Firstly, it's probably going to be 50 years before this turns into an actual medical procedure rather than a proof-of-concept experiment. Let's just get that out of the way.
So what they're doing is taking people with a defective retina, and adding a synthetic one. The retina normally receives photons and sends a signal along the optic nerve. What they're doing is implanting a silicon photoreceptor behind the retina of people whose retinas aren't doing the job. The chip receives the photons and sends an electrical signal, serving the same function as a "healthy" retina to some fidelity. The results are sort of low-fi since (a) it's just a proof of concept trial, and (b) the retina is a horrendously complex photodetector so it will take a lot of work to approach that in an implantable device. But dude, blind people. Seeing. Go, science!