Writing fiction can seem like a uniquely individual endeavor, similar to painting, but perhaps unlike recording an album of music, that is, unless you were Prince. Nonetheless, examples of collaborative fiction are readily available, with authors such as Henry James and Harriet Beecher Stowe having dabbled in the genre. And suspiciously, some of the most productive writers of the past and present often had a partner, or several, in tow. Now, is a new approach to writing nearly available to us, powered by the technology that underlies those words we’ve all heard, like “machine learning”, “big data”, and “recurrent neural network”? OK, that last one is new to me. But, explorations by Bay Area author Robin Sloan, of Mr. Penumbra's 24-Hour Bookstore fame, suggest we may be fast-approaching an era in which your word processor, like a trusty coauthor or cohabitator, can finish your sentences, and ultimately, your books.
The majority of fiction writers apparently work alone, and that is true even for prolific novelists. Danielle Steele, reportedly the best-selling author alive, has averaged about 4 books a year since the early 1970s and is succinct in dismissing any suggestion that she’s needed help. In interviews, she calls the idea of a ghostwriter a “shocking concept”. During his 50-year career, the French writer Georges Simenon wrote over 400 novels, ranging from an expansive, pulpy detective series to serious autobiographical novels. He claimed that a self-imposed 80-pages-a-day quota meant he rarely spent more than a week and a half on any given book.
And alternatively, many collaborative works have been one-and-done ordeals, and often not for public consumption. The great Ken Kesey led a group of graduate students at University of Oregon in writing a novel called Caverns that Penguin did publish in 1989 under the name O.U. Levon. The book, which describes an eccentric’s search for a desert cave in Utah containing ancient American treasures, seems ripe for a film adaptation with Nicolas Cage generously featured. Earlier in the 20th century, ten Italian Futurists, including the man who penned their manifesto, wrote a novel together, I would guess in the free time after their day jobs currying favor with Benito Mussolini. Lo Zar Non E Morto (The Czar is Not Dead), written in 1929, is about Czar Nicholas II of Russia turning up in 1931 in China, and the problems his unexpected survival posed for the young Communist regime in his homeland (Mr. Cage would also bring this story to life well). For better or worse, these two works are remembered, if at all, for the collective nature of their creation, more so than their quality or popularity.
Now, what about books that were decidedly a team effort, products of a so-called writing factory, but sold millions and remain widely loved? James Patterson and Alexandre Dumas are two writers well known for producing such books, though they are from different eras and, it is safe to say, on different sides of the canon of Western literature’s shiny gates.
Patterson has written about 150 books and counting, usually hefty thrillers, including the well-known Alex Cross and Women’s Murder Club series, where, respectively, a Washington, DC detective and a group of San Francisco women solve brutal murders, kidnappings, and other crimes in which the perpetrator often has a grotesque kink or two. Patterson has spent most of his career at the top of the New York Times bestseller list, become the most borrowed author at British libraries for years running, and enlisted other writers to help keep the books coming (most recently, Bill Clinton). A sprawling New York Times Magazine article from 2010 is a good primer on his legend: about 1 in 17 novels sold in America is written by Patterson, who was formerly an advertising CEO responsible for slogans such as “I’m a Toys ‘R’ Us kid,” but now coordinates his network of co-authors from an oceanfront mansion in Palm Beach, Florida. Accordingly, solo Patterson novels are few and far between these days, as Patterson has perfected his collaborative process: serving more as a supervisor than equal partner, Patterson establishes an outline and then reviews, edits, and sometimes adds to his co-author’s chapter drafts, before sending it off to his publishing company, which reserves a whole team of staff for Patterson alone. The circumstances of a Patterson novel more readily call to mind a conference room on Madison Avenue than a creaky attic apartment in the Latin Quarter, but if books are published to be read (and that’s usually how it works), Patterson’s success speaks for itself.
