Retinal prosthesis. Converts light to neurotransmitter release (as opposed to the older generation, which converted light to voltage). The problem: positioning it and encouraging it to integrate with the fibers of the optic nerve, the remnants of a damaged retina, etc, etc. Not a trivial task, but significant progress is being made.
Also important to note that once
one of these things really works, the race will be on for the dominant design in neuroprostheses - after that, we will see a lot of recombinative innovation focused on discovering the 'right' value equation. After that, costs will drop massively - fast. Think about Lasik - it's a rough analogy.