All you need is an algorithm to pen a hit pop song or write a bestseller

Writing a hit song or forging a masterpiece has never been easier — all you need to do is crack the code, says Rosamund Urwin
© Karen Karakhanian / Alamy Stock Photo
Rosamund Urwin26 May 2016

Art forgers, music scouts, literary agents and even poets: computers are coming for your jobs. There’s now an algorithm that can create a fake Rembrandt, another that can predict chart-toppers and a third that can work out whether a novel will hit the bestseller list. And this month, Google’s artificial intelligence system made a crack at writing poetry, though the result was even worse than the sub-U2 lyrics your emo friend recited to you in adolescence (a taster: “You’re right/ All right/ You’re right/ OK, fine.”)

Algorithms have swiftly been infiltrating the arts. Their most obvious use is to tailor material to an individual’s preferences. At the start of the year the BBC — with its 43,000 hours of “music content” — began to use algorithms to personalise music on its app. The idea is to build a one-to-one relationship with the listener by learning what you like, rather than the BBC’s traditional one-to-many approach.

Crucially for the money-makers, they can also be used to judge the likelihood of a work being a hit. In September a new book — The Bestseller Code: Anatomy of the Blockbuster Novel — will be released. Its authors, Jodie Archer and Matthew L Jockers, claim there’s an algorithm that can predict with 97 per cent accuracy whether a novel will top the New York Times’ bestseller list. Although details of The Bestseller Code are closely-guarded, it does claim that it pays for a novel to have a 28-year-old heroine, to court realism and to star a dog rather than a cat.

Algorithms can predict whether a song will top the charts too. Using data from The Echo Nest, a music intelligence lab, data scientists analysed 2015’s top 10 dance music tracks for tech site Motherboard, finding their popularity largely predictable.

More controversial is the use of algorithms to determine the worth of art. ArtRank, a service for deep-pocketed collectors, puts a “buy”, “sell” or “liquidate” rating on the creations of contemporary artists, in a similar vein to how an analyst rates stocks. The service’s founder, Carlos Rivera, claims to have developed his algorithms from investment banking. Unsurprisingly, many are aghast at the idea of decreeing the next generation of potential Jeff Koonses as “hot” or “not” before they’ve been allowed to develop fully.

The “Rembrandt” algorithm, though, is different because it created something new. The painting was made last month by the Dutch museums Mauritshuis and Rembrandthuis working alongside Microsoft, ING and the Delft University of Technology. The algorithm analysed 346 of Rembrandt’s paintings, which had been digitised with 3D scans, to make a new one. The result? A man with a moustache clad in black but for a white collar.

But as John Hyman, professor of aesthetics at the University of Oxford and the author of Action, Knowledge, and Will, makes clear, this is no artistic achievement: “What this shows is that an algorithm can generate a pastiche. Don’t think of it as a work of art. It’s as if you used a computer to make a forgery, only in this case it wasn’t made with a dishonest purpose. Judging by the photo, I don’t think the painting is convincing, though that’s easy to say when you know it’s a fake.”

Hyman judged the algorithm’s attempt as “not nearly as bad as some fake Vermeers which fooled people in the 1940s”, but not as impressive as the forgeries of John Myatt, who went to prison in 1999 for conspiracy to defraud.

“It wouldn’t fool an expert at an auction house, but perhaps one day there will be algorithms that can,” he adds. “It doesn’t put the artist out of business, but it could put the [human] forger out of business, especially if you took a less challenging artist to forge, like Mark Rothko or Jackson Pollock.” And if AI can try this with a work of art, why not an algorithmically-derived symphony No 10 from Beethoven (Barry Cooper attempted this in 1988, using the composer’s sketches), or perhaps an ending for Sanditon, Jane Austen’s unfinished novel?

The best apps for Londoners

1/12

A problem arises, of course, if an algorithm creates a fake that is passed off as genuine. “Introducing fakes into the canon messes up our understanding of history,” explains Hyman. “The Getty kouros statue is a famous example. Experts are divided over whether it was made in Ancient Greece or in the 20th century. And this makes it harder to tell the story of how Greek sculpture developed.” The implications of getting this wrong were so substantial that the Getty held a conference to work out whether it was genuine.

 In September The Bestseller Code: Anatomy of the Blockbuster Novel will be released

However, there may be an algorithm to prevent this too. Art Fraud Insights is working on a fake-busting algorithm for an online auction site, which would notice when a forgery was listed. The algorithm would spot bogus experts “authenticating” fakes online.

But what if a living artist started using algorithms to create new works? Titian was helped by pupils in his workshops; Andy Warhol had his factory; Damien Hirst’s assistants were responsible for most of the spot paintings that carry his signature. Would it be so different for an artist to outsource to an algorithm?

“You’d expect them to patent that algorithm,” notes Hyman. “Could there be a Damien Hirst algorithm? Why not?”

Follow Rosamund Urwin on Twitter: @RosamundUrwin

Create a FREE account to continue reading

eros

Registration is a free and easy way to support our journalism.

Join our community where you can: comment on stories; sign up to newsletters; enter competitions and access content on our app.

Your email address

Must be at least 6 characters, include an upper and lower case character and a number

You must be at least 18 years old to create an account

* Required fields

Already have an account? SIGN IN

By clicking Create Account you confirm that your data has been entered correctly and you have read and agree to our Terms of use , Cookie policy and Privacy policy .

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Thank you for registering

Please refresh the page or navigate to another page on the site to be automatically logged in

MORE ABOUT