Should we speak differently about Artistic and Artificial Intelligence?

_____

A thought experiment with our vocabulary about Art and AI

_____

Wondering about a conceptual vacuum

The revolution in generative Artificial Intelligence (AI) has impacted the visual arts in a disruptive manner: This might be what the iphone moment was for music. AIs ability to generate complex, aesthetically sophisticated images in seconds has radically altered the means of artistic production, especially when on top author rights are infringed and creators are not accepted and paid as the owners and rights holders they are.

Artificial Intelligence is too often built on the exploitation and even destruction of artistic intelligence.

Whilst artistic practice is coerced into this change (and others will call this a quantum leap forward), our language—and thus our thinking—is not undergoing a similar change. As a result it might not be fit for purpose any longer, not fit to describe what is happening. We are experiencing a new form of intelligence, leading to new forms of freedom and creativity, but are trying to continue with grown conceptual drawers. While slack in language development without doubt serves good reasons too, the disruption of AI is here to stay. We from the cultural world should not let technology also enforce a new language and terms upon us. However, if cultural players do not step forward with proposals for a new language for a new artistic practice, even if from a machine, we can not complain of being taken over!

Let us test the thesis: We are facing a conceptual vacuum. When an artist speaks with an AI model for hours, prompting it, challenging it, curating the results, recombining, and discarding them—what exactly is the artist doing? And what is the AI in that moment?

The metaphors we instinctively reach for fall short. If we call the AI a tool, we ignore its ability to surprise us with unexpected, emergent ideas. A hammer does not suggest an alternative position for a nail to its user. If we call it an assistant or co-pilot, we falsely humanise it, imputing an intention and consciousness it does not possess.

This article asks if these old terms are not merely inaccurate, but rather prevent a genuine understanding of what is happening. The core problem is a philosophical category error: we are attempting to grasp a fundamentally post-human phenomenon—the intimate co-creativity between a human and a non-human agent—using the vocabulary of humanism that centres around the ‘human genius’ cult. (This often serves as justification for a macho leadership style in culture, but this is for another article). This humanist toolkit is partially insufficient for a practice that dissolves the boundaries between author and instrument (as analysed by thinkers from Donna Haraway in her Cyborg Manifesto to N. Katherine Hayles or Artificial Extelligence by Siegried Zielinski). These traditional concepts might block the new thinking required for this new practice. It is time for a conceptual development that relates artistic and artificial intelligence in an innovative way to one another, overcoming vice versa (justified) accusations, yet blocking new forms of freedom and innovation, likely for Artistic and Artificial Intelligence alike!

Calling for this debate and thought experiment is taking us, and especially artists and right holders, out of their comfort zone. In fact it is a lot to ask as Artificial Intelligence is mostly trained and build on copyright protected works either without or without fair compensation or even permit of the rights holders. Artificial Intelligence is too often built on the exploitation and even destruction of artistic intelligence, at least this is what the majority of rights holders say. In many cases this leads to lawsuits, such as Getty Images vs. Stability AI or Authors Guild vs. OpenAI.

I wonder if there is a relation of the conceptual vacuum and the copyright vacuum; in any case my intention is to support fair compensation for rights holders, for obvious legal reasons, but also to make Artificial Intelligence sustainable in the long run, maybe even as long as Artistic Intelligence exists already?!

The inadequacy of old metaphors (the demarcation)

To develop a path for a new vocabulary, we must first analyse why the most common terms for AI in art are potentially misleading. Many stem from a mode of thinking that presupposes a clear separation between the human subject (the artist) and the passive object (the material or tool). However it is precisely this clear separation that is being dissolved by AI – at least from the point of AI inventors and evangelists. Let us take this for now as the starting point for the following thought experiment:

The ‘tool‘

The most frequent and seemingly harmless metaphor is that of AI as a tool—a paintbrush, a chisel, a camera. This term is seductive because it leaves full control and authorship with the human. Yet, it is fundamentally inadequate.

The language we use directly shapes the reality in which we operate.

