muPoint

muPoint, a creative project designed in collaboration with Artificial Intelligence, explores the intersection of human cognition and AI through the fusion of augmented reality, machine learning, and real-time device sensors.

In this work, we trained an AI model on Filipe Rocha da Silva’s textile drawings, creating a dynamic filter for an augmented reality app. This experience reinterprets the slow and precise craftsmanship of textiles through the lens of the model’s training. Both these processes shape an algorithmic landscape – a space where form is not static but emergent, shaped by both computational logic and real-time human interaction.

Designed for Apple devices, the work uses real-time tracking and procedural geometry to generate digital natural elements, such as rain and wind. These elements react to natural light and reflect the environment onto the digital surfaces. As users engage, the virtual world adapts to their movements, creating a unique experience each time. 

The visual design of the website—square-format GIFs—closely mirrors how the machine learning model receives, processes and interprets visual data: as images cropped to a square ratio. Rather than perceiving linear, narrative video, the model “sees” in fragments—loops of motion, rhythm, and pattern that reflect the nature of its image-based training. The GIF format reinforces this logic, evoking not only the repetitive cycles of machine learning iteration but also the looping gestures inherent in textile craftsmanship, bridging digital process and material tradition.

At its core, Neural Landscapes intends to challenge perceptions of autonomy and authorship in the relationship between human, machine, and environment.