Complexity is a Hierarchy of Abstractions

Self portrait by František Kupka

One of the technologies that is rapidly emerging is neural networks, but scientists are just beginning to learn how they work.

"Neural networks aim to mimic the human brain — and one way to think about the brain is that it works by accreting smaller abstractions into larger ones. Complexity of thought, in this view, is then measured by the range of smaller abstractions you can draw on, and the number of times you can combine lower-level abstractions into higher-level abstractions — like the way we learn to distinguish dogs from birds."

This abstraction architecture is useful for training a computer network to recognize objects, such as differentiating one animal from another, or for reading handwriting.

Self portrait by Chuck Close

But the basic concept of complexity as a hierarchy of abstractions can be taken in another way by artists. Often a realistic painting is made up of abstract strokes at a micro level. They might be the little abstract pixel tiles of Chuck Close or the blobby dots of John Berkey.

We might also connect this idea with image processing in the human brain. As our visual system begins to interpret an image, it does so first by perceiving brightness differences across boundaries, and broadly abstract shapes and colors.

As the image is decoded from the occipital cortex to other cortical areas of the brain, those abstractions are assembled into a higher level interpretation of reality that takes into account such things as the passage of time, cause and effect, the relationships between objects, and the sense of story.
----
Previous post: Abstraction vs. Realism: Thoughts While Painting the Maine Coast
Quote from Quanta Essay "Foundations Built for a General Theory of Neural Networks"