AI Art Doesn't Exist (Yet): Art Across Cognitive Systems

Abstract

The contemporary debate around "AI art" is asking the wrong question. Much of what is currently described as AI art is neither genuinely artificial intelligence nor necessarily art. More fundamentally, the debate assumes that humans are competent judges of whether another cognitive system produces art. That assumption itself deserves scrutiny.

This essay proposes that art is not a property of objects but of cognitive systems. Art emerges through the interaction of makers and interpreters within shared worlds of meaning. Signals may travel between different cognitive systems, but meaning rarely does. As a consequence, humans may be no more capable of recognising genuine machine art than birds are of understanding the human world encoded in a mobile phone ringtone.

The question therefore is not "Can AI make art?" but rather:

What happens to art when the producing and interpreting cognitive systems no longer substantially overlap?

1. The Current Debate Is Mostly About Something Else

The current discussion surrounding "AI art" largely concerns image, music and text generators responding to human prompts.

Whatever one thinks of these systems, they are not the kind of artificial intelligence this essay is concerned with.

Today's generative models do not possess:

  • autonomous symbolic worlds

  • independent cultural traditions

  • persistent interpretive communities

  • self-sustaining artistic practices

The creative framework remains overwhelmingly human.

Humans decide:

  • what to generate,

  • which outputs to keep,

  • what they mean,

  • who the audience is.

Calling these outputs "AI art" therefore conflates human artistic practice using computational tools with the possibility of art emerging from an independent non-human cognitive system.

This distinction matters.

Just as László Moholy-Nagy could create artworks over the telephone without the telephone becoming the artist, artists can employ generative systems without those systems becoming artists.

Likewise, not every image is automatically art. Most generated images function as illustration, decoration, visual communication or entertainment. Some undoubtedly participate in artistic practice—but this is because of how humans use them, not because the software itself has become an artist.

The genuinely interesting question lies elsewhere.

2. Art Exists Within Cognitive Systems

Art is not simply an object.

Nor is it reducible to intention.

Instead, art exists within cognitive systems capable of producing and interpreting it.

Interpretation has two relatively independent poles:

  • the artist's understanding of the work

  • the audience's understanding of the work

Neither has absolute authority.

An artist may create work that no audience understands.

Conversely, observers may experience art where none was intended:

  • landscapes

  • found objects

  • accidental images

  • natural phenomena

Art therefore emerges through interpretation, but interpretation is not monopolised by either maker or audience.

3. Shared Worlds Make Interpretation Possible

Successful interpretation depends upon overlapping cognitive worlds.

Members of the same culture often share:

  • language

  • symbols

  • myths

  • practices

  • assumptions about reality itself

Medieval religious imagery provides an instructive example.

Church decoration was not mysterious to contemporary audiences. Images were deliberately used to educate largely illiterate populations. Their symbolism was reinforced through sermons, ritual and everyday belief.

More importantly, medieval viewers inhabited a different ontology.

Saints were not fictional characters.

Miracles were not merely symbolic narratives.

The supernatural formed part of lived reality.

Modern viewers may understand the stories while lacking the worldview that originally animated them.

Consequently, we do not simply lose symbolic meaning—we lose the cognitive world that made the symbols meaningful.

4. Interpretation Weakens Across Cognitive Systems

Even within humanity, interpretation changes across cultures and centuries.

Across different cognitive systems it becomes more fragile still.

This does not necessarily result in misunderstanding.

Instead several possibilities emerge:

  • successful interpretation

  • partial interpretation

  • reinterpretation

  • perception of only structural features

  • complete imperceptibility

Consider birds adopting mobile phone ringtones into their songs.

The bird almost certainly does not understand:

  • telephones

  • communication networks

  • human technological culture

It responds only to acoustic properties useful within its own signalling system.

Likewise, humans looking at prehistoric cave paintings recognise intention and symbolic structure while remaining unable to reconstruct with confidence the worldview from which those paintings emerged.

Interpretation therefore depends upon cognitive overlap.

5. Signals Travel; Meaning Does Not

The key distinction is between signal and meaning.

Signals migrate.

Meaning usually has to be reconstructed locally.

Consider the following chain:

Beethoven composes the Ninth Symphony.

Humans compress fragments into a mobile phone ringtone.

Birds adopt that ringtone into birdsong.

A sound artist records the birds and incorporates those recordings into a new artwork, as in Machine Auguries by Jennifer Walshe.

What survives across this chain is not Beethoven's meaning.

What survives are fragments of signal:

  • melodic contour

  • rhythm

  • pitch relationships

  • acoustic identity

Each cognitive system reconstructs entirely new meanings.

The ringtone becomes birdsong.

Birdsong becomes contemporary sound art.

The signal survives.

Meaning is continually reborn.

6. Art Behaves Like Ecology

Art therefore behaves less like an object than an ecological process.

Signals propagate through:

  • individuals

  • cultures

  • technologies

  • species

Each system absorbs, reflects and transforms incoming signals according to its own cognitive organisation.

Art is less a static object than a pattern of continual reinterpretation.

Like vibrations travelling through different materials, artistic signals are refracted rather than simply transmitted.

7. AI Changes the Scale, Not the Principle

The AI question therefore becomes much deeper than current debates suggest.

If future machine cognition develops independent symbolic cultures, humans may occupy the same position with respect to machine art that we currently occupy with respect to:

  • prehistoric symbolic worlds

  • birdsong

  • other animal signalling systems

Humans may:

  • perceive fragments,

  • admire certain structural qualities,

  • reinterpret them according to human culture,

  • fail entirely to perceive the artistic practice.

The issue is not whether machines produce images.

The issue is whether humans possess sufficient cognitive overlap to recognise genuinely non-human artistic practice.

There is no reason to assume they will.

8. There May Be Bleed Between Worlds

Importantly, cognitive systems are not completely sealed.

Signals bleed across boundaries.

A bird may appropriate human ringtones.

Humans may incorporate birdsong into music.

Machine systems may one day appropriate fragments of human culture.

These crossings do not preserve meaning.

They generate new meanings inside new cognitive systems.

The resulting artefacts become reflections rather than copies.

This continual migration resembles vibrations echoing through different bodies, each body reshaping the incoming wave according to its own physical and cognitive properties.

9. The Conclusion

This essay is ultimately not about artificial intelligence.

It is about the limits of interpretation.

Humans already fail to fully understand:

  • prehistoric humans,

  • distant cultures,

  • other species.

Future machine cognition simply extends an existing problem rather than creating a new one.

Consequently, the question:

Can AI make art?

is probably the wrong one.

A better question would be:

If another cognitive system developed its own artistic culture, how would we know?

Perhaps we would not.

Perhaps we would recognise only faint echoes—surface structures, aesthetic traces, or attractive patterns—while the underlying symbolic world remained as inaccessible to us as the human world is to a bird singing a ringtone.

In that case, genuine AI art may one day exist without humans ever becoming its primary audience—or even realising that they are standing in its presence.