Tinkering with AI Prompts and Cardboard Construction
Over the past few months I’ve been experimenting with adding AI tools and topics into tinkerable explorations of art science and technology with NEXUS Experiments at University Freiburg and other partners. As I’ve been exploring new ideas, I’ve been trying to keep the learning dimensions of tinkering in mind so that I can evaluate for myself and workshop participants how AI affects the process of developing new ideas and prototypes. I’m trying to push the projects so that alongside AI, learners explore physical materials, iterate on ideas and collaborate with others.
As part of the nationwide Digitaltag festival in Germany, I’m partnering with the City of Freiburg Sustainable Management Office and the Stadtbibliothek Freiburg to lead a new workshop that incorporates many of these design principles. We’ll use AI image generator called “SchulKI”, which is less powerful than cutting edge models but is built for educational settings, to visualize ideas for a more sustainable city. Then using a mini sticker printer, we’ll make physical copies and then add them to a shared cardboard installation.
As I’ve been preparing for the workshop, it’s been interesting to iterate on image generation and see how experimenting with prompts can make more interesting images. I wanted to share some of my process before the workshop and reflect on how that might inform my facilitation at the event.
Some of the iteration on prompts helped to make the images less generic or “vanilla man” as my friend Eric Rosenbaum labeled the results that come from generative AI. For example the prompt “robot helping a senior citizen in Freiburg install solar panels“ resulted in a standard while plastic robot that I didn’t like so I added some description. “Friendly and scrappy homemade robot helping a senior citizen in Freiburg install solar panels“ gave me something that I fount to be more compelling and less of the stereotypical robot look and feel.
Creating images of a playground made from upcycled materials is an example where the prompts got more specific with descriptions of things to add and things to take out. To make an image that I thought reflected both the style of the playground and the futuristic theme I ended up with “a couple of kids wearing cyper-punk clothes playing at a sustainable adventure playground in Germany with scrappy play structures made from upcycled desks, bathtubs and pallets no plastic or bright colors.“
An example also with experimenting with different styles in the images. For this idea of making a lab-grown version of the traditional Freiburg extra-long ‘lange rote’ hot dog I went through a few iterations. Adding in key words like “futuristic but grimy” gave me the look and feel I wanted. And at the end going to “scientists and researchers working in a futuristic but grimy setting to create lab grown hot dog in the style of a comic book“ gave me something a bit more interesting. While the first image seems like a current day advertisement for a run of the mill laboratory, the second has the look and feel of a jumping off point for a sci-fi lab grown meat story.
As part of the workshop, I’ve been thinking about how to best use these images to connect to physical projects in the real world. I really like this mini printer that makes stickers because they are quick and easily customizable. I would like to try to go deeper by having each sticker be part of a larger scene (like windows in an apartment building or specific elements that we could cut out and add to our collaborative installation.
Going through the process that the learners will go through at the workshop has been an interesting experience. I would say that creating the images and designing the prompts in an engaging activity. I kept wanting to tweak and reconfigure the words of the prompt to see what the AI would come up with. The results of the images are sometimes funny, often unexpected and for me the prompt rarely worked in the way that I had expected.
I’m still not sure if the process of creating these images is more like the addictive quality of building a marble machine or scrolling through the instagram feed. One one hand there is some creativity and experimentation to tweak and edit the prompts but on the other hand it still feels like the machine is doing much of the work and you are just waiting for the next image. For me one way to counter this factor is by moving as soon as possible to the real world constrution and to try and create an experience where learners are organically moving between the screen and the physical materials. I’m interested to continue to explore this tension in different formats.
I hope that you can join us this Friday at the digitaltag event hosted by Stadtbibliothek in Freiburg from 4-6pm to experiment with these ideas, make your own images and add them to our collaborative installation! The workshop will be free to attend and open to adults and teens over 14 years old. I’ll post again about the results of the event and the learnings from trying this experiment with participants.