The human visual cortex employs a combination of bottom-up and top-down processing. So before we see a dog the brain sort of generates a picture of a dog top-do

The human visual cortex employs a combination of bottom-up and top-down processing. So before we see a dog the brain sort of generates a picture of a dog top-down, and compares that with the visual stimuli. My feeling is the problem with the "AI" described here is that it's still bottom-up only.