I was invited to present my current practices as an artist and musician working with artificial intelligence in audio at two seminars at Hochschule für Gestaltung and Hochschule für Musik in Karlsruhe together with students from University of Bayreuth.
The presentation was mainly about methods of direct and realtime intervention in latent space employing nn~ which I tend to call “Latent Jamming” because of its explorative nature.
While preparing, I’ve had the time for some introspection on what has happened in the past 10 months. When I started working with RAVE, vschaos2 and other architectures, two questions were the main motivators:
- Can AI assist in making an artist’s particular sound aesthetics perceivable? Or: What can AI see and hear what humans cannot?
- How can AI be fruitfully implemented into a human centered creative process? Or: Can AI spark creativity?
I can say that working with these tools has formally disrupted my writing and production practices to an extent where I find a lot of creative inspiration alone in inventing new techniques to map and explore latent space – and this is merely the functional or technical part of working with them.
In regards of the aesthetic quality of the output of these explorations, I find that the most interesting results were created in an equilibrium of control and the lack thereof – e.g. techniques of provoking repetition in model hallucinations.
Three hypotheses
- AI models can be considered generative instruments. Their role in the creative process is defined by the amount of control we’re willing to give up.
- Our decisions and the extent of our influence on what the models see and learn during training both enables and limits their capabilities to surprise us with the results.
- And our willing to play with rather than control these models/ instruments might bear the most fruitful outcome in terms of a joint creative act.
Since this is a relatively new field of artistic practice, I’m super interested to hear other people’s input, interpretations, opinions and any kind of reflection on the matter. Don’t hesitate to get in touch!
The whole playlist of presentation videos explaining key methods of nn~ with RAVE models can be found here:
