In the column “Computer Recreations” of “Scientific American”, September 1984, authored by A. K. Dewdney, we find this:
The failings of a digital eye suggest there can be no sight without insight
The “Perceptron” is, in this case, a thresholding “head demon” collecting votes from “local demons” that can arbitrarily (but fixedly, think lookup table) emit a 1 or 0 depending on the pattern that appears in their local field of view. Or at least that’s how I understand this.
The article discuss the fact that this “Perceptron” cannot reliable decide whether a pattern is connected or not. Yes, it’s too simple.
Note that this column was written 4+ years after Kunihiko Fukushima’s Neocognitron multilayer NN able to deal with handwritten digits, but the expanded edition of Perceptrons was still 4 years in the future.