Can AI Be Creative? Schelling’s Answer

Something strange happens when you watch a generative AI produce an image, a poem, or a piece of music. The output looks creative. It has form, variation, even surprise. And yet a quiet dissonance remains — a sense that what you are witnessing is not creation but a sophisticated form of reassembly. You cannot quite name what is missing. But you feel it.

That feeling has philosophical weight. It points to a distinction that Friedrich Wilhelm Joseph Schelling articulated more than two hundred years ago — one that the current conversation about artificial intelligence has almost entirely forgotten.

The Strongest Argument for Machine Creativity

Before criticizing it, the argument for AI creativity deserves to be heard in its strongest form. It can be stated in three steps.

First: Creativity consists of certain observable features — novelty, surprise, the combination of previously unconnected elements, the generation of something that did not exist before. Second: If creativity is identical with these features, then any system that produces them is creative. Third: Generative AI demonstrably produces novel, surprising, formally compelling results — therefore it is creative.

The argument has an internal consistency that must be acknowledged. If you define creativity by its results and then observe that a machine produces such results, you cannot meaningfully deny within that framework that the machine is creative. The framework is internally coherent. A reader who compares AI-generated images with Goya’s etchings or Bach’s fugues and notices no difference probably finds this argument convincing. It would be dishonest not to state it at full strength.

The real question is therefore not: Does AI produce creative results? But rather: Is creativity identical with its external features? Is novelty the same as creation?

Here lies the error — and it lies not in the third step but in the first. The argument defines creativity by the surface of the result and thereby confuses the property of the product with the nature of the process. But assembling and creating are not the same thing. A kaleidoscope generates infinitely many novel patterns without anyone creating anything. The confusion of novelty with creation is not a weakness of the argument — it is its premise.

The Question No One Is Asking

The public debate about artificial intelligence and creativity revolves around a curiously narrow question: Can machines produce outputs that are indistinguishable from human-made art? This is a question about surfaces, about results. It asks whether the product looks right — not whether the process that generated it has any relationship to what we mean by creation.

Schelling would have recognized this confusion immediately. He spent the better part of his philosophical life fighting a version of it. In his Ideen zu einer Philosophie der Natur, he drew a sharp line between two ways of understanding the natural world (Schelling, 1797). You can describe nature from outside — measure it, calculate it, map its regularities. Or you can understand nature from within — as a living activity that produces itself.

The first approach gives you what Schelling called natura naturata: nature as product, as finished thing, as the collection of objects that can be catalogued and quantified. The second gives you natura naturans: nature as creative process, as the inner striving through which forms emerge from a living ground.

This distinction is not academic. It cuts to the heart of what creativity actually is.

What the Mathematical Description Misses

Schelling was unsparing on this point. The merely mathematical description of nature, he argued in substance, is without real cognitive value, even though it offers accuracy. It is like describing Homer’s works by counting the characters. Of the inner movement, one knows nothing. Natura naturans — the inner striving of nature — must be understood. Otherwise it is no natural science at all (cf. Schelling, 1797, Introduction).

Natural philosophy insists on this distinction. This is not a rejection of mathematics or precision. It is a claim about what counts as understanding. You can map every pattern in a dataset. You can model every statistical relationship between words, pixels, tones. The result may be extraordinarily accurate in reproducing surface regularities. But if you have no access to the inner life from which those patterns emerged — if you are working exclusively with natura naturata, with the products rather than the productive activity — then you have not understood anything. You have only counted the characters in Homer.

Generative artificial intelligence operates entirely within natura naturata. It works with patterns abstracted from finished products — text already written, images already composed, music already performed. Its training data is, by definition, a vast collection of completed forms. The machine finds statistical regularities across these forms and generates new combinations that are consistent with the patterns it has absorbed. This is impressive engineering. It is not creation.

Nature Creates from Within

The decisive move in Schelling’s natural philosophy is the claim that nature is not assembled from outside but produces itself from within. “Life is not a property or product of animal matter,” he wrote in Von der Weltseele (Schelling, 1798). “On the contrary, matter is a product of life. The organism is not the property of individual natural things; on the contrary, the individual natural things are just so many limitations or individual modes of viewing the general organism.”

This reversal is radical, and it has direct consequences for the AI question. If creativity is the activity of a living whole producing new forms from within itself — if the creative act is a participation in something that is already alive and self-organizing — then a machine that recombines patterns from a fixed dataset is doing something categorically different. It is assembling, not creating. It is working from outside, not from within.

The creative human being, by contrast, stands in a relationship to the living whole. Goethe understood this when he developed his method of anschauende Urteilskraft — intuitive judgment — (Goethe, 1817), a thinking observation in which the observer participates in the phenomenon rather than standing apart from it. Novalis understood it when he wrote in Die Lehrlinge zu Sais that the thinking person “returns to the original function of existence, to creative contemplation, to that point where producing and knowing stood in the most wonderful reciprocal connection” (Novalis, 1798/99).

Creativity, in this tradition, is not the generation of novel combinations. It is the moment when a living being participates in the self-producing activity of nature and gives form to something that emerges from that participation. The artist does not assemble a work from pre-existing parts. The work comes through the artist — from a ground that is deeper than the individual.

