Machine learning is totally changing what we think of as literature

The definition of sci-fi is notoriously slippery. For some the genre is defined by its authors and landmark novels – starting with Twenty Thousand Leagues Under the Sea author Jules Verne, or pushing things even further back to Mary Shelley's Frankenstein. Others argue that it's more about ideas than people. The Encyclopedia of Science Fiction, for example, calls the genre the "literature of cognitive estrangement." But what if crunching the data on thousands of books could give us a more definite answer?