What is the bottleneck for an artificial eye? At a super basic level, cameras can handle high quality images in the visible spectrum, which is loosely equivalent to rods/cones + pupil + lens. Part size doesn't seem to be an issue, as cones are a few um, while transistors are nm sized now.

Is it software related? The brain does a shit ton of processing to merge the two inputs, generate a 3d image, motion tracking, etc.

Is it language/information related? As we could recreate it, but would have no way to knowingly convert our electronic description into the correct neurons firing?

Ignoring higher level functions(reasoning/reacting to the input), an artificial eye seems super doable in theory. Is there a bottleneck I'm not thinking about, or a failure to understand mechanisms that could be copied or at least recreated with our current computer/optical knowledge?