What AI Actually Does

It is important to be precise here. The argument is not that AI produces bad art or uninteresting outputs. Some AI-generated work is genuinely compelling. The argument is that what happens inside the machine has no structural relationship to what happens inside a creative human being.

A large language model processes sequences of tokens. It predicts the next most likely token given a context window of previous tokens. The prediction is based on statistical relationships learned from billions of examples. There is no interiority, no participation in anything, no relationship to a living whole. There is pattern and probability — nothing more, and nothing less.

Schelling would have called this the “economically-teleological” approach to nature that he found so troubling in Fichte: a framework in which nature “is only to be used, exploited. Its entire existence amounts to the purpose of its processing and utilization by human beings.” The AI does not encounter what it processes. It uses data as raw material for pattern recombination. The question of what the words mean, what the image shows, what the music expresses — these questions do not arise for the machine because there is no one inside for whom they could arise.

Pathogenesis, Not Progress

There is a broader context here that matters. The philosopher Jochen Kirchhoff identified a persistent pattern in modernity: what presents itself as progress often turns out to be a disease process (Kirchhoff, 2002). “Transhumanism is the logical continuation of abstractionist natural science,” he argued. “It did not fall from the sky but developed out of the ultimately materialist, reductionist natural science since Galileo.”

The claim that AI is “creative” belongs to this pattern. It sounds like progress — machines that can do what humans do, only faster and at scale. But the claim works only if you have already reduced creativity to its external features: novelty, variation, complexity of output. If creativity is participation in a living whole, then calling AI creative is not a discovery but a redefinition — one that quietly eliminates the very dimension that makes creativity meaningful.

This is not a minor philosophical quibble. It shapes how an entire culture understands what it means to create. If we accept that pattern recombination is creativity, we lose the capacity to distinguish between the poem that emerges from a human being’s encounter with the depths of experience and the poem that emerges from a statistical model trained on other people’s encounters. The outputs may look similar. The processes could not be more different.

The Living Thought and the Dead One

Schelling made a distinction that illuminates this further. There are living thoughts and dead thoughts. Dead thoughts are abstract entities circulating in the intellect — “mere units of information,” as Gwendolin Kirchhoff puts it. Living thoughts are embodied, generative, producing effects in life. The difference is not one of complexity but of aliveness. A living thought participates in something real. A dead thought merely represents.

AI generates what Schelling would recognize as dead thoughts at extraordinary speed and scale. The tokens have no felt dimension, no relationship to experience, no inner life. They are, in his precise terminology, “modifications of the infinite” without any contact with the infinite itself.

This is why the output of generative AI can feel simultaneously impressive and hollow. The form is there. The substance — the participation in something alive — is not. You sense the absence even when you cannot name it.

What Remains Irreplaceable

The question is not whether AI will replace human creativity. It cannot, because what it does is not the same kind of activity. The question is whether a culture that confuses the two will still be able to recognize creativity when it encounters it.

Schelling saw clearly that the mathematical description of nature, however precise, misses what matters most: the inner striving, the self-organization, the living ground from which all forms emerge. “As long as I am myself identical with nature,” he wrote, “I understand what living nature is as well as I understand my own life. I grasp how this universal life of nature reveals itself in the most manifold forms, in graduated developments, in gradual approximations to freedom. But as soon as I separate myself and with myself everything ideal from nature, there remains nothing for me but a dead object. And I cease to grasp how a life outside me is possible.”

The creative act is precisely this: a moment of non-separation, in which the creating person participates in the self-producing life of nature and gives it a form it did not have before. No amount of computational power can simulate this participation, because simulation and participation are different in kind.

What remains irreplaceable is not the human capacity to produce novel outputs — machines do that now, and they will do it better. What remains irreplaceable is the human capacity to stand within the living whole and allow something to come through that could not have come from pattern recombination alone. Schelling called this natura naturans. We might simply call it the creative ground.

Whether a culture remembers how to stand on that ground — or trades it away for the convenience of automated novelty — is the real question the AI revolution poses. It is not a technological question. It is a philosophical one. And it was answered, in principle, more than two centuries ago.

Sources

• Goethe, J. W. von (1817). Anschauende Urteilskraft [Intuitive Judgment]. In: Zur Naturwissenschaft überhaupt, besonders zur Morphologie. Stuttgart/Tübingen: Cotta.
• Kirchhoff, J. (2002). Die Anderswelt: Eine Annäherung an die Wirklichkeit [The Otherworld: An Approach to Reality]. Klein Jasedow: Drachen Verlag.
• Novalis (1798/99). Die Lehrlinge zu Sais [The Novices of Sais]. Unfinished novel, posthumously published 1802. In: Novalis Schriften, ed. Richard Samuel. Stuttgart: Kohlhammer.
• Schelling, F. W. J. (1797). Ideen zu einer Philosophie der Natur [Ideas for a Philosophy of Nature]. Leipzig: Breitkopf und Härtel.
• Schelling, F. W. J. (1798). Von der Weltseele [On the World Soul]. Hamburg: Friedrich Perthes.