Editor’s note: Nebula Award nominated author Jason Sanford is now publishing a monthly column in SF Signal called To the Ends of the Universe. These columns were originally printed in the Czech SF magazine XB-1. Jason’s novelette Her Scientifiction, Far Future, Medieval Fantasy, published last year in Interzone, is now available as a Kindle ebook.
It’s time for a quick thought experiment: Without delay, imagine your favorite novel. Do you remember how you felt the first time you read it? How many times have you read this story? Does the novel continue to hold a treasured place in your heart?
For me, the first novel which pops to mind is Arthur C. Clarke’s 2001: A Space Odyssey. I read the novel as a child, long before I saw the Stanley Kubrick film co-written by Clarke. 2001 was the first story which truly opened my mind to the far reaches of eternity and I still love—and reread—the novel to this day.
Everyone reading this essay is likely to have equally emotionally reaction to their own favorite novel. After all, the relationship between readers and fiction exists on an extremely personal level, with readers often developing human-like relationships with their favorite stories.
But now that you’ve pictured your favorite novel, ask yourself how you’d feel if a computer program had written it. Would you still love the novel? Would a computer-written novel change the human-like relationships readers have with stories? Could you even see a program writing the novel?
This isn’t a science fictional thought experiment but instead one rooted firmly in current technological developments. While the day hasn’t yet arrived when computer programs can write fiction, journalism is already at risk of being AI-sourced by the new generation of intelligent computer writing programs.
In fact, according to the provocatively titled “Will Robots Steal Your White-Collar Job?” on the news website Talking Points Memo, a program called StatSheet has already generated “a million pages of news” over the last four years. The program’s inventor, Robbie Allen, adds that “Within the next three to four years, it will be better than what a human can produce. And the reason for that is pretty much the foundation of computation: We can analyze and access significantly more data than one person can on their own.”
I have no doubt about that last point regarding data handling, and it will be interesting to see where this trend goes. I suspect these writing programs will be very useful in cranking out breaking news articles, which are essentially a simple retelling of facts. And if these programs are already used for journalistic writing, we can’t be far off from them trying their cyberhands at fiction.
Which brings up my original question: How will readers react to computer-created fiction?
As a lover of written fiction, I want to say that computer programs will never successfully write in the genre. However, as a science fiction writer gazing into the near future I also know this to be a lie. Just as the programs are already writing simple journalism, I see no reason why they won’t eventually write simple fictional stories. All they’d need are certain parameters and inputs from their human programmers, such as a base set of storylines, characters, and situations known to appeal to readers.
If you don’t believe me, take a look at how Hollywood creates most of their movies. Their “system” for creating blockbuster films uses a human version of programming, with scores of writers and directors working on each film. The teams slightly change the settings and characters of new films so everything appears fresh even as Hollywood movies are increasingly programmed around a few basic, reoccurring theme and plot archetypes.
If Hollywood can create successful stories using their mass-production process, I don’t see why computers can’t be programmed to do the same. But don’t expect these programs to create great fiction. Just as AI programs currently only create simple journalistic stories—and just as most Hollywood blockbusters hardly qualify as great films—I suspect the same will apply to programmed fiction.
My reasoning is simple: While computers have proven themselves to be very adept at crunching data, they have yet to provide the deeper meaning, context and understanding which the best writings create for both author and reader. This applies across all types of writing, whether journalism, fiction, poetry, and so on. This deeper meaning is also one of the main differences between writings which are merely good enough and writings which are great.
In Douglas Adams’ The Hitchhiker’s Guide to the Galaxy there is an extremely powerful computer named Deep Thought, which is programmed to find the “Answer to the Ultimate Question of Life, the Universe, and Everything.” After computing that problem for seven and a half million years, Deep Thought proudly proclaims the answer: 42. When Deep Thought’s creators express frustration at the answer, the computer explains that the problem is “you’ve never actually known what the question was.”
Obviously Adams was parodying humanity’s belief that we can ask a computer to crunch the numbers and provide all of the biggest answers in life. That will likely never happen. And since great fiction is one of the main ways in which humanity explores life’s biggest questions, I don’t see how computers could ever create the fiction which gives deeper meaning to our lives.
But this doesn’t mean computers won’t one day write more simple fiction. I can see these programs being used as a tool in the mass production of basic stories, along the lines of what Hollywood’s “system” now produces. But don’t expect greatness from these programs. At least, not greatness which resonates on a personal, human level.
Or maybe, as a fiction writer, I’m simply hoping this won’t happen. Perhaps I’m simply being naive. Perhaps not. But I do know what I love about the act of reading. And until a program can create deeper meaning and understanding in stories, I’ll remain skeptical of computer-written fiction.