Machine Learning Can Generate Appropriately Cryptic Artist Statements

We feature quite a few art installations here at Hackster, and there is a good reason for that. Modern kinetic and interactive art is a fascinating intersection of engineering and creativity, and the field is largely free from the constraints of practicality or profitability. The results, as we often show you, can be spectacular for audiences to enjoy.

But, one piece of the puzzle that is so often lacking is the artist’s description of their own work. These are the short paragraphs you’ve probably seen on a plaque below an art piece in a gallery. These usually contain flowery prose that is so vague it could apply to any other piece in the room.

To highlight the absurdity of this in the most meta way possible, Selcuk Artut built Variable. Variable is itself an art installation, and its entire purpose is to generate convincingly cryptic artist statements. Artut accomplished this using machine learning, and the algorithm was trained on Martin Heidegger’s Being and Time. This philosophy book provides all of the material necessary for a computer to create descriptions of art that are indistinguishable from the real thing.

Variable consists of eight screens which are connected to Raspberry Pis. When a visitor pushes the corresponding button, the system generates an artistic title such as “Movement.” Then, the machine learning algorithm takes over and comes up with a suitably intangible description. As you can see, Variable is perfectly capable of generating descriptions with just as much substance as any real artist’s.