As if competition from the giant online slushpile wasn’t enough, seriously writerly folk now have an even more alarming threat: computer-generated works. And if you’re a cookery writer, this is now a genuine and serious prospect. Because IBM’s artificial intelligence/cognitive computing system IBM Skynet … oops, I mean IBM Watson … has now been set to work on human gastronomy, in partnership with the Institute of Culinary Education (ICE), and has whipped up “a new book…introduced on April 14, 2015 – Cognitive Cooking with Chef Watson. It features more than 65 original recipes exploding with irresistible new flavors developed by Watson and brought to life by ICE chefs.”
According to IBM, “Chef Watson is able to learn recipes, dish types, and ingredients, understand human taste preferences, and then rearrange and redesign the data to generate unique combinations of savory ingredient pairings. Given the numerous different combinations of possible ingredients in the world, it’s impossible for a single person to imagine and reason about them all … Through the Chef Watson collaboration, ICE chefs provided input and feedback on the system’s ingredient choices, infused the cooking with their own interpretations, and together delivered creative recipes.”
Since IBM Watson is already active in the health field, less life-or-death decisions like what ingredients to put into a recipe should be … well, a piece of cake in comparison. Actual creations include the Spanish Almond Crescent, Indian Tumeric Paella, Caymankin Plantain Dessert, and Swiss-Thai Asparagus Quiche, which already sound surprisingly acceptable and appetizing.
I can foresee many more applications for IBM Watson to get to work on in partnership with Big Media. For instance, with genre fiction tropes so formulaic and much genre product in thrillers/crime and romance so uninspired, how about Genre Writer Watson? Can the first James Patterson collaboration with IBM Watson be far off? Or, how about Celebrity Ghostwriter Watson, so no human mind ever has to parse the burblings of a Zoella or one man and his dog? And look at the big picture: The longer we keep these machines busy churning our asinine moronic trash, the less time they’ll have for plotting to take over the world …
Well, a few years ago I read two books from a wildly popular urban fantasy series that were so much alike it seemed like the author simply did a few dozen search and replaces and … Voila! The hero battles werewolves in the woodpile in book one, vampires in the barn for book two, and probably hobgoblins in the pantry for book three if I had stuck around for more. And that’s a Hugo nominated author.
As a computer programmer, I have no trouble believing the Novel Writing Machine is not too far off. The genre forumas are there and would not be difficult to systematize.
Would readers notice? Would they care?
As someone who reads and has written mystery, romance, and science fiction, I seriously doubt that an AI could write any of them, particularly romance which is about emotion, not plot. All good fiction is about emotion and character, not tropes or formulas which are more about reader expectations.
Years ago, on April 1st, an article appeared in one of the publishing trades about a software program that could write novels. It was depressing how many editors wished it was real so they didn’t have to deal with authors.
I would propose that Watson may someday become excellent at writing the three genres you specify. Especially Romance, because of the volume of Romance novels available and trite themes. Watson’s process would be something like this:
Data Scientist asked Watson “Watson. Write me a Romance novel that will sell well and be well received by critics. Then Watson will get to work doing the following in less than 15 minutes.
1.) Select 1000 books from the best selling Romance novels written in the past ten years and correlate volume sold with positive critique review and narrow down to the top 50 best selling, best reviewed Romances. 0.0028 seconds
2.) Read all 50 books. Analyzing all plot themes, character names, locations, pacing, number of story arches, beginnings, growth and outcomes for all characters. 0.75 seconds
3.) In step 2 Watson found a pattern correlating certain European countries. Watson reads Romance has also observed that the best selling romance novels that are also the highest reviewed are based on historical stories. Watson reads 10 history books and 10 travel guides on the region. 1.2 minutes
4.) At this point Watson has decided to create a 320 page historical romance novel told in flash back based in the little village of St. Paul de Vence in France. Watson makes the decision to read the scripts of the last 15 highest grossing Hollywood Romances and decides on two motifs that it has determined will blend well together. Supporting characters and subplots have also been created at this point. Watson finishes it’s virtual outline then filters out concepts that may conflict with the critics, 3.53 minutes
5. Watson completes transforming its outline to a novel and e-mails Data Scientist the resulting PDF. 6.28 minutes from the request.
Data Scientist asked Watson to provide a rewrite focused more on what fans want than critics. Watson reads all Amazon reviews of the 50 books that it used as it’s research foundation. Then reiterates all steps used to create the novel with focus on what readers desire. The rewrite is available 10 minutes later.
While we are a ways away from seeing Watson accomplish this I don’t believe it’s the distant future. I think between 6-10 years we’ll be there.