Can AI create art?
Or, can AI be creative? In really every article and in every discussion about art and AI, these questions are asked again and again. The answers offered are very different. You often notice that the people discussing it have probably not dealt intensively with the practical application of image-generating neural networks. However, the clear answer to the question must be no (at least so far and when the question is posed in this way). But more on that in a moment.
Is AI really intelligent?
The term “artificial intelligence” has – unfortunately – already become established. The definition of intelligence in humans is complex and still controversial. Nevertheless, let’s just call neural networks and their algorithms “artificial intelligence”? The term seems misleading and counterproductive to me. It creates fear and defensiveness in many people. Even the word “artificial” has a negative connotation, because for many people the opposite, “natural”, has a positive meaning. And to describe neural networks as an intelligence is premature at best. So we understand “artificial intelligence” as an artificial personality with a consciousness. Because often people are already saying: the AI thinks this is so and so, the AI doesn’t know how many 7 or 12 cats are. Which implies that it “knows” other things instead.
You should always bear in mind that we are only talking about neuronal networks with an incredible number of neuronal nodes (neurons) and that amazingly simple computing operations take place in the neurons. However, we tend to anthropomorphize this technology due to the surprising results that are delivered with the help of these neural networks. AI is often referred to as an assistant – simply because it feels like one.
Back to the question of whether AI can do art. It can’t – without humans. The neural networks are completely passive as long as they do not receive any external input. In the case of image-generating neural networks, a human is therefore always necessary. The same applies to the question of creativity. It is always possible to achieve creative results with the help of “AI”. But the emphasis here is on “with help”.
What is also often discussed in the debate about AI is the comparison between AI and artists. This comparison makes just as much sense as comparing chainsaws and lumberjacks. The chainsaw does not go into the forest alone and decide which tree to cut down, how it should fall and then get to work. The lumberjack, on the other hand, can fell a tree much more easily with a chainsaw than with an axe. This is because axes and chainsaws are just tools and need a lumberjack to use them. It is the same with neural networks. They can be tools for the artist. Neural networks do not decide what kind of image and how they generate an image. Nor can they decide when the work is finished or whether something needs to be changed. Nothing happens here without the human being.
The image does not emerge from the AI itself.
Artists have an idea, a vision of a picture or a mood that they want to visualize with the help of AI. They choose the motif and describe it as well as possible. They can choose colors and techniques. And they can even bring in their own personal style and use their own sketches or images to determine the composition, style and mood. Artists select their results and modify them further using AI or other computer programs. They transfer the image to a medium of their choice and can rework it with color and other materials. The creative process does not have to end there. The results can then also become part of a concept idea. This is the practical side of artists using AI.
The other side of AI
Of course, there are also people who are not artists and use AI to produce “nerdy kitsch”, as the editor Dr. Wolfgang Stieler from Heise.de once aptly described it in an article. And it is unfortunately the vast majority of AI users who prefer to have AI generate half-naked women, cute kittens, funny cartoon characters, interiors, hearts, sunsets, sports cars, martial warriors, Halloween pumpkins, skulls and so on. The list of non-creative motifs could be continued indefinitely. However, the countless images that are generated in the style of … form a special category. This can be interesting or funny – but it is certainly not artistic. However, this also confirms that the neural networks do not produce any creative results without the creatives.
I confess here that I generated my first three pictures in March 2023 in the style of Albrecht Dürer. However, mentioning artists’ names in a prompt is perhaps not such a good idea. At least not if the artists are still alive or their works are not yet in the public domain. This is because most of the image-generating neural networks were trained with images by artists – without their consent. The issue here is the rights of use, which always remain with the authors as long as the artists do not sell or assign them. It is therefore not surprising that artists have already filed lawsuits against the operators of the major networks.
However, this raises another question. What actually constitutes use? If I go to a gallery and see pictures by a contemporary artist, for example, is that a use? Because I take the pictures and the impression in my mind and am inspired. Would I have to pay the artist something in this case? That is of course an absurdly provocative question. Of course not. But the artists who are suing are also complaining that their style is being copied, which took them many years to develop.
German copyright law is very clear on this, by the way. An artist’s style, like an idea, is not protected. Copyright law only protects the specific execution or design of a work, but not general artistic techniques. It is important to understand that this regulation promotes artistic freedom and innovation. If styles were protected, this could significantly restrict the development of new forms of artistic expression. Artists can therefore be inspired by the styles of others and develop them further without infringing copyright, as long as they do not copy specific works.
It should also be mentioned that the AI cannot be the author by law. Only a natural person can be an author. For example, an artist.
A suggestion
Personally, however, I think it would be good if the naming of artists whose works are not in the public domain could be prevented in prompts from a technical point of view. Certain terms that are not permitted according to the terms of use of the AI operators are already successfully blocked. This should also be easily possible with artist names. Then there would be no more images that can be immediately assigned to a specific artist. And there would no longer be this flood of plagiarism. Apart, of course, from “Klimts”, “Van Goghs”, “Schieles” and “Dürers”, whose works are in the public domain.
And what actually is art?
Neural networks are incredibly powerful tools. For me, they mean empowerment and the democratization of the artistic creative process. It is not true that this creative process must necessarily involve a paintbrush or a printing press. Because the brush and the paint alone do not make art. Neither does the use of neural networks. It is and remains the human being who creates art.
“The great achievement of the 20th century is that other categories have been found for important art than anatonmic correctness or classical craftsmanship. It’s always about form; what form of expression do you find for your message and for what you want to say.” (Florian Illies, May 3, 2023)
What is art? A question that is as old as art itself. Futurologist Bernd Flessner has a clear opinion here: “If a work of art says something to the recipients who look at a picture, listen to a piece of music or read a book, then it is art, regardless of how it was created”. That’s how I see it too. If I look at a picture and it triggers a strong feeling in me, then for me it is art. It’s actually quite simple.
I can recommend a book to anyone who wants to delve a little deeper into this fundamental and age-old question. The book “Playing to the Gallery” by Grayson Perry deals with this question in a very entertaining way and is by no means just gray theory.