If the visual system is not self-producing (as seems plausible), then it is not autopoeitic. But, if it is not autopoeitic, then it is not a robust, adaptative autopoeitic system either, so then, by Di Paolo's account, it is not a cognitive system either.
I mention this "problem" for Di Paolo, but it seems to me a consequence of any version of an autopoetic account of cognition.
Now, I put problem in scare quotes, because this may simply be what the autopoeitic folks want to say about the visual system. They do, after all, tend to have different target concepts for cognitive scientists than do, say, cognitivists. No cognitivist, as far as I know, thinks that plants, slime molds, and bacteria are cognitive systems, but at least some in the autopoeitic camp do.