Research and find a project (experiments, websites, art installations, games, etc) that utilizes machine learning in a creative way.
The project series “Learning to See: Gloomy Sunday” is an interactive art installation created by Memo Akten, who is a Turkish-born artist, researcher, and musician known for exploring the intersection of art, technology, and human perception. The project was first exhibited in 2017 and has since appeared in multiple exhibitions and online performances. In the installation, Akten sets up a simple webcam pointing at a table. On the table, he moves around everyday objects, for example, fruits, cups, books, cables, or his own hands. And the live footage is passed through a neural network trained to interpret the visual world through a very specific aesthetic lens.
The result is that a banana might suddenly appear as a twisted structure in a foggy landscape, or a hand might turn into part of a surreal painting. The machine is not “recognizing” the objects in the way a standard AI model would, instead, it’s hallucinating based on what it has learned from its training data.
Memo Akten curated a collection of paintings, photographs, and textures that reflected a very particular mood. The training data included:
By training the model exclusively on this dataset, Akten effectively gave the machine a “visual vocabulary” and “emotional palette.” When it sees new input, it uses this learned vocabulary to interpret the world according to its own learned dataset.
Memo Akten intentionally used a machine learning model because Akten might have wanted to question how machines “see.” The project invites the viewer to reflect on how much of what we call “perception” is shaped by the data we’ve been trained on, and every change in the webcam input leads to a new hallucination. This creates an interactive, improvisational performance between human and machine.
HandPose Project—stretches a “😯” emoji
