Sophisticated algorithms allow Argentinian artist Sofia Crespo to invent insects, while American Robby Barratt can modify the characters of classic art paintings, two examples of the revolution that artificial intelligence-assisted painting can bring to visual artworks, according to its advocates.
It's "like a human-machine dance," says Jason Bailey, a collector and one of the most popular crypto-art bloggers.
The majority of these digital artists use supercomputers and software known as generative adversarial networks.
Two types of artificial intelligence algorithms called "neural networks" compete to provide the most complete picture of the artist. The human artist first provides the source images and adjusts the settings to get the desired result from them.
Sophia Crespo, 30, uses these algorithms to recreate animals. The goal is "not to avoid real nature, but to establish contact with nature through a medium in which we spend a lot of time, which is the digital medium," she told AFP in a video interview from Lisbon.
Crespo insects are very realistic, with antennae, wings and bodies that look like they came from an entomology book, but what they're all missing are the heads, and their bodies seem to have undergone multiple genetic mutations.
Intelligence that needs direction
The astonishing progress of artificial intelligence points to a world in which computers can learn and create like humans. But for now, AI still needs guidance, and Sophia Crespo's bug series requires a great deal of effort and revisions to match the models the artist wants with those produced by neural networks.
"The computer is an integral part of the creative process", notes Camille Langlois, curator of the Pompidou Museum in Paris, but believes that "the ability to create realistic images does not make everyone an artist", as artistic-quality images require "a critical and creative ability".
Like Sofia Crespo, Robbie Barrat began his work around 2018, inserting thousands of classic artworks into his computer and dialogue with the machine until he got what he was looking for: a series of amorphous busts, combining the styles of Salvador Dali and Francis Bacon. .
"When I work in this way, I don't create an image, but a system that can recreate images. In a way, I create a tool," he explains.
This 22-year-old has a great proficiency in digital art, and one of his works was sold last March at an auction by Sotheby's for more than 700,000 euros.
The problem so far has been that "the inputs are text and the outputs are images," explains Ugo Cazel Debray and Gauthier Vernier of the Opvius group.
The biggest difficulty was getting the computer to be able to take in huge amounts of images with their descriptive texts.
This gigantic task can only be accomplished with well-funded projects such as the Dall-E 2 model from California startup OpenAI that is privately funded by billionaire Elon Musk, or the Imagine project. (Imagen) competitor from Goal Research.
Based on a single sentence, a computer with all this information is able to mix concepts and create multiple representations of "two radishes on a skateboard" or a "monkey astronaut".
Sophia Crespo, who tested Dal-E2, said it was "best in terms of generating images overall". Experts believe that these programs can revolutionize the image creation and retouching industry. On social networks, several examples are already being circulated.
In contrast, OpenEye and Google have not yet published the consumer tool, or brought it to market because of the risks of harmful use it entails.