How might the See agent work? The answer to this question is much more complicated than introspection would reveal -- you have to deal with object permanence issues and hidden surfaces, for example.
Minsky's approach to this problem is to use frame-arrays, in which different views of the details into the same structure. (NB: Cool thing here is that the object-centered "coordinate system" really becomes a concatenation of different viewer-centered perspectives!). "When you represent a thing's details with a frame-array, you can continue to move around yet 'keep in mind' all that you've observed from those different viewpoints, even though you've never seen them all at once. This gives us the wonderful ability to conceive of all of an object's different views as aspects of a single thing.
NB: From Pylyshyn's perspective, the role of a finst would be to provide an index from a retinal cluster to terminal nodes of the frame array interpretation.