For my project, I asked ChatGPT to imagine a world where indigenous cultures were never colonized and then I asked it to write lyrics for a song about indigenous people mourning the future they would never have because of colonization. Then I went to splash and played around with different sounds before putting the lyrics into the AI voice. One thing the song really struggled with was how certain words were pronounced and how to put spacing that seemed more natural since I had no control over how the AI would present certain words. For example, every time it would pronounce “resilience” it wouldn’t be phonetically correct, so eventually I just changed the lyrics manually. I also asked midjourney to create certain prompts that would try to imagine ancient civilizations and what that would look like through a futurist lens but it really struggled with that because if it is only fed information of a colonized world it is really hard for it to think of what possibilities it could look like without colonization since it is so embedded in our world today. It also struggled to depict visual cultures accurately, for example the Ainu indigenous peoples of Japan would tattoo women’s mouths as part of their culture and had a very distinct textile pattern. But no matter how specific my prompts would get (ie: impressionist painting of Ainu woman with black mouth tattoos wearing traditional AINU indigenous robes and headgear while holding a flower and looking to the side) they refused to show any mouth tattoos. I think this shows that AI hasn’t really been trained on specific cultures and that ultimately a visual history of what has been recorded (and how) from the past plays a big part in modern AI.
Project Video Link
By

Julia Galang
CAMD AI Projects