A blog related to original AI artworks. I make AI artworks using coined words and emoji as prompts.

Is AI good at abstract paintings?!

There are many tools that use deep learning (AI) to generate images from English sentences. I’m trying to make various abstract painting-like images using it. Looking at those images, I feel that AI (deep learning) is good at abstract paintings.

First, let’s look at some examples. As I have already introduced in this blog, when I draw a cat in Mondrian’s style, Stable Diffusion draws a picture that captures the characteristics of the cat well.

The reason why I can draw such pictures may be because I have learned various abstract cat pictures, but I think that is not the only reason. Deep learning has given computers the ability to abstract (generalize). I believe that this ability is being demonstrated. Not all humans have great abstraction ability, so isn’t it possible that AI has more ability than the average human?

Let’s take a few examples where it doesn’t seem like you’re drawing a picture that you’ve learned without abstracting it.

In the next picture, a white cat is embedded in a black cat. How did AI learn this technique?

Note: This article is a revised and translated version of an article in Brog from Kanada (in Japanese).