The bark of a tree, for instance, is composed of a multitude of different local image patterns, but the computational model can determine that all these local images represent bark and are all part of the same tree, as well as determining that those same patches are not part of a bush in the foreground or the hill behind it.
Whether the theory is exactly correct or not, in thinking about it, I realize that we take understanding and perception quite a bit for granted. Our comprehension of the world is supported by complex internal processes that, like all characteristics of living systems, can vary from one “bundle” (human being) to another. We evolve diverse assumptions about the world based on our diverse interpretations. When we communicate, we generally think we share assumptions about reality, but our assumptions are only similar, and less similar where we have differences that will color perception (culture, language, internal process). Communication is far more difficult than we imagine. How do you communicate more of your substructure, so that the things you write or say are understood because your specific neural context is understood? I studied Irish author James Joyce quite a bit when I was in college, and I recall that, with Finnegans Wake, he created a book that could only be understood if you read what he’d read and could get into his head. He understood that a reader’s experience of a book will differ from the experience the author might have intended, depending on the reader’s background, perception, interpretation, culture… so each reader’s experience of a book is different, and if you’re trying to convey your unique sense of the world as you sensed it, you have to demand that the reader crack your code by following a path similar to the path you followed in reading, thinking, writing.