You could feed a machine with context, on top of feeding it with existing imagery, or any kind of additional data you can imagine. A more sophisticated version could scrape a certain amount of "drama" for instance. Input from different news sources could be programmed to affect art style. You could enable artistic "moods" based on a set of random parameters. Each new source of data would affect its ability to generate new content. Given the fluctuating definition of art, I see no reason why AI wouldn't be able to generate original artworks in the future and trigger human emotions in the process. A new machine-generated artistic movement sounds much more complex but may well happen further down the line.