A tool is passive. It extends the human hand but possesses no agency of its own. A paintbrush does not react to the semi-finished painting; a hammer makes no suggestions. Generative AI, however, is an active dialogue partner. It operates in a multidimensional “latent space” of learned concepts and reacts to the human impulse (the prompt) with emergent, often unpredictable behaviour. The AI surprises the artist. To call AI merely a tool is to ignore the dialogical core of the process, reducing a complex interaction to a simple command.

The ‘assistant’ and the ‘dialogue‘

This second group of metaphors is more complex. It begins with commercial terms like assistant or co-pilot and extends into the sophisticated academic discourse on dialogue. The popular terms (assistant) are the easiest to debunk as misleading. They humanise the AI in a way that obscures its true nature. An assistant implies understanding, intention, and a form of consciousness. An AI, however, understands nothing in the human sense; it is a sophisticated statistical prediction model completing patterns.

Far more nuanced is the dialogical approach. Thinkers like Grégory Chatonsky (with his concept of co-création) or Sergi Jordà (as a pioneer of tactile, musical human-machine interaction) have explored this dialogical practice positively and profoundly, seeing it as a genuine partnership.

However—and herein lies the philosophical difficulty—even the term dialogue suffers from a humanist bias. A dialogue, etymologically a conversation between (dia) thinkers (logos), traditionally presupposes two conscious, understanding subjects. This is precisely not the case here. When we speak of a dialogue with AI, we are using a metaphor that still suggests a human counterpart. The term dialogue is thus far superior to co-pilot because it describes the process, but it remains imprecise because it misconstrues the nature of the non-human partner.

The ‘cyborg’

The cyborg (as a fusion of human and machine) comes philosophically closer, especially in the sense of Donna Haraway’s influential concept that dissolves rigid boundaries. For describing artistic practice, however, the term is unsuitable.

The cyborg metaphor is mechanistic, cold, and often carries dystopian connotations. It implies a physical fusion, a prosthesis, a cybernetic optimisation. What we experience in AI art, however, is not a hardware upgrade for the artist, but an immaterial, poetic, and inspiring dialogue (in the limited sense above). The term cyborg captures the technical aspect of fusion but completely misses the creative, muse-like, and inspirational spark central to the artistic process.

The ‘aleatoric procedures’ from Leonardo to Richter

In the article, Artistic intelligence vs. artificial intelligence, the different concepts of creativity by humans and machines are being debated. For artistic and human based creativity aleatoric procedures are indispensable, creating rooms for the unknown, not only the unexpected.

The conclusion is that, implicit in this statement, programming and aleatoric processes are contrary to each other. According to computer art pioneer Frieder Nake, at most, we can speak of calculated randomness. On the part of computer science, it could be argued that random factors can now be implemented in computer programs and even unexpected processes can take place. However, even if processes that are so complex that they can no longer be understood by programmers take place, they are still calculated processes.

Following this argumentation, the conclusion can be to position artistic versus artificial intelligence.

Extelligence

Siegfried Zielinski offers with the concept of Extelligence, a profound and nuanced contextualization of Artificial Intelligence. By seriously engaging with the philosophical fundaments inherent in artistic practice, he suggests that the common term ‘intelligence’ can be a limiting and potentially misleading anthropomorphism.

He thoughtfully proposes the concept of Extelligence to help reframe the discussion. This term shifts the focus away from an internal, human-like ‘thinking’ process and towards the actual machinic function: a dynamic, external act of projection and transformation.

The post-human muse is a generative entity drawing from an immeasurable cultural data archive.

This approach allows for a crucial distinction that honours the unique qualities of human creativity. Zielinski carefully differentiates artistic intelligence—which thrives on the vital integration of chance (the aleatoric) and critical self-reflection—from the capacities of AI, which he characterizes as a powerful, complex system of imitation rather than creation in the reflective, human sense.

Finally, by situating AI within his expansive framework of the ‘Deep Time’ of media, Zielinski provides a valuable historical grounding for the current moment. This perspective thoughtfully reframes AI not as an unprecedented rupture, but as another significant variation within the long, ongoing co-evolution of human expression and technology.

A thesis: the post-human muse

If the old metaphors of the tool, the assistant, and the cyborg all fail, what metaphor can aptly describe the artistic practice of AI?

