The Film Machine
Are we on the verge of losing our grip on cinema? Not losing it as an art form entirely, but ceding control of it over to the machines?
by Daniel Benneworth-Gray
I had a horrible, horrible thought in the shower. This happens far too often. I’m not sure if it’s been plumbed in incorrectly or something, but every time I go in there, I end up having some kind of terrible epiphany. It’s like a little cubicle for emptying the mind, rinsing away all the mental detritus and imagination grit. All that’s left is a dark pearl of an idea. I should’ve learnt by now that if I want to stay happy, I should just avoid the shower at all costs. To hell with cleanliness, I need my sanity.
Today’s horrible thought: we, mankind that is, might be on the verge of losing our grip on cinema. Not losing it as an art form entirely, but ceding control over to the machines.
Most stories, particularly those within the narrow genre-fication of cinema, follow certain rules of structure, style and repetition. We’ve enjoyed a century of perfecting the tropes and conventions of mainstream story-telling on the big screen. The audience can identify these narrative iterations and variations, because they are essentially no more than familiar patterns. Patterns that, in theory, a machine could understand.
Now, I’m no hardware designer, but picture this: a gleaming, brushed-aluminium wood-chipper. At one end, thousands of films are being thrown in. And a copy of Robert McKee’s Story, outlining the principles of screenwriting. And Kurt Vonnegut’s rejected anthropology master’s thesis, The Shapes of Stories. And maybe something like Billy Mernit’s Writing the Romantic Comedy. All thrown in to the chipper.
Beneath that shiny casing, whirring happens.
Some time later, something ker-plunks out the other end. Not just a finished screenplay, but a completed film. No need for cameras or actors or costly night-time location shooting permits. The machine has written it, directed it, edited it, all using the finest CGI that the near-future has to offer. A random synthesis of all the cinematography and performance and cliché that it’s been fed. Nothing more than an act of pattern recognition and replication.
This idea of computerised creativity was touched on in Spike Jonze’s film Her. Throughout the story, a husky operating system (think UberSiri) composes music for its user, simply using every piece of music that has ever come before as instruction. It’s not AI as such, rather a massive database of human-created art reduced to a series of variables and formulae.
They may be vast and elaborate, but there are rules that dictate how notes and rhythms are pieced together. The same applies to film-making, essentially sound and light and movement. All it takes is a machine big and shiny enough to do it, and we’re just a couple of technological steps away. Once those steps have been stepped, all it takes is for someone to decide that this new automated model is financially viable, preferable even.
I dread to think what the results will be like. Like I said, it’s a horrible thought. But don’t think for a second that horribility is enough to prevent it from happening. Look at any other industry: profitability and horror go hand in hand. And the worst thing about all of this, it’s not just fanciful shower-based conjecture. Primitive models of the wood-chipper (word-chipper?) are already in use elsewhere.
Chicago-based startup Narrative Science provide — in their words — business intelligence reporting services. Basically, this means they make software that analyses data and transforms it into narrative. Business can turn impenetrable records of stock figures and sales projections and other number-gristle into human-readable prose. Given the right data to mine, annual reports can write themselves.
And it’s not just in the boardroom. Various news outlets are already using software like this to write stories. Journalist and programmer Ken Schwencke has created an algorithm to write up particularly statistic-laden news reports. In March of this year, the LA Times used this to break the news of an earthquake occurring in the city, minutes after it had happened.
A shallow magnitude 4.7 earthquake was reported Monday morning five miles from Westwood, California, according to the U.S. Geological Survey. The temblor occurred at 6:25 a.m. Pacific time at a depth of 5.0 miles …
Okay, so it’s not the most scintillating of reads, but the few paragraphs met the core requirements of any news story: they reported the facts in an accessible and timely fashion. This style of reporting is perfectly suited to finance, science and sports, fields in which statistics buried in easily-mined big data is the essence of the story. This is just the beginning — beyond replicating the basic syntax of human reporting, the software can be programmed to incorporate journalistic catchphrases and flourishes, even an editorial stance on the events being reported.
Narrative Science’s cofounder Kristian Hammond predicts that 90% of news will be written this way within the next fifteen years, and that a computer could win the Pulitzer Prize even sooner. When the line between automated and creative writing becomes that blurred, there’s nothing to stop it being applied to any narrative-based form. Tropes and clichés within films may be trickier to extract and replicate than earthquake statistics, but to a computer, it’s all just data.
And then where will this take us? Presumably, computers will continue to get faster and more portable. There’s no breaking the incessant march of exponentiality. The film-making wood-chipper will become tiny and fast and efficient, nothing more than an app that’s able to churn out a bespoke, unique lump of entertainment on demand.
“I want to see a romantic comedy”, you’ll bark into it, horribly, “something along the lines of Sleepless in Seattle … but with a more Steve Martin meets nouvelle vague vibe … and significantly more velociraptors.”
Cinema is still a relative young art, and it’s not one that keeps still for long. Size, format, girth — everything about it changes all the time. Maybe disposability and convenience and automation are the next logical progression. Maybe it won’t be that horrible. Maybe.