Computer Algorithm Identifies History's 'Most Creative' Paintings

                                                                 And for their next challenge, the algorithm's inventors will attempt to convince the art world that their definition of creativity should be the prevailing one. Once they have accomplished that, they will presumably build a causal model that correlates creativity with value.

At which point, all artists will then spend eternity painting the subjects that equate with the highest historical prices. Or, despite the interest of hedge fund squillionaires, Chinese real estate magnates and Russian oligarchs,  is it a  mistake to assume that art is a purely economic enterprise? JL

Daniel Culpan reports in WiredUK:

The researchers themselves concede these results are impossible to confirm or disprove -- the software is measuring a painting's originality rather than its intrinsic merit. In most cases the results of the algorithm are pieces of art that art historians highlight as innovative and influential.

A visual algorithm has been developed that its researchers believe can accurately rank historical art works according to their creativity, a study published in arxiv reveals.

Computer scientists Ahmed Elgammal and Babak Saleh from Rutgers University define creativity as "the originality of the product and its influential value", and used this definition to create a kind of "art network" based on how similar paintings are to earlier works. This barometer of originality, dubbed the "time machine experiment", looked at elements including everything from colour and texture to the type of scenes depicted.

The pair then applied these measurements to a database of some 62,000 paintings, and enabled the algorithm to draw parallels between these creative works, from more modern paintings to those from the distant past.

The paintings at the top of the chart are those judged by the algorithm to be more original, while those languishing towards the bottom have been rated to be more derivative. The results are interesting, if not altogether surprising; Edvard Munch's instantly recognisable The Scream is considered exceptional, along with Pop artist Roy Lichtenstein's bold Yellow Still Life and Monet's serene Haystacks at Chailly at Sunrise. In contrast, works by some of the Old Masters, such as Ingres and Rodin, slip down the list -- perhaps simply proving the age-old adage that beauty is in the eye of the beholder.

ADVERTISING

However, as the researchers themselves concede, these results are basically impossible to confirm or disprove -- and the software is measuring a painting's originality rather than its intrinsic merit. They commented: "In most cases the results of the algorithm are pieces of art that art historians indeed highlight as innovative and influential".

Elgammal and Saleh have said that the algorithm could be adapted in the near future to also analyse music and literature, but only time will tell if we'll one day turn to a machine -- rather than a critic -- to successfully rank history's greatest art works