Look at the image. It is painted in a surrealistic style and depicts a landscape that could be taken from a dream. When you look at the painting, you might wonder what the different parts of the painting mean. The face in the moon, the man on the horse-bike, the train crossing the bridge. What are they symbols of? What does the creator of the painting want to tell me? Is it a dream that the artist has painted?

This painting was created by an AI image tool (DALL-E). I only wrote a prompt to describe the motif. You might wonder what I wanted to say with this painting? What was my purpose with the image? What exactly did I ask the AI to create?

My prompt was very simple: "a surrealistic painting of a landscape with people." The prompt contains no details about trains, horse-bikes, or other elements. Nor that the motif should express a certain feeling, idea, or symbolism. I did not specify that it should be composed in a particular way or asked the AI to use certain colors or even be inspired by a specific era or artist. My idea was simply to create a painting that looks like it was painted by a surrealist.

But what did the AI want to say with the motif it produced? The image is well-composed. The colors match, and there are many exciting and dreamlike elements that we recognize from other surrealistic paintings, such as eyes, vehicles, strange shapes, and creatures. But unlike a surrealistic painting by a human artist, there is no thought or idea behind the motif. An AI is not conscious; it creates an image that is a statistical model of, in this case, a surrealistic painting. So, what you see in the image is an average surrealistic painting, not a unique painting.

One way to describe the new art style that generative AI creates could be to call it probabilistic art. An art style where the AI uses large amounts of data to generate images that represent a probable or average, that is, probabilistic version of a motif. It is an art style that provides insights into how different elements and details tend to look based on statistical patterns and trends in the training material used.

So, when you write a prompt, you do not get a unique image but a collage of average images of how all previous artists and creators have interpreted the different parts of your prompt. Just like a collage, the whole can be perceived as unique and innovative and creative, but just like a collage, the AI image is only composed of probabilistic layers of motifs. In that perspective AI art is a natural development of the cut-and-paste and remix culture that has developed on the internet over the past decade. Generative AI can be seen as the ultimate way to remix, renew, and reuse our cultural history.

Throughout history, we have seen several times how artists have created new art styles by reusing and renewing older visual material. During the Renaissance, ancient art and culture were rediscovered, in the mid-1700s, antiquity returned through neoclassicism, and during postmodernism, mainly in architecture, older art styles were taken up and reused. AI art is the last step in this development. Probabilistic art uses mathematical algorithms to create the most representative and probable image of different motifs. This is also why, many images looks generic and similar to each other. Many users use similar prompts and create images in a few popular styles, resulting in similar motifs with few variations.