Creativity and art are usually thought of as the domains of humans.
But computer scientists from Rutgers University has designed an algorithm that shows that computers may be just as skilled at critiquing artwork. By judging paintings based on their novelty and influence, the mathematical algorithm selected the most creative paintings and sculptures of each era.
The study, published in arxiv, found that more often than not, the computer chose what most art historians would also agree are groundbreaking works, like Edvard Munch’s “The Scream” and Pablo Picasso’s “The Young Ladies of Avignon.”
Scroll down to see which paintings made the cut, and why.
The algorithm's network included over 62,000 paintings spanning 550 years and some of the most well-known names in art history, from the Renassaince era to the age of pop art. This painting by Lorenzo di Credi is often called the Dreyfus Madonna, after Gustav Dreyfus, one of its longtime owners.
The paintings were arranged on a timeline according to the date it was made, so each painting could be critiqued with a historical point of view. The algorithm looked for paintings that differed from the work that came before to measure its novelty. This fresco mural by Andrea Mantegna decorates one of the walls in a castle in Mantua, Italy.
The computer algorithm also weighed how influential each painting was by looking at paintings that imitated its style. Leonardo da Vinci painted this portrait of St. John the Baptist late in his career, leading an artistic era called Mannerism, which is characterised by exaggerated poses.
The network also included sculptures, like this marble statue by Michelangelo. He worked on this up to the last days of his life.
Computer vision scientist Ahmed Elgammal and his colleagues examined only Western art, so religious images take up the bulk of the subject matter of the early paintings, like Diego Velazquez's 'Christ Crucified.'
The algorithm could be changed to focus on different characteristics, like subject matter, colour, use of perspective or even texture, said Elgammal. At the time that Johannes Vermeer painted this view of his hometown, paintings of cities were not popular, so it would have scored quite high in novelty.
Elgammal and his colleagues determined that subject matter and composition were the most important factors to consider. Art historians and critics have long discussed the subject matter and symbolism of this Johannes Vermeer painting of a woman balancing a scale.
This depiction of Jesus' crucifixion by Francisco de Goya almost a 150 years after Velazquez's 'Christ Crucified.' It scored higher than other paintings of its era and on originality, but low on influence.
This 1865 painting of haystacks by French impressionist Claude Monet predates a series of 25 paintings of the same subject matter. Monet's later haystack series, painted in the 1890s, weren't included in the algorithm to ensure that this painting's influence score hasn't been exaggerated.
The algorithm seemed to favour paintings done late in an artist's career. Vincent Van Gogh painted this view of the asylum where he stayed for a year toward the end of his life.
'The Scream' by Edvard Munch is one of the few outliers that scored extremely high in its era. According to the study, 'this painting is considered as the second iconic figure after Leonardo's Mona Lisa in the history of art' and is one of the 'most reproduced paintings in the twentieth century.'
Elgammal, a computer vision scientist, said that the algorithm was designed with the goal of creating a computer that has human-like creativity, a subset of artificial intelligence called computational creativity. He said a computer that is creative enough to pass as human should be able to assess its own and other's creativity. This early Gustav Klimt poster was made at the beginning of a movement called the Vienna Secession scored relatively high in its period.
Pablo Picasso's 'Young Ladies of Avignon' is one of the seminal works of early Cubism. The computer algorithm ranked it as the top scoring painting from 1904 to 1911, Algammal said.
Picasso made this cardboard and rope sculpture five years after 'The Young Ladies of Avignon.' The algorithm ranked it as the most creative piece of cubist art.
Cubism dominated early 20th century art. Believe it or not, this Piet Mondrian painting is of a tree, in classic cubist style.
This simple Cubist painting called 'Red Square' by Russian painter Kasimir Malevich ranked off the charts in creativity.
Elgammal also tested the algorithm by running a series of 'time machine' experiments. He moved paintings back and forward on the timeline to see how the creativity scores changed. He found that moving paintings back increased creativity scores, while moving a painting forward decreased it. This Kasimir Malevich painting, like other cubist paintings, would have seen an average increase of 89% in creativity when moved back to the 1600s, according to the study.
This 1919 Georgia O'Keeffe painting of what looks like an enlarged flower was extremely creative for its time, paving the way for an artistic movement called American modernism.
Elgammal also ran the algorithm on a separate set of 5,200 religious paintings made during 1410 to 1993. The creativity score of this 1957 lithograph by Marc Chagall scores off the charts.
Roy Lichtenstein's work defined the American pop art movement during the 1950s and '60s, along with artists like Andy Warhol. This still life of 'Bananas and Grapefruits' ranks the highest in its era.
Business Insider Emails & Alerts
Site highlights each day to your inbox.