Safeguard is an autonomous live-coded performance. The performer is a machine learning algorithm, ​Cibo​, trained to perform (read, change, execute, and repeat) Tidal code. ​Cibo was trained on recorded Tidal code performances by the authors and relies upon neural network architectures that are employed for various natural language processing (NLP) problems such as language translation and interactive dialogic agents (chatbots). As such, the current implementation of Cibo focuses on the textual performance of code. ​Safeguard demonstrates the ability of the machine to learn code patterns reminiscent of the human performers it is modeled on while also maintaining a high degree of unpredictability. This performance by ​Cibo is at times schizophrenic while also retaining a distinct aesthetic style, periodically returning to earlier themes while continually producing new content.

The primary intention of this project was to develop a collaborative performance partner, with secondary objectives of determining if the algorithm could be ​taught to perform in the likeness of specific live-coders, and could the algorithm become sophisticated enough to make ​learned aesthetic code decisions. This performance is the first step towards those ends. ​Cibo​ will perform/demo (you the audience decide which) by itself.

We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.

  • Year: 2019
  • Media: Tidalcycles, PyTorch