Refik Anadol — “Unsupervised (Machine Hallucinations — MoMA)”

⭐ What the project is

“Unsupervised” is a large immersive art installation shown at the Museum of Modern Art (MoMA). It uses machine learning to “dream” new artworks based on the museum’s collection — showing what a machine imagines after seeing thousands of artworks.

Visitors see constantly changing abstract visuals projected in a huge space, generated from AI analyzing art history.

👉 Core idea:

What would a machine imagine if it experienced all of human art?

🧠 What type of machine learning models were used?

The project uses generative deep learning models:

• DCGAN (Deep Convolutional GAN)
• PGAN (Progressive GAN)
• StyleGAN

These models:

• learn visual patterns from large image datasets
• generate new synthetic images
• create continuous variations of style, color, and structure

How they work (simple explanation)

GANs contain two networks:

• Generator → creates fake images