In 1993, I wrote a short article for Wired magazine titled “Fishing.” The story was about how a computer scientist named Clifford Pickover wrote software to generate art. Making generative computer art, Pickover said, was like catching fish:
"Computer programs and ideas are the hooks, rods, and reels. Computer pictures are the trophies and delicious meals. A fisherman does not always know what the waters will yield; however, a fisherman may know where the fishing is good, where the waters are fertile, and what type of bait to use. Often the specific catch is a surprise, and this is the enjoyment of the sport."
For the past couple of weeks, I’ve been fishing for art myself. My interest in generative art was sparked by seeing the incredible work of Unlimited Dream Co. Here’s an example:
The Unlimited Dream Co. (consisting of a human and several AIs) also posted a series of excellent articles on how to create AI art. Using the instructions, I set up a Google Colab Notebook (which runs the AI art software) and started fishing for art.
The software asked me to provide a text description of what I wanted the AI to generate. Additionally, I could upload an image for the AI to use as a starting point. Here’s what the software created after about 20 minutes using a simple sketch and the prompt “colorful plastic space deity.” It was amazing!
Since then, I’ve been experimenting with a couple of different AI art generators. My favorite so far is a website called Conjure. Like the Google Colab Notebook, Conjure asks you for a text prompt and/or a starting image. Unlike Google Colab, Conjure is easier to use and much faster. Another cool thing about Conjure is that you can select from several different artistic styles. For example, here’s what Conjure generated from the description “DNA paranoia cavern” in the styles of “retrofuturism,” “utopian,” “cyberpunk,” and “solarpunk.” I liked them all!