Hidden Layer Interaction: A Co-Creative Design Fiction for Generative Models
Imke Grabe, Jichen Zhu

TL;DR
This paper explores a speculative co-creative interface that allows users to interact with the hidden layers of generative models, potentially enhancing human understanding and control over AI-generated content.
Contribution
It introduces the concept of hidden layer interaction and proposes visualizing and manipulating neurons within hidden layers to extend co-creation beyond input-output interfaces.
Findings
Proposes a new interaction paradigm involving hidden layers.
Suggests feature visualization for neuron manipulation.
Enhances understanding of generative model inner workings.
Abstract
This paper presents a speculation on a fictive co-creation scenario that extends classical interaction patterns with generative models. While existing interfaces are restricted to the input and output layers, we suggest hidden layer interaction to extend the horizonal relation at play when co-creating with a generative model's design space. We speculate on applying feature visualization to manipulate neurons corresponding to features ranging from edges over textures to objects. By integrating visual representations of a neural network's hidden layers into co-creation, we aim to provide humans with a new means of interaction, contributing to a phenomenological account of the model's inner workings during generation.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsData Visualization and Analytics · Artificial Intelligence in Games · Aesthetic Perception and Analysis
