It is a fact that there are specific neurons in the retina that take in the raw input from the eye and detect specific features such as red, blue, lines,edges, and more.
These features (or objects) are sent further up in the processing center of the mind to form more complex objects like buckets, cars, pencils, etc. These are like primitives that can be reused in a compositional way for all visual comprehension.
I suspect that the Endorhinal Cortex (EC), especially grid cells do the same thing, but for structures and relationships of concepts. I have no idea waht those primitives are, but I’m going to try and collect as many abstract structure primitives that I can.
lineness – lines in a picture, line of ducks, line of people
size comparison of objects – a > b
transitivity – a >b, b > c, so a > c
fixed object orientation – sun shields, umbrellas, roofs
fixed configuration of parts – bicycles
free combination of parts – vehicles
stackable objects – blocks, most objects that are not circular
wearable objects that come in pairs – gloves, socks
breakfast machine – coffee maker, toaster oven
adjectives + nouns – fat cats, big balls
[What could that look like in a computer]
[how could recognizing an object move to structure]
[multi-scale periodic representation as generalization, abstraction, and time?]