- Title: Could A.I. be the next Rembrandt?
- Date: 20th September 2018
- Summary: (SOUNDBITE) (French) CO-FOUNDER OF OBVIOUS COLLECTIVE AND PHD STUDENT IN MACHINE LEARNING, HUGO CASELLES-DUPRE, SAYING: "We were going directly to galleries in Paris. We were presenting our work, trying to talk to gallerists, and they laughed in our face, saying 'What you're doing is nonsense.'"
- Embargoed: 4th October 2018 15:29
- Keywords: artificial intelligence algorithm art Christie's Obvious computer fine art portraits
- Location: PARIS AND JOINVILLE-LE-PONT, FRANCE
- City: PARIS AND JOINVILLE-LE-PONT, FRANCE
- Country: France
- Topics: Art,Arts / Culture / Entertainment
- Reuters ID: LVA0058YFMQ1L
- Aspect Ratio: 16:9
- Story Text: What if the next Rembrandt is a computer algorithm?
Obvious, a collective of Paris artist-entrepreneurs, is using artificial intelligence to create fine art that they hope would one day be hung alongside Renaissance portraits in prestigious galleries. One of their creations will be auctioned in Christie's in October.
The author of the art works is an algorithm called the Generative Adversarial Network (GAN), which learns to generate new images by being fed a database of samples, without replicating any one of the individual pictures.
"Can a machine be creative?" is the question they've thrown not just amongst themselves but to their audience.
Last year, Obvious launched a project which applies GAN to a database of 15,000 paintings that were available in the public domain. In turn, the learning tool churns out portraits of noble-looking figures, that the trio baptised with titles such as the Countess, the Marquis or the Baron of Belamy. After some adjustments on contrast or brightness, the images are then printed out on canvas, with durable ink that resembles paint.
"We are artists with a different type of paintbrush. Our paintbrush is an algorithm developed on a computer," said Hugo Caselles-Dupre, a computer engineer, who co-founded the group with his childhood friends Gauthier Vernier and Pierre Fautrel, who both have a business background.
Obvious populated their database with portraits from the 14th to the 20th century, saying that such a genre would be the easiest for drawing comparisons between paintings done with human hands and those by artificial intelligence.
"It gives a reference point to see what the machine is capable of," Gauthier Vernier, one of Obvious' three co-founders, said.
As the algorithm has not yet been perfected, the portraits are fuzzy: The Baron's nose looks disfigured, the Cardinal's face is murky.
Paris experts first laughed at the group's art works, the trio said, until they found their first buyer in Nicolas Laugero-Lasserre, an art collector and director of a cultural management school.
"What was astonishing was that they knew nothing about art, none at all. In the beginning, I took them for crazy people. And finally, are they crazy, are they genius? We'll see," he said.
Laugero-Lasserre displayed the portrait he purchased for around 10,000 US dollars, 'The Count of Belamy', at the Art42 Gallery in Paris. The painting shares a space with his collection of street art, hanging inside a classroom where computer engineering students learn coding.
Researchers are continuing to develop the GAN algorithm to make more realistic images. But the trio of Obvious say it is not their goal to create perfect portraits, but rather to educate people about how artificial intelligence works.
Some artists are not warm to the idea of calling algorithm-generated creations "art".
"If there was no anger from Picasso, the Guernica would never have existed... There's always a feeling behind a painting - whether it's anger, yearning, desire. And artificial intelligence is - well, you have the word 'artificial' in it," painter Robert Prestigiacomo said.
While Parisian galleries sneered at Obvious' creations, Christie's in New York approached the group and proposed to sell their portrait "Edmond of Belamy" as the first algorithm-authored oeuvre to be auctioned. The estimation is between 7,000 to 10,000 US dollars.
Pierre, Gauthier and Hugo's experiment is only in the nascent stage. They are looking to build new databases, so the algorithm can perhaps create sculptures, or traditional Japanese art. - Copyright Holder: REUTERS
- Copyright Notice: (c) Copyright Thomson Reuters 2018. Open For Restrictions - http://about.reuters.com/fulllegal.asp
- Usage Terms/Restrictions: None