A structured description by making it more elaborate

This preview shows page 24 - 26 out of 38 pages.

a structured description by making it more elaborate and detailed. The advantages of thinking of this as the elaboration of a description, rather than of the image being more pictorial (apart from the vague- ness of the notion of "degree of pictorialness"), are that (a) this interpretation gives recognition to the fact that the elaboration is done within the constraints of available concepts rather than by the addition of arbitrary pictorial fragments; (b) "reading off" becomes a well-defined symbol-matching operation rather than involving all of the perceptual apparatus driven by intentions; ( c) no matter how much elaboration of detail is carried out, there will always be an arbitrarily large amount of indeterminateness in the resulting rep- resentation (it will always fail to be determinate with respect to
IMAGERY AND ARTIFICIAL INTELLIGENCE 43 some aspects which are not only determinate in the scene but which no picture would leave uncommitted), and furthermore, as noted in section 2 ( 1), the representation is not homogeneous in the amount of determinateness of various aspects; (d) this interpretation dis- courages the view, invariably associated with the term "image," that there exists an object that is interpreted the way a scene is inter- preted (i.e., visually), that has a stable simultaneous existence so that it can be scanned perceptually for new readings, and that in- herits certain intrinsic properties from the material medium in which it is embedded (e.g., rigidityundervarioustransformations-seethe discussion below). Replacing "images" with "images under descriptions," as Fodor does, frees the term from many of the philosophical problems that plagued it in the past. The trouble with this move is precisely the problem of making clear the sense in which images under descrip- tions are to be distinguished from descriptions. Fodor (p. 190) puts it this way: "Images under description share their nondiscursiveness with images tout court. What they share with descriptions is that they needn't look much like what they represent." Thus discursive- ness seems to be the crucial property. But, as we have seen, symbol structures are not discursive in the sense that sentences are-i.e, they need not be read in a prescribed order. The order of "scanning" is determined by the accessing algorithm and makes use of the relations that are the access paths of the structured description, just as a visual scan of an image would presumably be determined by the intentions of the perceiver together with something like peripheral vision. This amounts to saying that we have yet to see a viable distinction among images, images under descriptions, and structured descriptions when any of these is embedded within a representational system-i.e., when paired with the appropriate semantic interpretation function.

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture