The Museum of Modern Art (MoMA) announced AI-generated imagery as its latest exhibition piece. The work just unveiled by MoMA (Refik Anadol: Unsupervised) is a huge installation piece measuring 7.3 × 7.3m. It is built using an AI model developed by Refik Anadol. The basis for producing works by AI is 380,000 images from 180,000 collections owned by MoMA.
According to MoMA, it is an unprecedented meditation on technology, creativity and contemporary art, focused on re-imagining modern art, paying homage to history and dreaming of the future.
The artist released unsupervised digitally in 2021 as part of a series (Machine Hallucinations) that began in 2016. The details of this work say that the art was created from publicly available resources using several types of GAN algorithms an adversarial production network for the production of the work. After that, the work was sold as NFT. According to reports, the AI used in his series was created using Nvidia’s AI device, Nvidia DGX Station A100.
Unlike those released digitally, you can enjoy the way the images change, such as light and movement, ambient sounds, and the weather, while viewing them realistically in the museum. The artist says she has been working at MoMA since 2021. However, the image used in the non-inviting director, presented in 2021, is said to use Github meta data that MoMA releases.
When it comes to images generated by AI, things like Stable Diffusion or Dali 2 come to mind these days, but generating images through a GAN algorithm is a bit different. Although both are said to distort, blur and reconstruct images, static diffusion uses machine learning to perform an iterative process in creating a realistic image, while GANs use discriminators to base adversarial systems on how to decide which parts of image to make. As a result, fixed diffusion is likely to become an annoying image mixture, but images created with GAN models are said to be completed as independent images. The work is due to be on display until March 5 next year. Relevant information can be found here.
