I read the paper a bit, looks like a rant more than a paper. He's just arguing that you can assign anthropomorphic features to any computing system regardless of substrate as if the researchers working on it didn't know that already. I don't see how implementing a perceptron on AoE is any relevant to his argument at all.

He also talks shit about connectionism even though it is rooted in neuroscience and still used very often to explain all sorts of behavior.

We use anthropomorphization in science because that is how we make sense of things. We say that the ball "wants" to go downhill not because it we assert that it has a conscience. The same goes for analyzing behavior in LLMs.