Dumas, like Patterson today, wrote a lot of books. Or at least, he put his name on the cover of a lot of books, around 650 of them, including brick size classics like The Three Musketeers and The Count of Monte Cristo. Unlike Patterson (at this point…), Dumas has a permanent place of high-regard in Western literature. James Joyce was so fascinated by Dumas’ work that he framed much of Stephen Dedalus’ adolescence in A Portrait of the Artist as a Young Man as explicitly mirroring the trials of Dantès in Monte Cristo.
Dumas probably had a hand in almost everything that was published in his name; however, he was consistently aided by a variety of assistants and collaborators who were called on to fill barebones outlines, flesh out characters’ setting-appropriate garb, pitch story ideas, or produce whole chapters. Dumas even published “first hand” accounts of far-away places, like California during the gold rush, sourced exclusively from newspaper articles and, at best, conversations with returning travellers. Maybe we can sympathize with his detractors in mid-19th century Paris who accused him in pamphlets of operating a system of “mercantilisme littéraire”. However, the literary tastemakers of Paris (the friends of Dumas, naturally) and modern history did not let these accusations worry them. One intrepid accuser brought a case before the writer’s association Dumas had founded with Balzac and Victor Hugo, La Société des Gens de Lettres. The accuser claimed Dumas’ operation was harmful to the public, his assistants, and even literature itself, but ultimately got jail time for libel in 1845. You can read all about it in this somewhat fawning but truly entertaining biography of Dumas. Meanwhile, the Maison Dumas, as some critics referred to his army of assistants, rolled on, sometimes publishing whole sections of books without their general’s review.
It is easy to forget, when marveling at the writing factories of Patterson and Dumas, that it was simply their talent as writers, not managers or salesmen, which first attracted the readership that eventually justified a mechanized, impersonal production process. Yet soon, a simple application of machine learning could mean that early, even one-off, displays of unaided literary talent are no longer prerequisites for popular success. Evidence of this comes via a programmatic demonstration in a blog post by Robin Sloan, a Bay Area novelist who has achieved popular literary success in a tried and true way: by writing acclaimed books that hum along with engaging plotlines, and touch on topics of contemporary interest (how we let technology govern our daily lives, homemade bread making, etc.).
In his blog post, and at a recent talk hosted by Ars Technica in Oakland, Sloan described an “autocomplete” tool he built using a recurrent neural network (a way for a computer to connect the dots between similar bits of information), a simple text editor, and a ton of old sci-fi stories. The tool finishes sentences with phrases that mimic the sci-fi stories that it “learned” from. Sloan likens it to writing with a “very well-read parrot” on your shoulder. The allure, for Sloan, is not having a program that writes for you, but an AI program that works only in collaboration with humans, rather than independent of them. This view is interesting in its own right and is compatible with those of the techno-optimist crowd, who expect tech and innovation to continuously better our lives. However, it is equally easy to imagine a tool like this evolving to be used exactly in a way that Sloan is “absolutely 100% not talking about”, turning our word processors into a Dumas assistant who can enhance a simple plot line with the imagery of a 19th century novel of manners or a Beat generation road odyssey. Say an author like Patterson wanted to sell more books in Scandinavia: instead of calling on a popular Swedish crime writer to write with him (which he did for that very reason), he could train his word processor on the books of Steig Larsson and Karl Ove Knausgaard, maybe throwing in some Strindberg for good measure. Come to think of it, it’s hard to say whether the latter route would have been worse.
With Dumas and Patterson in mind, any autocompleted book of the future might not be so different from a lot of the books that are published and sold by the millions today. A divergence would arise, though, in who gets access to the writing factory. Up until now, the assistance of and dependence on a team of collaborators has been, ostensibly, earned with early displays of individual brilliance by an author. A collaborator in the form of a word processor would be available to everybody. The democratizing aspect of Sloan’s exploration is then encouraging, if we accept that a certain kind of cookie-cutter book has been rolling off the assembly line for some time. In one sense, whatever gets more people writing and reading should be celebrated. Whether it distorts further the perception of writing as an unadulterated reflection of an individual’s worldview and imagination is up for debate.
// jamie a.