Perhaps not impossible - sometime after 2040, a deep learning model reconstructs the entire "Titanic" (the James Cameron film) from a total of 1,000 frames, the script, and voice samples of the actors.