With the recent hot news of the Stable Diffusion (AI) algorithm for image generation based on a short user prompt, Marina Velikova, the founder of Creazzia, came across a powerful demo provider allowing on a small scale to try out the algorithm.

As a trained AI scientist, she was truly impressed by the efficiency and sophistication of the AI image output. Having played around with some fun prompts, she went to check on her dream of building a new inspiring learning space that provides lots of light, connection with nature and a cozy feeling of home.

Using the prompt “creative study/learning place students” yielded standard results as the rooms looked very similar to what we already know at schools and workplaces. She decided to add the concrete term “glass roof” to look for a light-giving place. That helped in getting a diversity in designs but still the “study place” was associated with tables and chairs, ordered in the rooms.

On the next "design" step, Marina was thinking of a cozy place for studying and she added the words "oval" and "sit on pillows" to the prompt. That helped to get into the creative realm where unexpected and very different results emerged thanks to Al in no time.

What are the first preliminary learnings from these play experiments?

AI (Stable Diffusion Demo) seem to associate “study place” with the familiar images of learning places as we know them in daily life, likely as it is fed (trained) mostly with that type of images, available widely. The word “creative” seems too broad/abstract for an AI algorithm to deal with.

Once prompted with concrete words as oval and pillows, the pictures appeared very different and closer to the originality looked for by the user. That shows that AI is still in the concrete ‘child’ phase of learning, though catching quickly and with impressive results.

The question that arises now is “Would the jobs of professional artists, photographers, architects be threatened by AI or rather strengthen by co-creation with AI?”

The future will show.

In the meantime, try the algorithm for yourself and share your experiences with us: https://huggingface.co/spaces/stabilityai/stable-diffusion