Such AI is still limited by the same resolution (480
interlaced) and it can't create detail where it finds none. To compare, it's like any one of us getting some LSD and asked to draw a picture of the Statue of Liberty. There will be omissions, and bits that were never on the original.
Besides, even under ideal conditions, film still suffers from "the vinegar effect" over time. If for no other reason, I'd be making a digital high-res backup of the original material so
something could be saved. The film stock used will certainly last a few more decades; usually color fade is the first problem - especially reds... once the vinegar effect starts, then problems really begin. After x number of lost frames, no AI would be able to extrapolate the number of moves to mimic the original even adequately. Not going to happen. Ask me again in 50 years, 4 months, 3 weeks, 2 days, 1 hour, and 60 seconds.