Blind Camera

Diego Trujillo Pisanty

Artwork description:
Blind Camera is an AI-powered device that creates images from sound instead of light, mimicking a point-and-shoot camera experience. By aiming the horn at a sound and pressing a button, sound is converted into an image, framing scenes with ears rather than eyes. The resulting 'photo' represents the surrounding soundscape rather than the visual scene.

The project employs a custom-made artificial neural network (ANN) to link sound and image, similar to how 'dog' can refer to both a picture of a dog and its bark. This ANN was trained using bespoke videos from Mexico City, with each frame linked to the preceding second of sound. The training involved encoding the sound into a vector, decoding it into the corresponding image, and convincing another network that the generated image is a photograph.

Blind Camera thus integrates elements of Autoencoders and Generative Adversarial Networks. Limiting the training data to Mexico City imparts a local perspective. Unlike most AI, which aims to reduce bias, Blind Camera embraces its identity as a Mexican urban object, interpreting sounds from this unique viewpoint. This approach invites speculation on other non-global and underrepresented perspectives that AI could take. For an artist, developing an ANN and its training dataset was daunting but essential for artistic control over the image aesthetics and narratives. This process allowed the creation of code that foregrounds artistic intent. It also prompted reflection on how artist-created AI might differ from existing models, highlighting the potential for diverse, artistically driven AI innovations.