The ghostly portrait Edmond de Belamy, created by the French collective Obvious, was the first work produced by an artificial intelligence auctioned by a large auction house. The vaporous image proposed in October 2018 by Christie’s of New York was valued at around $ 10,000, a trifle in this speculative overheated market for decades. Its inclusion in the catalog wanted above all to test the interest of collectors.
The ultra rich have responded beyond expectations. The portrait has been auctioned for 50 times the estimate, relegating to the background the results of the works of true great masters. Obviously, Obvious had just launched a small revolution that paid off.
“The portrait made a lot of noise”, humbly acknowledges Gauthier Vernier, one of the three co-founders of Obvious with Pierre Fautrel and Hugo Caselles-Dupré. Forbes France did not go with the back of the keyboard, calling them “French geniuses”.
“We went to school and high school together,” explains Vernier. We each did our studies. We have remained very friends. We decided to set up a creative project together. Suddenly, Hugo, who is now a doctor of artificial intelligence, showed us what algorithms can do. We were very surprised that these abilities weren’t better known. We therefore imagined our own project against the backdrop of the question of the ability to create algorithms. “
The image Edmond de Belamy was “generated” as part of a serial production after the analysis by artificial intelligence of some 15,000 portraits painted between the XIVe and the XXe century. The program first mastered roughly what a portrait is since the European Middle Ages before getting down to business by increasingly refining its “style”, which is ultimately printed.
The result evokes a unfinished, or the product of a sponge rubbed on a wet oil of the XVIIe century, or the failed restoration of an old fresco. Some pixels visible here and there betray the real origin of the signed canvas min G max D x [log (D (x))] + z [log (1 – D (G (z)))], or the name of the mathematical formula which guided the machines.
The equation used for the pioneering portrait of 2018 itself adds deep questions to this already unusual work. Obvious used an artificial intelligence “recipe” posted online by young Robbie Barrat, who did not touch the sale three years ago. The success sparked a controversy around the definition of author, copyright, ethics and aesthetics, of course.
Obvious followed with Japanese prints, African masks, street art variations, a recent project by a Marianne, a synthesis of thousands of photos of current French women to represent them all, “in their diversity as well as in their consistency”.
Mr. Vernier speaks of “AI art”, in the English style, in reference to artificial intelligence, to describe these various creations. Others prefer to say “digital art” or “crypto art”. “We were pretty pioneers in the world and only a handful of creators were doing what we were doing,” he says. Today there are thousands of artists who work with algorithms to make creation. “
Machinart
The emulation between art and machine obviously does not start with the macBook. During the Renaissance, the invention of the camera obscura already seemed to threaten the artist’s exceptionality, just as photography allowed a certain automation of the representation of reality. The avant-gardes of the early twentiethe century, often fascinated by technique, continued experiments with tools for capturing reality made overpowering in the digital age.
“For me, there is a difference between, on the one hand, computer-aided creation, which offers one more means, and, on the other hand, the use of algorithms, of programming to open up on new possibilities to stimulate and create meaning, ”summarizes Mehdi Benboubakeur, Managing Director of Printemps Numérique, a Quebec organization that democratizes issues related to the technical revolution.
The exhibition Towards an imaginary digital presented this fall at the UQAM Design Center questions the role of this tool “in the formal conception process in design and architecture”. Google’s DeepDream Generator generates psychedelic images using sarongs created by algorithms. The result is frankly quite boring.
Emmanuel Durand, co-director of research for Metalab at the Société des arts technologie (SAT) in Montreal, cites the example of a new Nirvana-style piece composed by artificial intelligence. “It’s interesting,” he said. But what we are not told is that, potentially, 150 other songs that performed much less well were generated and discarded. “
Artificial intelligence-assisted creation is nevertheless achieving more and more astonishing (and sometimes worrying) feats in all artistic fields. In recent years, we have seen and heard musical compositions, film scripts, texts books, poems and even cooking recipes produced by more or less autonomous software. Deepfake easily imitates photos, videos, sound recordings. Algorithms produce fake journalistic texts and malicious intoxesat will.
“Artificial intelligence is very present in society, but it is not necessarily wanted,” says Durand. Police departments use AI to classify passers-by. It is a subject in itself for artists. “
Terminator and creator
It is one thing to create, copy, divert an image with a computer or a printer. Gauthier Vernier asserts this lineage in the reinterpretation. “The variation from known references in traditional art allows us to reach as many people as possible,” says the co-founder of Obvious. Instead of asking ourselves if the algorithm can create completely autonomously, we show that it is able to create a classic table and then compare it with what humans do. “
To imagine a machine creating by itself, in a completely autonomous way, yet here is a completely different problem which seems to tip into another universe.
The Terminator-style security fears are well known. Stephen Hawking, Bill Gates, and Elon Musk have all pointed out the terrifying potential dangers of artificial intelligence. But should we be concerned if autonomous robots come to generate grand masterpieces and symphonies instead of shooting at us?
This possibility (conceivable at least theoretically) refers to the nature of the creation, of the art, of the work. The philosophy of digital arts is already looking at this problem, which evokes the Turing test to qualify a machine as conscious.
British scientist Simon Colton is leading the project The Painting Fool. The assumed objective is to make the creation of this crazy machine completely autonomous and to have it recognized as an artist in its own right. On its site, this artificial intelligence in need of recognition presents itself to the “I”. “I was built to demonstrate behaviors that could be considered talented, esteemed and imaginative,” says Machine.
“The question of identification seems central to me,” says Nicolas Bouillot, another co-director of Metalab from the SAT. Just the fact that it is a machine that produces will not put off the public, or at least part of the public? If a machine writes a high-level work, its appreciation will be up to everyone’s taste. And many people prefer to feel a human behind the works. “
His colleague is still thinking about it. “At what point is art art? asks Emmanuel Durand. Is it from the moment we take pleasure in the experience? If this is the case, the day when we can generate films with machine learning, it will perhaps be very pleasant to view these works. And why not ? “
Mehdi Benboubakeur, for his part, proposes, in closing, the idea that an artistic creation, to be conceived as such, must be new, surprising and rich in meaning. For now, machines seem especially adept at understanding patterns of production, when the personal and social complexities necessary to create elude them.
“The deep learning used by Obvious for portraiture Edmond de Belamy is very complex, says the director of Digital Spring. The result is new and surprising. But is it rich in meaning? “
He reminds us that an artist is anchored in a community, relationships, an audience, issues. This human is human, what. “The resonance between his creation and his universe means that the work tells a story, takes a stand, speaks of experiences and feelings,” he says. An algorithm is not so lucky to exist in an organic environment. I don’t think an algorithm is going to cut an ear… ”