We propose a term that takes one of the oldest humanist metaphors in art—the muse—and radically re-contextualises it: the post-human muse.

This term incites a conceptual framework that helps artistic and artificial intelligence to overcome traditional prejudices as well as intellectual models hindering artistic freedom:

•               It preserves the concept of inspiration: The term muse has always described an external source of inspiration. It focuses on the spark, the unexpected, the given that ignites the creative process. This perfectly matches the experience of artists using generative AI: they are inspired by the model’s outputs to develop ideas they would not have arrived at alone.

•               It preserves the artist’s role: The muse (in mythology) does not create herself. She whispers ideas to the artist, but it is the artist who interprets them, gives them form, and completes the work. This metaphor solves the problem of authorship: the AI is the source of inspiration, but the artist remains the author, retaining final authority on meaning through selection, curation, refinement, and interpretation.

•               It is post-human (not inhuman): By adding the prefix post-human, we take the decisive step. We detach the muse from her humanist origins (a divine, anthropomorphic figure) and correctly situate her within the philosophical discourse of the 21st century. The source of inspiration is no longer divine or purely human, but a non-human, generative intelligence.

The post-human muse is not a partner in the human sense, but a generative entity drawing from an immeasurable cultural data archive. It is the ideal term for this new relationship: it respects the tradition of artistic inspiration whilst precisely naming the new, non-human nature of that inspirational source.

We are aware that this term—the post-human muse—is likely to meet with irritation from both sides of the debate. The traditional artist may reject the cold, philosophical post-human, as it seems to cement the dehumanisation of art. The AI programmer or engineer, conversely, will dismiss the muse concept as unscientific, esoteric, and imprecise.

We argue: this precise dual irritation is a good sign. It proves that the term hits the sore points of both camps, leaving the comfort zones of the purely humanist just as much as those of the purely technological. It is the perfect third term to start a long-overdue conversation about the true nature of this new creativity.

The characteristics of the post-human muse (the analysis)

The post-human muse is not a monolithic concept; it defines itself through its specific properties and behaviours. To understand its character, we must analyse how it operates within the artistic process. We can grasp its characteristics through four central metaphors:

The mutable muse (the aspect of mutability)

The humanist muse was static; she was the goddess of the epic or the tragedy. The post-human muse is fundamentally mutable. She is not a fixed entity but pure potentiality. She is fluid and malleable. She is redefined at every moment by the artist’s input. With every word in the prompt, every iteration, every re-roll, she changes her form. This mutability is the core of the interactive process: the artist not only shapes the work, but also permanently shapes the muse herself through dialogue.

The mirror muse / the echo muse (the aspect of reflection)

The post-human muse acts as a resonating body (echo) or a calculating mirror (mirror). She is not a passive recipient but an active counterpart in the dialogue. She reflects the artist’s intention back at him, but never 1:1. Like an echo shaped by the topography of a valley, the artist’s call (the prompt) is shaped by the topography of the latent space and returns altered. The mirror muse shows the artist his own idea from an unexpected perspective, thereby compelling him to react.

The latent muse (the aspect of the hidden space of possibility)

The muse lives in the latent space—that immeasurably vast, multidimensional space of learned concepts, styles, and connections that the AI model has built through its training. She is latent in the sense of hidden, slumbering. She is the embodiment of this infinite archive of possibilities. The artist becomes an explorer who, with his prompts, awakens this latent muse and navigates this hidden space to find things he did not know he was looking for.

The oracle muse (the aspect of interpretation)

Unlike an assistant who follows clear instructions, the post-human muse acts like an oracle. She responds to a question (the prompt) often enigmatically, ambiguously, and generatively. She does not deliver a finished result, but an answer that requires interpretation. This restores the artist’s role: he is the priest who must interpret the oracle’s enigmatic images, selecting, rejecting, and placing them in context. The AI provides the generative riddle; the artist creates meaning and the final work through interpretation.

Conclusion: implications for a new practice

The introduction of the term the post-human muse is more than a semantic exercise. The language we use directly shapes the reality in which we operate. By changing our vocabulary, we also change the way we create, teach, and critique AI art.

