top of page
portfolio anne arzberger
THE DRILL-HAIRDRYER
Ready to try out something new?
What can exist between a drill and a hairdryer
Joining forces with our artificial partners, we developed a process that helps us to envision the space between and beyond categories. While we often wrongfully assume that AI is thinking as binary as we are, the opposite is true.
Curate your data
We curate a dataset of 300 drill images and 100 hairdryer images. It is important to make the distinction between a primary and a secondary dataset. They should be visually related, yet carry different social meanings and roles.
Train a StyleGAN on primary data
Once we curated our data, we can train the StyleGAN for 5-7 hours on our primary data. The model will learn to create images of drills.
Train secondary data on drill-StyleGAN
In a second step, we train our model, that is already familiar with drill data, just for half an hour on our secondary data of hairdryer images.
We will find objects that show familiarity with drills, however they started adapting hairdryer aesthetics, rendering the drills functionality useless.
Collect inspiration
With plenty of visual inspiration it is now time to start designing the actual toy. How closely you relate your concept to the AI stimulus is up to you. However, keep trying to break the stereotypes!
%20(1).jpeg)
.jpeg)
.jpeg)
.jpeg)
Prototype design ideas
Prototype the ideas that work best for you, and explore them in 3D. Lasercutted or 3D printed models can allow for quick testing of different ideas.
Finalize Toy
bottom of page