Last week, the first ever work of AI-generated art to be sold by a major auction house fetched US$432,500. The work, entitled Portrait of Edmond Belamy, is a depiction of a fictional character and sold for an astounding 45 times its upper estimate. However, many in the AI art community are frustrated such a basic example of algorithmic art has achieved this success.
The artwork was created by a Paris-based collective of artists called Obvious. Utilizing a type of machine-learning called Generative Adversarial Networks (GANs) the algorithm is designed to generate novel works of art.
"The algorithm is composed of two parts," explains Hugo Caselles-Dupré, one of the members of Obvious. "On one side is the Generator, on the other the Discriminator. We fed the system with a data set of 15,000 portraits painted between the 14th century to the 20th. The Generator makes a new image based on the set, then the Discriminator tries to spot the difference between a human-made image and one created by the Generator. The aim is to fool the Discriminator into thinking that the new images are real-life portraits."
Portrait of Edmond Belamy is one of 11 portraits in a series generated by the algorithm. The series is designed to resemble classical portraiture and encompasses a fictional family called the Belamys. The name is a reference to Ian Goodfellow, a machine learning researcher renowned for developing GANs. Goodfellow, roughly translated into French, is "Bel ami", hence the name Belamy.
Completing the final artwork, the 70 x 70-cm (27.5 x 27.5-in) portraits were printed and framed in a classically inspired golden frame. The artist's signature at the bottom of the artwork is a core component of the algorithm, printed in a cursive Gallic script.
Separate to the obvious massive question of whether a machine can ever be genuinely creative in the same way as a human, many in the AI art community have expressed frustration at Obvious's simplistic form of AI art becoming so successful. It has been suggested that Obvious's Belamy series makes no technical innovations in the field of AI, simply lifting other previously composed algorithms, printing out the work and signing it with code.
Caselles-Dupré, in an interview with Artnome, readily admits his team's work expands on a great deal of innovation from others, and agrees with many of the criticisms directed at the project.
"Some of the comments, we really agree with this," says Caselles-Dupré. "It's like, what we do is not that complicated, or it can be qualified as not really very original compared to what is done in the AI art community, and we totally agree with that because it was just the first project. It's the first project. You've got to start somewhere. And we started there."
While the Obvious project is not exactly the most innovative example of AI-generated art, it is still the first artwork to end up in a major auction house. Initially only expected to sell for between $7,000 and $10,000, the extraordinary final price does suggest major interest for AI art in the high-art auction world.
And, credit where credit is due, Obvious is the first group to infiltrate the art world with AI-generated art to this degree. Whether this is good art, or even art in any sense of the word, is a philosophical argument sure to be debated for years to come, but what we can be sure of is that this AI-generated portrait will not be the last machine-made portrait to hit the market.
Source: Christie's