Circuit Training: Machine-made Art for the People

"My interests are manifold and in constant evolution, involving artificial intelligence, deep learning, generative and evolutionary art, glitch art, data classification and visualization or robotic installations."

Mario Klingemann

Module 1: Acquisition

The first module in the circuit serves as the data input channel. Here the machine tries to learn about the visual appearance and the behaviour of humans. It consists of an empty white photo studio with a camera and a separate viewing area that will show the non-participants the data acquisition process in real-time.

Human interaction

The machine guides the participants through the process of collecting the images and biometric data like face markers, pose data or voice samples.

Presenting the data

The analysis of the data acquisition is presented on a screen outside the photo studio visible only to the waiting audience.

It will show the video feed together with the acquired biometric data. It will also show additional search processes, like similarity rankings with previously acquired data "most similar face", "estimated age/gender/mood" as well as novelty scores for poses or faces. It will also try to measure and display the current biases it has in its data set by analyzing clusters of similar looking humans, estimated gender, age and ethnicity.

Module 2: Curation

In the second module the machine tries to learn the preferences of its audience by collecting feedback on its training process. It continuously presents the progress of its learning process and allows visitors to rate the results.

Human interaction

On each screen the participants can decide if they like a certain generated image or not by the well-known Tinder left/right swiping mechanism. Every vote will influence the training process in which it tries to optimize its output to maximize its likeability.

Presenting the data

Since the machine tries to maximize the amount human attention it gets it would be logical that at this stage it also publishes its art on various social channels, like Twitter or Instagram. Doing this it could collect more data on how its creations are being received by measuring views, likes or retweets.

"If there is one common denominator it’s my desire to understand, question and subvert the inner workings of systems of any kind. I also have a deep interest in human perception and aesthetic theory."

Mario Klingemann