RNN Descriptions

I maintain a poorly trained, semi-wild recurrent neural network to act as inspiration for this artwork. This RNN has been fed descriptions of thousands of artworks and is then run through multiple output cycles and re-trained on its own output. Only the freshest, most absurd artifacts are kept for the next round of training. The result of this process is a neural network that will generate artwork descriptions that are interpreted here.

Additionally, Thorin Klosowski and I have an ongoing collaboration wherein he writes stories based on descriptions from the RNN and I sometimes interpret those into artwork. He has written an excellent FAQ that does a much better job of describing what this is and how it works here. And subscribe to the newsletter here!