It’s a tradition in France: each president chooses his “Marianne”, the symbol of the French Republic that will then appear on postal stamps. For the next five years, the Obvious collective, which uses AI to create artworks, is planning a new Marianne more representative of the French woman: it is asking all women who want to be part of this project to send their photo and will create a new Marianne portrait thanks to a GAN (Generative Adversarial Network), an AI algorithm.

Pierre Fautrel, Hugo Caselles-Dupré and Gauthier Vernier, members of the Obvious Collective, are childhood friends and had the intention of working together. Hugo Caselles-Dupré, a researcher in artificial intelligence, discovers GANs, algorithms invented by Ian Goodfellow in 2014 that have produced a huge craze in the field of machine learning and made him famous, the collective decides to use them as tools for creation.

GANs as creative tools

A GAN is composed of two neural systems that confront each other. The generator generates a sample of images that it offers to the discriminator, which is responsible for differentiating between the images in the database and those produced by the generator. The three friends, art lovers, then decided to create classical portraits thanks to these algorithms able to replicate the creation process. One of their first paintings was purchased by the director of the Institute of Artistic Careers (ICART) and collector, Nicolas Laugero Lasserre, which made them known. In 2018, one of their works, the “Portrait of Edmond de Belamy” estimated at most 10,000 euros, was sold for $432,500 by the famous American auction house Christie’s. For this painting, the collective trained its algorithm with approximately 15,000 portraits painted between the 15th and 19th centuries. The signature, as with the other paintings produced by Obvious, is a mathematical formula.

“We are Marianne”

The Obvious collective, which has become famous abroad, thought about a project that could symbolize France and decided to create a Marianne representative of today’s French woman in her diversity. In an interview with RTL, Pierre Fautrel explained:

“Our work is rather known to collectors abroad but we wanted to work with France so we started thinking about different symbols of the country. Marianne stood out as a relevant choice because its current representation is a bit outdated, in our opinion. Many Marianne were chosen for their career or their plastic. We think it’s possible to do more representative things with digital tools.”

The Marianne project proposes that all French women of legal age who want to participate send their photo to the website noussommesmarianne.fr to allow the algorithm to draw the features of the next symbol of the Republic, ” a Marianne that is representative of all French women in their diversity as much as in their concordance,” before April 10. More than 5,200 of them have already sent a bust photo, on a white background and without glasses as required, but it is not enough. The Internet users will vote, among several Marianne that the collective will propose, for their favorite from April 17 to 30. The new Marianne, which will not be a photorealistic face but a stylized artistic representation, will be unveiled on May 1st. The collective hopes to propose it to La Poste to be printed on the next stamps, and then send it in physical form to the future president.

Translated from « Nous sommes (toutes) Marianne » : créer la femme française grâce à l’intelligence artificielle