Forget Facebook manipulating our timelines as a mass social experiment, or Google mining our personal emails for advertising keywords. When it comes to algorithms, nothing gets the blood boiling more than the subject of creativity. Questioning whether an algorithm will ever create a work of art on the level of a Franz Kline or Franz Kafka (or, hell, even a Katy Perry) is enough to have many people fleeing for the hills to smash up some looms.
But the reality is that algorithms are already playing a big part in the creative industries -- and this will only increase over time. In my book The Formula: How Algorithms Solve All Our Problems -- And Create New Ones, I discuss Epagogix, a London-based company which uses vast artificial brains called neural networks to challenge screenwriter William Goldman's famous assertion that "nobody ... knows the least goddam [sic] thing about what is or isn't going to work at the box office." While it may be true that one person alone cannot predict hits and misses with total accuracy, Epagogix's data-mining and deep learning algorithms have shown that they can accurately forecast how much money a movie is going to make at the box office.
But here's where things get weird.
When Epagogix submits its final report to a movie studio (CEO Nick Meaney won't reveal which studios he works with, although he says they're among the biggest) it presents a thin dossier -- just two or three pages bound by a single staple -- culminating in two different numbers. The first of these is the projected box-office forecast for a film as written. The second, more mysterious, number is usually around 10 percent higher than the first figure, although it can sometimes be up to twice its value.
This figure is the predicted gross for the film on the condition that certain recommended tweaks are made to the script. Since regression testing is used to analyze each script in forensic detail, the neural network can be used to single out individual elements where the potential yield is not where it should be -- or where one part of the film is dragging down others. For better or worse, this is the point at which an algorithm makes creative decisions.
Okay, we might say. As far as creativity goes, the movie industry has always been ruled by technocrats. Writing excitedly about cinema in its earliest days, Soviet filmmaker and montage pioneer Sergei Eisenstein claimed, "One does not create a work, one constructs it with finished parts, like a machine."
But algorithms also have plenty to say about supposedly less industrial forms of creativity.
In literature, a growing group of scholars have called for an "algorithmic criticism" that would turn literature studies into a branch of computer science. For instance, algorithms can be used to determine "vocabulary richness" in text by measuring the number of different words which appear in a 50,000-word block of text. In doing so they can produce unusual insights, such as the fact that a "popular" author like Sinclair Lewis -- sometimes derided for his supposed lack of style -- regularly demonstrates twice the vocabulary of Nobel Prize laureate William Faulkner, whose work is considered notoriously difficult. Algorithms can also be used to determine authorship, as was the case last year when they were used to reveal that The Cuckoo's Calling -- a crime novel written by supposed first-time author Robert Galbraith -- was actually the work of Harry Potter author J. K. Rowling. Rowling later admitted to writing under a pseudonym.
With knowledge of the high-level details that characterize an artist's work (Rowling's recurrent word pairings, for example, or the temp and time signature of a particular musical act) it might even prove possible to generate new works in an existing style.
Looking for a new Beatles album that sounds like it comes from their 1967 "Sgt. Pepper" era? Lior Shamir has created an algorithm capable of sorting the Beatles' music into chronological order by tracking their musical progression. A future mission? Using this data to create new music. "Theoretically it's certainly possible," Shamir told me earlier this year. "The problem right now is the amount of computing power you would need to let the computer do that kind of composition."
Of course, in an age where no two users have to show the exact same Google results for a search, we might argue that the day of "mass appeal" is over. Companies like Epagogix offer a modern spin on an old-fashioned idea: that there is such a thing as a work of art that will appeal to everyone. But algorithms, neural networks, and assorted other similar apparatus can go much further than this. In an era of digital reproducibility why not democratize art by finetuning it for each person?
It is into this space that the likes of contemporary Italian architect and designer Celestino Soddu fit. Soddu uses what are referred to as "genetic algorithms" -- which replicate evolution inside a computer -- to generate endless variations on individual themes. Adopting the idea that living organisms are the consummate problem solvers, and using this to optimize specific solutions, Soddu first inputs what he considers to be the "rules" that define, say, a chair or a Baroque cathedral. His algorithms can then conceptualize what a particular object might look like were it a living entity undergoing thousands of years of natural selection.
Calling the results "idea-products", Soddu's genetic algorithm means a trendy ad advertising agency could fill its offices with hundreds of chairs, each one slightly different, while a company engaged in building its new corporate headquarters might generate thousands of separate designs before deciding upon one to go ahead with.
Another illustration of this possibility is the work of Alexis Kirke, a former Wall Street quant and Research Fellow at Plymouth University's Interdisciplinary Centre for Computer Music Research. In 2013, Kirke created a short experimental film which changed the direction of its narrative based on an averaging of the biometrics of those watching it. At screenings, audiences were transformed from passive consumers into active participants by being fitted with special sensors capable of monitoring their brain waves, heart rate, perspiration levels and muscle tension. These indicators of physical arousal were then fed into a computer, where they were analysed in real time, and the reactions used to trigger different scenes. A calm audience could be jolted to attention with a more dramatic sequence, while an already tense or nervous audience could be shown a calmer one. The branching narrative ultimately culminated in one of four different endings.
Some of these examples challenge our perceptions of what creativity looks like. It is clear that, just autonomic law enforcement and identity in the age of Google, creativity in the algorithmic era won't look exactly the same as pre-digital creativity. But why should it? Cinema and popular novels reflects the Industrial Age with its focus on standardized products designed to sell to the largest number of people possible.
We're only starting to find out what the next step in that evolution looks like.
Some of the answers we are faced with will be challenging, as art and creativity should always strive to be. For instance, what is the difference between algorithms creating and algorithms being used to help create? Many artists over the years have employed assistants to help them create their works. Damien Hirst painted the first few dozen of his famous spot paintings, but left the others to an army of (human) assistants. Well over a thousand paintings later, and even the so-called "art experts" can't tell you which were painted by Hirst. To him, it doesn't matter. The art comes from the idea; not the execution. What about in cases where the "idea" comes from an algorithm, then?
We can argue that algorithms which predict hit pop songs might be cynical, but the creation of art has often been calculated to please a particular audience -- such as the great Renaissance artists who created paintings and sculptures to impress their patrons. What is the significant difference between artists creating commissions for the church, and an algorithm generating a music track to match our mood?
The apparent gulf between computer science and the humanities harks back to the era in which we saw computers as dumb, number-crunching machines. Today we know that they can do so much more than that: from driving cars to diagnosing medical conditions. Why, then, should the creative process be beyond them?
Luke Dormehl is the author of The Formula: How Algorithms Solve All Our Problems -- And Create New Ones, out now from Perigee. He also contributes to Fast Company, Wired, the Guardian and other publications.