I found the article a bit roundabout. Perhaps I'm not part of the intended audience; I understand convolutions, but have only limited familiarity with the internals of neural networks.
This sounds like a big assumption. I think this is a deep-learning oriented blog post, and not for people with a background in systems & signals (basically information theory). I don't understand mathematical convolutions super well (I know the vague derivation, of integration of f(t)g(t-T)dT, and the metaphor of multiplication in frequency space), but I do get how convolutions work with neural networks, and maybe this is a failure on my part, but it doesn't feel like there's a super strong connection between the two. Is there something I'm missing out on?