There is confusion about who should get credit for the artwork the machine learning model borrowed from and who should get credit for the prompt the machine learning model generated. The Museum of Modern Art has a new exhibition on display this Saturday and you can give it to the first three. The last part is not completely clear.
The exhibit called Refik Anadol: Unsupervised will run through March 5, 2023. The artwork uses an artificial intelligence model designed by Anadol himself and 380,000 images of art pieces from MoMA to create a stream of moving images. According to a museum release, the new work on display is a singular and unprecedented meditation on technology, creativity, and modern art which is focused on reinventing the trajectory of modern art, paying homage to its history, and dreaming about its future.
Unsupervised was a part of his Machine Hallucinations series that he started in 2016 and was originally published on the platform Feral File. According to the project description, Anadol has used several different kinds of generative adversarial network (GAN) to create the generative artwork. The images were sold as NFTs, which is something MoMA has been investigating. According to Fast Company, the team that worked on the artificial intelligence that powers Machine Hallucination also trained using an Nvidia DGX Station A 100, a desktop-sized box used for computing.
MoMA said the in-person exhibit also monitors changes in light, movement, volume of the surroundings and even the weather which has an impact on the moving image.
MoMA confirmed in an email with Gizmodo that the artist had been working with the museum since early 2021, before the original Unsupervised art went up online. The work is a direct result of MoMA’s publicly available GitHub.
Stable Diffusion and Openai’s DALL-E 2 are examples of models used to power artificial intelligence art generators. Diffusion is an iterative process that uses machine learning systems to attempt to generate realistic visual images while using a kind of gaussian blur to distort and then reconstruct images. GANs are more of aversarial system that uses a discriminator to decide whether a part of the image belongs. Anadol’s GAN model let the artificial intelligence generate its own images without being watched.
According to the release, a leader in organizing the exhibit said that artificial intelligence is often used to classify, process, and generate realistic representations of the world. Anadol has a visionary work. It explores dreams, hallucination, and irrationality, posing a different understanding of modern art.
In an interview with Fast Company, Anadol said that this kind of image generator was becoming its own entity.
This new art installation may benefit from the fact that it isn’t using the most high-profile kind of image generation. The biggest names in artificial intelligence art generators have been the subject of controversy, with many of them coming from traditional artists. These systems have used millions of images from the internet without permission to train the models. Artists fear that their art styles have been coopted by image generators, leading to fears that their work could become indistinguishable from people who simply type in a prompt such as: “art in the style of Picasso”.
The question of whether an individual could ever claim to own art that was created using an artificial intelligence system that had generated images from other copyrighted work is sticky.
MoMA’s senior curator said that the new exhibit “underscores its support of artists experimenting with new technologies as tools to expand their vocabulary, their impact, and their ability to help society understand and manage change.”
