manifesting latent space

Latent space is a hidden, liminal part of the neural network’s algorithm. It is often concealed from the user, while it is exactly this space where the structure and logic of trained neural network manifest themselves, a representation of how an algorithm interprets the given data. It harbours biases and ideological artefacts introduced with the training data set. This project invites participants to explore the latent space by playing with it, in order to make it visible and palpable, accessible for interaction to a human user. It simultaneously poses a question: who can play with the structure of powerful commercial/corporate neural networks, and who is affected by the consequences of these games?