Concept
During our conversation in class last Wednesday, I felt a tension about AI and making art with it. The tension stems from the fact that AI image and video generation use an inordinate amount of electricity. The title of the video, 2.907 kWh per 1,000 inferences, is a reference to a study done by Carnegie Mellon University, which cites this figure as the average electricity used for generating 1,000 images with LLMs. “As the paper notes, the average smartphone uses 0.012 kWh to charge — so generating one image using AI can use almost as much energy as charging your smartphone.” (The Verge)
Another conversation inspired my framing of the video around a single user. My friend noted a recent encounter observing someone else using an AI chatbot on a flight. My friend observed the man using the AI as an advanced Google search for anti-balding products, best place to buy posters, and how to place proper poster dimensions in a bedroom. My friend asked if we really needed to use that much energy for purposes as frivolous as that. This question informed my exploration of using AI images.
This video follows the journey of a single AI image produced from one man’s request through data centers and cyberspace. Who knows what other data might pop up along the way?
Process
As my English major brain desires, I started by writing a script of my idea.
I went shot by shot, starting with the real life footage I recorded. I used my dorm bedroom and iPhone, which turned out well.
For the AI images, I wanted to produce as little AI images of my own as possible. I can’t make a movie about the environmental cost of AI and use it a lot. Also, there are a lot of AI generated images on the internet, so I felt I did not need to add more for this project.
However, I do not blame the costs of AI on the users as much as I do on the corporations who develop and control our access to AI.
To achieve reducing my AI footprint, I searched Reddit and Adobe Stock again for AI Generated images. Thank you to NYU for free Adobe Stock image licensing. Below is a screenshot where I filtered images for AI generated only on Adobe Stock.
For the videos, I searched YouTube. I may or may not have used cobalt.tools to download videos, who is to say.
I overlaid the videos in Premiere. I put the Aphex Twin song in there because it felt right for that journey.
Success?/Always Room for More
I felt I captured the spirit of my concept well. However, I wanted to include the Internet as more of a character in this journey. I wanted to showcase how AI acts as a harvester from all different sites and uses this wide range of good and bad data to spit out output at us. I also wanted to add more humor with memes and odd juxtapositions because the comedian in me wanted to lighten the mood where I could.
However, I did not have enough time to include more things. Also, I did not want memes to distract nor pull attention from what was important to witness in the videos.
Difference in Approach?
“How would I approach the project differently knowing what you know now?”
I would animate it by hand, if I could.
But I can’t. So here we are.
Ryan Rotella 
Leave a comment