Upscaling made some sense, frame generation did too, but predicting the future? No chance.
Maybe if you can single out only certain parts of the frame, like characters doing an animation, yes.
I do believe there's something fundamentally wrong about using AI to try and increase the graphics...