I want a machine learning enhanced analog loop. I want to be able to point a device at a screen and have the machine learning clean everything up and restore hi

I want a machine learning enhanced analog loop. I want to be able to point a device at a screen and have the machine learning clean everything up and restore high quality sound and video. Training data would be easy. Take a bunch of videos, play them on a screen, record the screen. You now have the input and the output, x and y. It's just supervised learning from there. I know there are better options currently, but in the end we have this, unless they take away general purpose computing (but that would a huge blow to our economy).