If the AI is no longer a tool or assistant but a muse, this could have at least three consequences:

•               For artistic practice: The focus shifts from pure prompt engineering—the attempt to technically operate a machine—to a genuine dialogue. The artist evolves from a technician to a hermeneuticist (one who interprets the oracle muse’s answers) and an explorer (one who navigates the latent muse). The creative act is no longer the generation of the image, but the entire process of interaction, curation, and interpretation.

•               For teaching: We can stop teaching AI in the arts as a pure software course (like operating photoshop). Instead, we must begin to teach the aesthetics of interaction and post-human theory. The questions are no longer: How do I get the AI to do what I want? but rather: How do I react to what the AI offers me?, How do I recognise the hidden biases in the muse’s mirror? and What does authorship mean in a dialogue?

•               For critique: Art criticism can move away from dismissing AI-generated images as mere fakes, imitations or cheap copies. It must recognise that the artistry lies not in the final pixel, but in the quality of the dialogue the artist has conducted with the muse. Criticism must evaluate the artist’s intention in dealing with the generative oracle, not the oracle itself.

The metaphor of the tool helped us to grasp the technology. The metaphor of the post-human muse —with all its facets as mutable, mirror, latent, and oracle — helps us to understand the art that emerges from it. It gives us a language for a mode of thinking that has overcome the assumptions of humanism – assumptions some might want to uphold, others might view them as obstacles. With this thought experiment I do not want to take sides in this debate, but rather help clarify arguments and thoughts by formulating more precisely what we talk about.

________

Pictured: AIS25, Copenhagen (l-r: Martin Brynskov, Scientific Director, AIS25; Curator of Society & Community, University of Copenhagen; Rasmus Larsen, President, Danish Society for the Dissemination of Natural Science; Professor at the Technical University of Denmark (DTU); Maria Russo, Director for Global Approach & International Cooperation in Research and Innovation, Directorate-General for Research and Innovation; David Dreyer Lassen, Rector, Copenhagen University; Serge Beglonie, President of ELLIS.

________

More proposals are welcomed, the post-human muse is hopefully a conversation starter for more profound and scientific concepts for the middling of artistic and artificial intelligence.

We hope this new terminology invites interested players for more mutual learning: a state in which artificial intelligence can learn more from artistic intelligence—from its freedom of expression and its capacity for creating meaning with large quantities of data —and, conversely, the artistic can profit more from the artificial, by accessing history in a new productive way, by learning quicker and discovering new possibilities in otherwise closed and clogged knowledge about culture and society.

Artistic intelligence has been trained over thousands of years, yet slowly and on low quantity. Artificial intelligence is being trained just a few decades, yet quickly and in masses. What is there to learn from each other? Are artists part of AI programming teams? or vice versa?

Artistic intelligence meets Artificial Intelligence – it has just begun.

_________

About the title picture ‘Self Portrait Claude Cahun’

Claude Cahun (1894–1954)

Born as Lucy Schwob in Nantes, Claude Cahun was a key figure of the Parisian avant-garde and Surrealism. Cahun identified as non-binary and used the gender-neutral first name as an artistic statement. Together with their partner Marcel Moore, Cahun created groundbreaking photographic self-portraits and photomontages, as seen in the major work “Aveux non avenus” (1930). From 1937, Cahun lived in Jersey, where both engaged in active resistance during the German occupation, were arrested in 1944, and sentenced to death. Following their liberation in 1945, Cahun remained in Jersey until their death. The work, only rediscovered in the 1980s, is considered pioneering for Gender Studies and 20th-century photography.

Cahun’s Surrealist Self-Conception Cahun’s self-portraits are a radical rejection of the humanist portrait and the idea of a true essence. Instead, Cahun demonstrates that identity is a theatre—a performative masquerade where beneath every mask lies only another mask. Through the play with dozens of roles (from dandy to doll), the artificiality of gender was exposed. To realise this fragmentation of the “I” visually, Cahun used surrealist techniques such as reflections (the self as object), double exposures (the “ghost image”), or confrontations with mannequins. The portrait we are using (c. 1928), with its shaven head, is exemplary of this: it rejects traditional norms and presents the subject as a neutral, ambiguous projection surface.

_________

This text was written with the help of Gemini.