Automatic Stories

If pressed to choose the types of jobs or tasks for which humans will longest be better suited than computers, I suspect that most people would point to creative fields, like writing or art, and to invention.  What underpins these disciplines is, ostensibly, creativity, an attribute considered to be uniquely human and beyond the ken of modern computers.  Prognosticators of many stripes have leant into this assumption, including the common conception of futuristic utopias in which machines can conduct all labor and leave humans to the ‘higher pursuits’ of artistic expression.  This, I think, is evidence that prognosticators possess insufficient imagination.

A few months ago, I shared with you an article I read from Nature detailing how artificial intelligence tools are helping scientists craft their papers.  The technology itself is interesting, and I somehow doubt that it will lead to more readable papers and clearer communication, but today’s post is not so much about the details of the technology, or pondering whether we will one day live in some kind of post-scarcity utopia in which our machine-slaves can solve all of our problems, generate optimal art, and fulfill our every whim in addition to freeing us from manual labor and rote tasks, as it is about reflecting on the nature of creativity and the process that we are really going through when we attempt to ‘create.’

When I tell people about my writing, a common response I receive is ‘oh, I could never do that; I’m not creative enough.’  People who don’t try to create new things, whether they be new technologies, elaborate paintings, or literary fantasy worlds, often have a misconception that ideas pop into creatives’ heads fully-formed, like Athena, or maybe like Terry Pratchett’s ‘inspirions,’ magical particles of inspiration that sleet through the universe and lodge in people’s brains to give them new ideas.  In truth, creativity is more like a muscle, or any other gradual, developmental process.  Ideas don’t come to me fully formed; they are the product of compiling observations and approaching stimuli with an openness to the possibilities they present for interesting combinations.

Most people who create things understand this, but we do like to think of the process as having an element of spontaneity, that we are crafting something whole-cloth (even if it requires some revisions after it’s finished).  That’s certainly how it can feel while you slowly stitch words into sentences, sentences into paragraphs, paragraphs into chapters, and chapters into a new, unique novel, something that did not exist until you conjured it into being.  Reflecting on how artificial intelligence writing systems go about crafting text led me to wonder if, instead, we are taking a more reductionist approach than we perceive.

This earlier post about choosing words started to get into this question of how we go from an idea to words on a page.  It seems like plucking the right words out of the lexicon, but what if it’s more like a process of elimination occurring at a subconscious level, happening so fast that we are only aware of the end result?  This makes a certain degree of sense, and it’s essentially how artificially intelligent writing software functions.  You look at a situation, and based on all of the factors of the context of the writing, what you’re attempting to express, and the rules you know about language, your brain runs rapidly through all of the words you know to find one that would fit there, and that’s what you put down on the page.  It just happens so quickly and naturally that it feels as if you are arriving at the decision spontaneously, as it were.

With this in mind, it does not seem so far-fetched to imagine a not-too-distant future in which these artificially intelligent writing tools could be used by publishers or authors to generate whatever stories they want without the messy, time-consuming, imperfect step of actually writing them.  You could feed your ideas into one of these systems, and it would spit out a finished novel based on those ideas.  If you don’t like that version, just tell it to try again.

If you think that the idea-generation would render it necessary to retain creatives in this storytelling process, you might want to examine more closely how we get our ideas, or, perhaps more significantly, how publishers choose stories.  Give an advanced writing software a set of parameters that describe a desired market, genre, audience, and so forth, combine it with a database of story archetypes and examples, and add in the wordcraft capability that we already discussed, and you have a computer that could probably write the next bestseller or blockbuster.  Don’t think that literary novelists are safe, either – just because you write depressing things that no one reads doesn’t mean that a computer can’t meet those parameters as easily as it meets those for a romance or police procedural.

To be clear, I don’t mean this to seem like a painting of some dystopian future where AI has supplanted us in all ways.  It’s easy to get into dichotomies and extremes with discussions like this, but just because a computer system could write as well as a human doesn’t mean that they would replace all human authors.  It would change the storytelling market, yes, but I think there would be people who prefer a human written story to an AI written one in the same way that some people prefer Van Gogh to Picasso; not because of some moral position but because of subtle differences in the storytelling that make us prefer one author to another.

These technologies are still a long way from being ready to write the next great novel or even surpass us as storytellers by volume; however, I think it is fascinating to think about, not just for purposes of idle prognostication (as I have indulged in a bit in this post), but for attempting to better understand how humans write.  In many ways, developing and understanding algorithms for computers forces us to understand how we ourselves think, which means that thinking about how to create a computer storyteller might just make us better storytellers, too.