When AI Begins to Create — Different Boundaries in Knowledge Production and Artistic Creation

I recently watched a video featuring a doctoral student in composition at a music conservatorium in Germany. She was broadly opposed to the idea of AI-created music. In her view, AI can produce music that meets basic formal requirements, but it does not possess the inspiration of a human composer. Her argument prompted some thoughts of my own about the different effects AI is having on knowledge production and artistic creation.

AI does not present the same challenges to knowledge and art. On the surface, the threshold for AI to create art is clearly higher than the threshold for it to carry out ordinary knowledge production. Knowledge generally operates within an existing logical framework, whereas the distinctive quality of art often lies precisely in breaking an established pattern. Yet knowledge production and breakthroughs in knowledge are not the same thing. Treating them as though they were easily creates a conceptual imbalance.

The threshold is not especially high for AI to summarise, explain, combine and reason within an existing body of knowledge. Asking it to produce a theory such as general relativity, which changes the coordinates of an entire system of knowledge, is another matter. That task is no easier than creating a genuinely original work of music. The shared difficulty AI faces is not producing content. It is stepping outside the framework that produced the content in the first place.

The difference between knowledge and art is that most knowledge production does not require a break with the existing framework. Knowledge demands stable concepts, logical consistency and reliable evidence. It cannot cross established boundaries merely for the sake of novelty. If AI can operate accurately enough within an existing framework, it can already perform a great deal of valuable work. Art is different. A work of music must still follow certain basic principles. Rhythm, pitch, harmony, tension and auditory expectation must form a perceptible structure. They cannot change without reason and completely disregard the natural habits of human hearing. But a truly outstanding work cannot merely be correct. It must also create surprise within the rules and may even redefine what can be heard as music.

In this sense, an outstanding musical work requires something like a Beethovenian breakthrough, just as scientific theory can contain an Einsteinian breakthrough. It may not overthrow every musical rule, but it opens possibilities within those rules that did not previously exist. This is why creating art presents AI with such great difficulty. Most of what it generates may follow the general principles of music and may even sound quite good, but that alone is unlikely to give it genuine artistic distinction. By contrast, knowledge produced by AI is not necessarily defective simply because it remains within an existing framework. Respecting the established framework is precisely what most knowledge production requires. AI therefore faces very different conditions in knowledge production and artistic creation.

The question is not only whether AI can generate music that sounds good, but whether music that sounds good is necessarily art. Most ordinary listeners primarily consume the function of music. They care whether it is pleasing, whether it can evoke emotion and whether it suits a particular setting. From this perspective, a composer’s personal experience, creative impulse and source of inspiration are not necessary conditions for a listener to be moved. A listener may be genuinely affected by a piece of AI-generated music. Once music enters human hearing, the emotion it produces does not simply disappear because its creator was not human.

Fruit grown hydroponically may look more uniform, and it does not lack nutrition. Its nutrients may even be more balanced. What it seems to lack may not be any measurable ingredient, but the imagined scent of the soil. AI music is much the same. Judged solely by melody, rhythm, harmony and sound, it may lack nothing. It may even sound better than the work of many mediocre composers. Most composers in the real world are, after all, ordinary, and some of their work genuinely does not sound as good as music created by AI. Yet when people know that no particular person stands behind a work, and that it arose from no lived experience, desire for expression or irresistible urge to write, some listeners will still feel that something is missing.

The question is whether that missing element exists within the work itself or within our understanding of its creator.

Serious listeners, critics and music historians do more than listen to a work. They ask why it appeared, what personal experience and historical circumstances it answered, and where it changed the musical language that preceded it. At this level, a work of art is not merely a product made of sound. It is an expression made by a subject in response to the world. Beethoven’s music matters not only because it sounds good, but because, under particular historical conditions, it expanded what music could express. If AI generated a work of equal structural quality but possessed no experience, intention or self-awareness, it might complete the musical form without completing the artistic act in its traditional sense.

This does not mean that AI has no artistic value. Art has never existed only at the creator’s end. It is also generated anew in the listener. Even if AI has no emotion, the listener still does. AI may have no creative impulse, yet the listener may discover personal experience within its work. AI may therefore be unable to express itself as a human does, but it can create a structure through which people express and discover themselves. In this sense, AI art may not be art without a subject. Instead, the position of the subject may have shifted partly from the creator to the selector, editor and listener.

AI is good at generating results within existing structures, but that ability means different things in knowledge and art. In knowledge production, it is often useful enough because most of the work requires adherence to a framework. In art, following the framework can produce work that is competent, pleasing and even beautiful, but not necessarily work that redefines art. AI can replace a great deal of mediocre intellectual labour and a great deal of mediocre artistic production. Whether it can replace Einstein or Beethoven, however, comes down to the same question. Can it do more than generate content within a world? Can it establish a new set of coordinates for that world?

There is, however, a more important difference between works created by AI and works composed by composers themselves. Creation is not only a way of producing something for an audience to consume. For the composer, it is also an enjoyable process in itself. An artist does not necessarily endure the process of creation simply to add another piece of music to the world. Often the reverse is true. The work comes into being because the artist needs to create and takes pleasure in creating. Conceiving, experimenting, revising, failing and suddenly finding a sound are all part of an artistic life.

From the listener’s perspective, AI may replace some of the artist’s work and may even produce works that better satisfy the demands of consumption. From the artist’s perspective, however, AI cannot undergo this part of life on the artist’s behalf. A car can run faster and further than a person, but that does not mean people no longer need to run. A car replaces running as a means of transport; it does not replace running as a human physical activity. In the same way, AI may replace some of the productive function of artistic creation, but it cannot replace creation as the artist’s own mental activity. People still eat with their own mouths, and artists must still create with their own minds. Eating is not merely a way of making food disappear, just as creating is not merely a way of making a work appear. Even when a machine can produce a more polished result, the artist still has reason to go through a process that is inefficient, uncertain and often marked by failure. This is not a denial of AI’s capabilities. It is part of what it means for an artist to be an artist.

The final question, then, is not only whether humanity wants music that sounds good or a person addressing the world through music. There is another side to it. Does the artist still need to become who they are through the act of creation? If all that is required is good-sounding music, AI is already a powerful competitor. If music also contains lived experience, the urge to express and the process of creation itself, then AI can mainly replace the production of the work. It cannot replace the reason the artist creates. Both can produce music, but they may not produce art in the same sense.