Instant Variety
As everything evolves towards more complexity, generative AI tools give us a way to accelerate this evolution. The hype about these tools capabilities is on point when they’re viewed through the lens of variation generation. Generative AI can quickly mash up many different items together.
I’ve looked at this through the generation of infinite detail using Midjourney by creating Wimmelbilder-style drawings to looking at how AI can generate faces across “latent spaces” — essentially, different grouped characteristics.
If one looks at simpler styles, it’s even easier to create fantastic variety. The emoji above illustrate this clearly. In fact, generative AI does a much better job at this than trying to create fonts. With fonts, Midjourney falls flat. It hallucinates letters that don’t exist, at least to my knowledge, and will repeat others. If you ask for a single letter, you can find success but the alphabet is a challenge.
As we see more integration between large language models and image generation tools, we’re going to get much better performance in generating images that are inline with what we’re imagining. It’s also going to help hone our preferences by presenting many options to start our selection process.
That’s going to be exciting.
