I keep reading about the Ai monster that is coming to take away everyone’s jobs, but my on-going exposure to ChatGPT and Ai image generators is leading me to believe that those stories might just be misinformed fear-mongering. The reality is that these seem to be immensely powerful creative tools, with no appetite for my job…just yet.
For this test, I decided to use DALL-E and StableDiffusion 2.1. I have had some success in the past creating interesting images with Mid-journey, but this was meant to test new ground for me. here is no touch-up or post production here and no in-painting or editing of any kind. The exercise here was to just test the generative abilities of each, based on provided prompts.
I have loved Mobile Suit Gundam since as far back as I can remember and have owned (and subsequently destroyed) more than a few models over the years. Naturally, the kid (and the Mexican) in me wanted to see what would happen if a Gundam were designed to look like an Aztec warrior so I went to work. The first takeaway: language referring to either Aztec, Incan, or Mayan had no implications on the design.
The results:
DALL-E
StableDiffusion 2.1
In the end, I have no explanation for why DALL-E produced such illustrative images when both generators were instructed to produce “realistic” images. For that matter, regardless of the language used in the prompt, I could not get StableDiffusion to render anything other than this very anime look. I also had to change the background to “downtown Tokyo” because the landscapes were unrecognizable and quite unappealing to look at. Still, both are fun and entirely too good at what they do. I look forward to working image generation into my workflows, but the jury is still out on which platform I like best.
The biggest takeaway from all this was not a feeling of overwhelming fear for my livelihood, but more sheer curiosity of just who is the captain of this pirate ship. Ai feels like what I would imagine the wild west felt like, in that there really are no rules yet. I couldn’t help but think of artists on both sides of this equation, though – the inspired and the plagiarized. It was exciting and terrifying all at the same time to create in this manner but as copyright laws are formalized I wonder where this technology will come to rest, if at all.