Opinion | Artificial Intelligence may prove to be the right stuff for writers

Can we use Artificial Intelligence (AI) to write a novel? The writing fraternity might scoff at the idea, but it has already happened. Ross Goodwin, a novelist, went on a road trip across the US in an attempt to emulate Jack Kerouac—to experience being on the road and find something essential to write about. The result, 1 The Road, is being marketed as the first novel written by AI. This writer is a microphone, a GPS, a camera hooked up to a laptop, and plenty of linear algebra.

Writers may not recognize AI as a fellow-author, but they themselves do make use of technology in their craft. Besides the more obvious kinds of software that writers employ, a program that actively assists their creative process might meet their approval. One aspect of the process that could benefit from AI is continuity.

The alert reader or viewer often spots continuity errors in books and films. Perhaps the most famous and funny instance of gaping plot holes is in Raymond Chandler’s cult detective novel, The Big Sleep. Several dead bodies turn up in the book, and one body, that of a chauffeur, isn’t accounted for. When he was asked “whodunnit”, Chandler famously retorted, “Damned if I know!”

In Daniel Defoe’s classic Robinson Crusoe, there is a huge continuity error that went unchecked even by his editor. When Crusoe is famously shipwrecked, he strips naked and swims back to the shipwreck still afloat on the sea to retrieve food supplies. And he stuffs as much as he can into his pockets and swims back to shore. Popular fiction too falls prey to such errors: Enid Blyton stories, the Hardy Boys, Nancy Drew, Harry Potter, and the Dan Brown conspiracy thrillers, for instance. They seem to especially happen when the novel has a rich or layered plot, or is of epic proportions.

It was because author Vikram Chandra feared such problems while writing Sacred Games that he began to look around for a software program that would help him keep track of his fictional world. He never did find one at the time, but he is developing one now that could be of immense help to authors. The plan is that Granthika, Chandra’s software, can eventually collaborate with authors by interacting within their fictive world. Let’s say one of your characters is introduced as being vegan, and some hundred pages later she’s slurping curd rice and pickle off a plate or eating bread omelette from a street cart, the software will flag you. The timeline is all off in the plot, but you just don’t know it yet—the software will check you.

The handful of word processing programs available when Chandra was writing Sacred Games—FocusWriter, Scrivener and Ulysses—didn’t come even close to meeting the demands of his projected novel, which was going to be nearly 1,000 pages long with some hundred characters and spanning sixty years. These programs let him store information that would go into the novel, but they could not be intuitive, they could not fact-check him, point out contradictions, go over details, description and minutiae, and untangle plot elements. In short, what Chandra was looking for was an AI collaborator, what in filmmaking is called a “Continuity Director”.

He writes in his blog: “Now I understood well a problem faced by all writers: how to keep track of the logistics of a narrative, of who, what, where, and when. This might sound easy to the non-writer. After all, even if a novel has a dozen speaking characters, how hard can it be to manage them all, and their relationships, and their movements? But, this is surprisingly difficult. Trying to remember where you left a minor but crucial character a hundred pages ago is error-prone and scrolling back and running word-processor searches to find answers is laborious and annoying. The traditional writerly tools of dealing with these problems include index cards to carry character notes and details, a timeline drawn on a wall, perhaps a spreadsheet to keep track of birth dates and ages, and maps festooned with marker pins. I know of writers who build entire “suspect walls”, complete with coloured threads to mark relationships. Despite all this effort, errors slip through into print, past the copy-editors and fact-checkers. I’d received letters and emails from readers myself, pointing out—usually kindly—slippages of age, impossibilities of travel time, contradictory accounts of relationships.”

Chandra eventually used MS Project to keep things in some order, but felt that a program should do more than just store your text and help you retrieve it—it must keep track of your narrative, and bring order and perspective to it. Chandra writes: “Granthika’s text and knowledge-base is built from the bottom up to be amenable to reasoning and AI routines, so that as we develop the system, we will increasingly be able to add more and more intelligence to the system, such that the editor becomes an effective assistant to the writer, taking care of much of the boring bookkeeping. The writer should then be able to concentrate more on what really matters in a narrative: the people, the plot, the emotions, the themes.”

We are at a crossroads: how much technology is good enough? Can we ever make that decision? 

V.R. Ferose is an inclusion evangelist and bibliophile, and also senior vice-president at SAP, based in Silicon Valley.