Modem collaborated with Google DeepMind and Studio Ross Lovegrove to explore the potential of AI as a collaborative design tool. The project materialized as a metal 3D-printed chair, born from a custom model fine-tuned on Lovegrove’s sketches.
The project explores new design strategies through human–machine collaboration. By positioning the model as a partner, the workflow creates a feedback loop between human intuition and algorithmic logic. The model did not replace the designer, but augmented the creative process, offering thousands of form iterations based on specific semantic prompts.
The team fine-tuned a Google DeepMind model on Lovegrove’s extensive archive, building a tool fluent in his biomorphic style. To avoid predictable outcomes, prompt strategies deliberately excluded the word “chair,” forcing the model to rely on semantic descriptions of form and structure. From thousands of iterations, “Seed 6143” was selected as the definitive output.
Seed 6143 was transformed from a 2D generation into a complete 3D silhouette and CAD master model. The form was sliced and subdivided for direct robotic-arm printing. Built in continuous layers of metal, it developed a tactile surface reminiscent of growth rings. This approach maximized continuous flow while keeping the number of parts to a minimum.
Seed 6143 demonstrates a full pipeline across dataset design, model behavior, prompt logic, CAD translation, and robotic fabrication. It frames generative AI as a disciplined collaborator within industrial design, not a shortcut for aesthetics. The result honors Lovegrove’s principles while proposing a geometry that human intuition alone might never have conceived.
Seed 6143
Partner: Google DeepMind
Vision & Design: Studio Ross Lovegrove
Design & AI Strategy: Ila Colombo
Strategy & Direction: Modem
3D Fabrication: Oxido Studio
Photography: Enric Badrinas
Art Direction: Ilya Schulte