You just pointed out the biggest limitation with this. Even if it can apply the style, that's all it is doing. It doesn't understand the context or content of the film. Everyone knows in the first fifteen minutes of the film of the importance of the monolith and how imposing it is, but the algorithm doesn't. Even image recognition couldn't help you there. In visual art where filters and different media are used, it is rare to non-selectively apply them. Computers with cognition and creativity are a long, long time away. That is why I can't stand people who are scared of some skynet scenario; even if AlphaGo mastered Go or Deep Blue mastered Chess, real-life problems are not so strictly defined and require much more intelligence and insight.