand it, issues of perception in AI today are taken to be issues of feature-recognition. But since no set of features (including spatial and temporal ones) can ever possibly uniquely identify an object across time, it seems to me (us) that this approach is a priori doomed to failure. Feature recognition cannot be the way to accurately simulating/reproducing human perception. Now, since I (we) are novices in this field, I want to open the question up to those more knowledgeable. Why are AI/perception people barking up the wrong tree? Or, are they? (One more note: PLEASE remember to put "For Alan" in the headers of mail messages you send me. ITT Corp is kind enough to allow me the use of my father's account, but he doesn't need to sift through all my mail.) --Alan Wexelblat (for himself and Izchak Miller) (Currently appearing at: ..decvax!ittvax!wxlvax!rlw)
jim@ism780.UUCP (05/24/84)
#R:wxlvax:-28000:ism780:20200003:000:519 ism780!jim May 23 11:52:00 1984 > Why are AI/perception people barking up the wrong tree? Or, > are they? While a possible long-term goal of AI research is to simulate/replicate human intelligent behavior, most AI goals are toward machines which perform specific difficult ("intelligent") functions. While it is now well understood that the fundamental key to general intelligence is representation of knowledge, practical short term goals of object recognition can be met through feature recognition. -- Jim Balter, INTERACTIVE Systems (ima!jim)