One of the fun things about making is AI art, is seeing whether an algorithm recognizes an artist. Dall-E 2 (which I am eagerly waiting for an invite) seems to recognizes a great variety of artists. For now I am using the wombo app, and from what I can gather, it only recognizes the very famous and popular ones. So we start with a very popular artist, Georgia O’Keeffe. She has a distinct style and a frequently used subject matter (flowers!), so she is a good starting point to see what the app can do. Some of these pictures were generated from scratch, and some used a seed picture.
The “Iris” pictures did not use a starting picture, and they seemed the most O’Keefe like – especially the one in the bottom center. One of the funny things is that the app sometimes things “iris” is a part of the eye, instead of a flower, and you can almost see that here. These are the artifacts that make AI Art interesting now, and I believe that when they are eventually resolved, we will miss them.
The bottom center picture was made with the “wuhtercuhler” art style, so I generated some more with that filter:
Very Georgia O’Keeffe like! As you might imagine, this is a dangerous thing for artists, as it could make them obsolete. When the machine learns to recognize your style, what is stopping anyone from ripping off your work? Also, these algorithms are trained on actual photos and images, and the originators of the content will likely never be compensated for this. Apps like Wombo and Dall-E 2 stand to make a great deal of money from these tools – the original artists – not so much. In the near future, there will be many things written on this controversy. For now, I am only interest in seeing what these tools can do.