Latent Voyage invites you to trust the intelligence of the body as you navigate the hallucinatory space of a generative AI model.
The image is determined entirely by the pose of your body.
Image generators like Generative Adversarial Networks (GANs) can create an immense range of uncanny images. But they're controlled via hundreds of entangled parameters which don't follow any logical system of reasoning. But as embodied beings, we also have the intelligence of the body.
My hunch is that, similar to riding a bike, we can feel our way through the latent landscape of an AI Model, even if we can't really explain what we're doing. Just like riding a bike.
In the Sonified Body project, Panagiotis Tigas and I worked with dancers to train AI models that function as multi-dimensional interfaces, what we call Latent Mappings.
As the interface is trained on how a human moves, it is optimised for a human's movement capabilities. It senses the entire body and maps it to a complex and entangled parameter space. The black-box approach of Machine Learning means we can't explain how to use it by breaking it down. Instead, you have to play, explore and practise.
Latent Voyage speculates on what interaction with computers might be like in a future where interfaces are defined by AI trained to optimise human expressivity.
It's part of research into devising complex interactions with machines while retaining the embodied state of mind we bring to dance and social connection.
For a technical overview of the mapping see our NIME paper on Latent Mappings.
For a more in-depth philosophical exposition of what we're trying to do, see our journal paper Emergent Interfaces: Vague, Complex, Bespoke and Embodied Interaction between Humans and Computers.
Events
Development residency
- 7–13 June 2021, Cove Park, UK.
Interactive Installation
- 27 Nov 2021, Mixed Reality Show and Tell, Surge, Glasgow
- 28 Jan 2022, Centre for Contemporary Art, Glasgow
- 3–5 Jun 2022, Electromagnetic Field festival, Ledbury, UK
Acknowledgements
Latent Voyage developed out of the Sonified Body project, which was created in collaboration with Panagiotis Tigas, produced by Feral, supported by Creative Scotland and Present Futures festival.
It uses the StyleGAN2 and StyleGAN3 models from Nvidia.
Technical production by Preverbal.
