The missing human quality of imagination

The influence of artificial intelligence (AI) is also steadily increasing in art. Here is an interview with art educator Prof Dr Jochen Krautz from the University of Wuppertal

Mr Krautz, can a machine be an artist?

Krautz: No, a cordless screwdriver, a camera or an image editing programme are not artists either. But you can make art with them. However, I believe that AI-based image generators, because that's what we're talking about here, offer the opportunity to sharpen our understanding of art in a new way. After all, even after the invention of photography in the middle of the 18th century, there were rumours that painting was dead from then on. But it still isn't today.
Basically, we must be careful not to extend our tendency to view reality anthropomorphically, as Robert Spaemann put it, to machines. In other words, microelectronic systems are not people. Therefore, they cannot make art.
Even if AI-generated images appear impressive and creative at first glance, they are based on probability calculation, pattern recognition and combinatorics, which has little in common with human artistic creation.
The terminological blurring also contributes to this deception - or rather: our careless self-deception: AI is not "intelligent". It does not think, it does not decide, it does not feel and it is not physically situated, which is the specificity of human intelligence. One colleague put it very aptly that AI has as much to do with intelligence as the logistics industry has to do with logic. In other words, at best formal procedures that are used functionally - and can also be misused.

How does AI's image-making differ from that of humans?

Krautz: In art didactics, we ask every artistic process the question of what exactly we are doing when we paint, draw, sculpt, build an installation or shoot a video. This is central to being able to teach the processes with the appropriate aids. In doing so, we always ask what the respective relationship is between vivid perception, pictorial imagination and visual representation.
If you ask yourself these questions for the AI's image generation, it becomes quite interesting: it has no perception itself, but is fed with image data that we have created. Above all, it does not imagine anything, it does not have the central human characteristic of imagination. Image representation is a purely technical output process on screens or by means of printers. AI therefore amalgamates the perceptions and ideas of millions of people of all times into technical representations. These work according to the collage principle. However, it is unclear how an AI makes creative decisions, or more precisely: whether one can speak of creative decisions at all. This is because it "knows" nothing about a motif, image content and their meaning. It calculates figurations according to principles of probability. An average horse is generated from tens of thousands of objects with the tag "horse" - without knowing what a horse is and what significance it has in terms of motif, culture or art history. AI is a calculating machine and calculates according to specifications, just like a blender mixes.
In a way, it reflects a kind of global imaginary space represented in images. However, because it only operates with what is available, it necessarily operates very schematically. AI therefore initially produces stereotypes, which can be problematic from an ideological point of view and usually results in kitsch in art.

Artists are already using AI to support their work. To what extent can AI become a legitimate partner for an artist?

Krautz: Be careful here too! Talking about "partners" personalises a technology in a very problematic way: Is my bike or my car my partner when driving? Are Alexa or Siri my partners? Certainly, many people now feel that way. But what are the actual consequences?

In art, AI can be an instrument just like brush and paint, chisel and chisel, camera and image editing programme. However, art is not necessarily created when I have bought all this from an artist's supply store (even if their advertising likes to suggest this). Just as little if I use AI (even if the equally financially motivated tech companies like to suggest this).

However, AI has a logic of its own that we no longer understand, which elevates it beyond its status as an instrument. A paintbrush is largely controllable with experience; an AI is not. Because it increasingly makes decisions that we can no longer comprehend, it appears "creative" to us. In fact, the user formulates a prompt and the programme generates the most likely intersection from the combination of algorithms (as instructions for the computer) and a (huge in the case of AIs) database.

There is a "setback" for the machine: those who prompt not only issue commands, but also receive them because they have to adapt their logic to the machine programme. Sure, I also have to adapt to the logic of the brush I'm painting with or the piece of wood I'm working on. But here I adapt to the logic programmed by others, even if I try to work against it. In other words, I limit the scope for artistic freedom from the outset.
This is why AI also changes our perceptions and ideas. This is not unproblematic, because the machine actively intervenes in real space and our human interactions. It shapes our perceptions and attitudes. So if you want to use AI as a "partner" for your own art, you have to take a very close look at who is actually using whom.

Can AI create art on its own?

Krautz: Basically no. But the phenomenon helps to clarify what art actually is: if we consider the imitation or mixing of art styles and global image patterns to be art, there is something wrong with our understanding of art.
Visual art results from our relationship to the world: we somehow try to wrest a meaning from this life, even and especially when we think it is meaningless. Art therefore creates a horizon of meaning, it results from an existential relationship to the world. As artists, we therefore place ourselves in relation to the world, to others who view our art and, through both, to ourselves.
Since AI is not alive, it has no existential questions. Therefore, it cannot create art. If it creates images or objects that we consider to be art, we have to ask ourselves why we do so. Either our judgement of quality is wrong or we are interpreting our human questions into the AI products.
To put it another way, using artists' slang, so to speak: they like to say they are "working on a problem" or "I am currently interested in ...". However, an AI cannot itself have an artistic problem and is not interested in anything. We may be able to use it to work on our own problem, but understood in this way, the AI remains an instrument and is not itself an artist.

Some artists are worried that they will abolish themselves through AI. So if an algorithm can be just as creative as a human being, this concern is justified, isn't it?

