[ AI ] dancer is a dance-video installation investigating on the potential of training a computer programme to learn dance improvisation. Choreographer Evangelos Biskas and creative coder Mauricio van der Maesen combine their disciplines, both looking for connections between the mechanics of a dancer’s body and of artificial intelligence.
Mauricio creates an algorithm which is trained by Evangelos’s skeleton tracking data out of a large database of dance video material divided into numerous short movement sequences, here are some examples:
The algorithm predicts what is the next best sequence to follow the previous one, re-composing the dancer’s improvisations based on its training. The training visually translates the information by taking into consideration the keypoints of the skeleton in response to music:
In the live version of the installation, the audience will chose what they want to see on the screen (skeleton, dancer, both ec.) and test how the dancer and the program interacts to different music. They will also have a pool of effects that they can manually select, creating their own [ AI ] dance improvisation:
In the future, Evangelos and Mauricio want to develop this research to a live performance installation which will reverse the process and ask the question; What can the body learn from artificial intelligence?