In art, humans are depicting using a varied range of perspectives. Yet in each instance, there seems to be a "common sense" that enables the user to say "This is human". Though, what is this essence? Through giving text to image generative AI tools: DALL-E 2, Midjourney and Stable Diffusion with simple prompts, the below image was created: - "a human, depicted in a minimalist style" - to sample artistic perspectives - "a Mongolian person" - to sample diverse but photorealistic images ![[Pasted image 20230910205331.png]] Keeping the prompts simple, let's the AI do the "imagining" - revealing it's capabilities, limitations and biases. It's important to show these AI tools and their flawed outputs as they are, so that we can explore rather than avoid important questions the below: - How can we prevent them from perpetuating harmful biases and stereotypes? - How can we deal with the imbalance of cultural knowledge in these tools? ---- *Credit: Commissioned by Science Gallery London Created by: Munkhtulga Battogtokh, Department of Informatics, King’s College London Collaborating artist: Alice White* [[Foundational Models MOC]] [[Artificial Intelligence]]