Krautz: If an algorithm were creative, then that would be the case, yes - but that's not the case. Those who programme it are creative.
But the first part of the question points to the real problem: many artists have long been adapting to technical images in a kind of "anticipatory obedience", as our colleague Gesa Foken analyses it. And not only aesthetically, but also in a mimicry of the technical mode of production: They imitate quasi-machine work processes even when they work in analogue. In this way, artists are in fact abolishing themselves by subjecting themselves to machine logic.

For art, the AI phenomenon should lead to a reconsideration of the question of the human core of art, to a re-emphasis on its relationship to the world, its materiality and responsibility.

These systems learn with vast amounts of data and process images and texts from the Internet whose creators have not consented to this use, but can also do nothing about it. Are artists being robbed of their identity?

Krautz: Adapting the works of others is fundamentally part of artistic work. The principle of "imitatio auctorum" (imitating the works of others) was and is a central artistic learning and working principle.
If we understand "mimesis" as creative imitation of a role model, as Christoph Wulf puts it, then it becomes clear once again that AI is not a person who is capable of this. It processes vast amounts of data and combines it with other text or image data using arithmetic operations that are almost impossible to comprehend. In this respect, yes: in my opinion, the aforementioned danger does indeed exist, because AI does not give any account of this and the operating companies ultimately utilise intellectual property on a massive scale without being asked. It is already the case that AI operators are in fact using the works of others - or to put it more sharply: stealing them - in order to generate profit themselves, but this threatens to destroy the livelihood of the applied creative industry. This should also be analysed more closely from a legal perspective.

Even if artists are allowed to object in future or are asked for permission, the current generation of generative AI has learnt from their work. Some of the systems are open source and therefore freely available. Anyone can use them to imitate styles. If these tools become standard sooner or later, will the creative industry have to fundamentally change its processes and skills?

Krautz: Well, as I said, "imitating styles" is not an art. As far as the creative industry in its applied fields is concerned, the question is indeed different. From what we hear in design, for example, AI has long been used in design, testing and quality assurance processes. In the latter, this saves analogue or human work processes. In design processes, AI is used to create initial design series in order to gain an overview of the breadth of what is known and obvious. And you may continue to work with them, but they do not replace the decision.
Things are getting really tight for catalogue photographers, illustrators, layout artists, campaign designers, web designers, advertising film makers and others, because their standard work has long been done by AI: You no longer need models, a studio to photograph a fashion catalogue or products, a film set for advertising videos, etc. There are already predefined prompts for everything from layouts and colour sets to entire advertising campaigns.

This seems to me to be central to art and design: If AI generates images and designs quasi automatically, the central task of the designers shifts to their judgement. In other words, the artistic and creative judgement of quality becomes increasingly central. This is because AI cannot judge design quality. Such judgements depend on intentions, form, content and contexts.
Artistic and creative judgement, in turn, can only be formed through one's own practice, especially analogue practice, and reflected reception experience, which is established knowledge at least in art education. The fundamental artistic and creative competences are therefore not changing; in fact, they are becoming increasingly important if we do not want to be at the mercy of AI. In this respect, the focus is shifting to the formation of artistic and creative judgement.

Everything is changing, including art. Image-based AI is already being used in art lessons at school. Are there any ways of controlling the use of AI in the arts?

Krautz: We have to differentiate here: The art sector is basically free in what it does. We can only hope that those working in this field deal with the issue responsibly and are aware of the far-reaching implications and dangers. However, the issues of copyright protection that have been raised urgently need to be addressed at a political level.
Art teaching has a normative horizon, i.e. it is part of school education which, according to the constitutions, is still aimed at maturity and humanity. In this respect, AI cannot be part of teaching without very careful pedagogical and didactic consideration. When it is used, its problems, possibilities and limitations must always be considered. Otherwise we lose the humane reason for art.

And it's not just art lessons that need to ask themselves whether we don't need school time to acquire the basic skills that we need to use digital technology in a reasonably responsible way, but which we cannot acquire from and with the technologies themselves. Perhaps we need to think in a much more complementary way: what should be taught in school today that children and young people are no longer learning in an almost completely digital world?

For art lessons, this means: isn't analogue drawing and building the prerequisite for dealing with digital technologies in a meaningful and creative way? And aren't the sensory and physical experiences that can be gained so essential that we should make them possible for young people, precisely because they are unlikely to have many of them after school? Don't we need to place the perceptible world, the formation of one's own imagination and the possibilities of shaping it in a formative and tangible way all the more at the centre of school the further it disappears from the horizon of children and young people? Is it possible that school needs to become more analogue and physical in general as the world becomes more digital and disembodied? Don't we need more "analogue skills in the digital age", as Gernot Böhme aptly put it?

These are outmoded thoughts, I know, because they run counter to the seemingly inevitable technological development and its economic exploitation interests. But: thirty years ago, the artist and writer John Berger once formulated, in view of the increase in disembodied images, that painting today is increasingly becoming a kind of resistance against the regime of virtuality. Perhaps it is therefore particularly timely to do the seemingly outmoded. In art, in art lessons and beyond.

Uwe Blass

Prof Dr Jochen Krautz studied art, Latin and educational science in Wuppertal and Cologne. He has been a professor of art education in the School of Art and Design at the University of Wuppertal since 2013.