98 Iterations presents a new form of production based on machine-learning techniques. From design to form this process is automated trough code and electronic devices, invoking a provocative question: Can a machine take over a designer’s creation process in addition to the production? The results of this investigation has lead to an artificial intelligence algorithm where a machine has learned to generate its own decorative objects, improving with each iteration. My input as a designer has been limited to namely that—to feed the machine with input to learn from.
This new form of production was presented at the Graduation show at Willem de Kooning Academy. The production line specializes in ornamental artefacts and gets its “inspiration” from 3D model sharing community Thingiverse where it automatically downloads models under the tags vase,bowl and bottle. These tags are pre-hand selected by the designer and entail its only design choice. As these tags determine the type of input that will be processed by the machine, the designer forfills the role of an input designer. This first and only step of the designer initiates a sequence of operations that ultimately leads to its final product. After the tagged models have been donwloaded, they are converted into binvox files, a format based on 3D pixels (voxels). This format enables the machine to distillate shapes trough pixel formation, making it possible to recognize objects. The machine repeats this process 98 times, becoming more sufficient with each repetion. These training sessions are called iterations.
From what it has learned the machine now tries to generate its own form that suits within the visual characteristics of the models it has downloaded. This results in interesting and unusual looking objects that represent the machines abillity to learn and evolve. Trough 3D printing techniques, the generated designs are translated into physical artefacts where it becomes a tangible documentation of its evolution and obtains a function within our physical enviroment.
This new form of production was presented in real-time at the Graduation Show at the Willem de Kooning academy where the machine-made designs where 3d printed.
This project is a critical, speculative, but also a topical and practical experiment with the impact machine learning could have on my chosen profession. The working production line specialises in ornamental artefacts and the machine uses my input as inspiration to help generate its own objects, as well as reference material to critique its own creations. In this case, selecting tags such as vase, bowl and bottle entail my only consequential design choice. It raises the question if the designer is the soul artist of the installation or that its tool has become the creator. In a near future, will we design shape or shape a process that designs for us?
> 1/98 Iteration
> iteration 55, iteration 56, iteration 74, iteration 95
Judith van der Heiden
- Input Designer
juudithvdheicen@gmail.com
Credits:
A special thanks to Takumi Moriya for providing coding and guiding me trough the complexity of machine learning systems.
> iteration 55
> iteration 95
> iteration 52