The MoMA (Museum of Modern Art) in New York, where great artists such as Picasso, Van Gogh, Frida Kahlo and Kandinsky immortalized their souls in their paintings, has opened its doors to work that could revolutionize our understanding of art. Unsupervised is the name artist Refik Anadol chose for his work and it was created using artificial intelligence.
Measuring 7.32m x 7.32m, Anadol introduces us to the controversial use of artificial intelligence in art. While the great artists of the past measured every stroke to the millimeter, Refik Anadol presents us with a universe where neural networks take control of the work. Of course, at the direction of the artist.
Unsupervised work is hard to describe. Not only because of its changing appearance, but also because of the dynamism it offers to unexpected visitor looks. The work was also designed to interact with the environment. It is constantly adapting and changing, just like the room and the lighting.
Unsupervised represents the AI’s conquest of territory as hostile and purist as art.
Unattended by Refik Anadol
The best part of Unsupervised is the technology behind the screen. This time the credits don’t belong to any AI like DALL-E, Midjourney or Stable Diffusion; it is the artist himself who built his work from scratch. Refik Anadolu was tasked with developing his own artificial intelligence model, and it is on this model that the most basic functioning of Unsupervised is based.
The museum itself described the work as “a unique and unprecedented reflection on technology, creativity and contemporary art” aimed at “rethinking the trajectory of contemporary art, paying homage to its history and dreams of its future.”
But what exactly does Unsupervised show? The work is a collection of 380,000 images taken from 180,000 works of art in the MoMA collection. In this way, Anadol created a kind of collage, which is presented as movement and dynamics on a giant screen.
To create this AI model, the artist relies on Generative Adversarial Networks (GANs), which have also been used in many other image-based AI projects.
New artistic trends are recognizable by their dynamism.
Unsupervised live exhibitions are more than a movement. According to MoMA, the work is able to adapt to light, movement, the volumes surrounding it, and even urban climates. All this affects how it is presented to the thousands of visitors that the museum receives every day.
And where did the name “Homeless” come from? This can be understood as a reference to the functioning of the AI model. Unlike others like DALL-E 2 or Stable Diffusion, GAN-based models use “hostile” systems. They are responsible for creating believable data, which is then passed to the discriminator, which decides which part of the image each element belongs to. In this way, GAN models can generate their own “unsupervised” images.
The first two, on the other hand, are based on distortion and reconstruction systems that attempt to form more realistic images.
Of course, we have not dealt with the issue of copyright yet. Unsupervised is a collection of many works. Therefore, will anyone in the future be able to claim ownership of such a system? Another chapter in the artificial intelligence copyright black hole.
To display Hide table of contents