At the beginning there is the text: "It should rain red roses for me", translated into English. There follows a word that will not be revealed here. To run. The parameters are set, the neural network is working. The result is initially - nothing. The first picture is almost monochrome. Then the second run follows, the third, fourth, gradually red spots of color form. One could think of the early universe, where the beginnings of the galaxies presumably form soon after the Big Bang. After the fiftieth time you can already see the flowers. And finally, a hundred runs later, the picture shows four roses that appear to be dripping and slightly squashed to stick to a surface.
You no longer have to be a programmer to use artificial intelligence (AI) to generate images.
At the beginning of this year there was a real boom in AI art on the internet.
The trigger was the publication of CLIP by the company OpenAI, one of the giants in the development of artificial intelligence alongside Google.
CLIP is a neural network; it assigns a suitable description to an image.
It wasn't long before programmers and artists figured out what to do with it.
You can connect a second machine to it, an image generator.
Then you just have to enter a little text, a so-called prompt, a kind of command, and the program generates images that are matched to what the AI thinks is appropriate to the input.
A song title becomes a still life with roses.
In the world of the neurographer
AI art has developed rapidly.
The artist Mario Klingemann was already doing generative art with computers long before Google published the psychedelic images of “Deep Dream” in 2015.
If you ask Klingemann, in his early 50s, about the rapid technical progress in his work, he laughs briefly and says: “I'm already used to that.” For example, he spent months working on having faces painted realistically.
When he finally got it out, one of the big companies with their software completely passed it by.
“But that's part of the whole thing,” he says.
“You have to see where you can find your niche, where you can still say something new.” In any case, he is glad that he has already made a name for himself.
Now it is easy to get lost in the "cacophony".
At the beginning of the zoom conversation, the artist apologizes for the rattling in the background. In the room next door to his Munich studio, a robotic arm works and scans slides. Automation, as you can see right away, is a far-reaching idea. Klingemann describes himself as a "neurographer". Based on the photographer who depicts parts of the world out there, a neurographer is on the move in the world of neural networks and takes pictures in this abstract space.
The neural networks have been a breakthrough in the development of AI.
Instead of specifying rigid decision-making patterns, they allow the machine to find the criteria it needs in order to understand a text or to recognize objects in an image.
It does that through training.
But because so little is given, it is difficult to understand how exactly the network ultimately distinguishes a dog from a cat.
The unknown, the surprising, the mistakes, all of this makes neural networks interesting for artists.