It is the summer of 2021 at 1am. I sigh and rub my eyes. Rolling my shoulders back and sitting a little more upright on the sofa, I go over the draft one more time.

“You are one of the most hard-working people I have ever met - and this includes myself, so that really means something.” I think back to the first months of our relationship when he complained about the many late night study sessions and 6am wake ups. “Men are pussies”, I mumble to myself, smiling.

Then I continue to write what would eventually become my wedding vows. I didn’t want the speech to merely sound polished, but for every word to feel earned.

When I wrote my vows, ChatGPT didn’t exist. I spent an absurd amount of time on finding the right words. I am no longer sure many people still would, now that beautiful prose has become incredibly easy to generate.

But then a vow is not just a beautiful text. It is an attempt to hold still inside your own feelings long enough to arrive at something worthy of being spoken aloud in public on what is meant to be the most important day of your life. It is not so much a writing exercise as a small act of authorship and a confrontation with your own emotional reality.

And that, I think, gets at the heart of the confusion in the debate about whether AI will kill creativity. Because when we ask whether AI can be creative or whether it makes us less so, we are often collapsing different things into the same question. The quality of the output. The originality of the idea. The presence of a human self behind the work. And lastly, the inner process of becoming equal to what one is trying to say.

We are all suffering from creative confusion

Every day, I hear some version of the curious tale that is AI and creativity. There is the hero’s perspective: Some are elated about AI superpowers that turn even the biggest bores into Hollywood directors.

There is the villain’s warning: Others whisper that the death of human creativity is upon us. That we are creatively withering away, shackled by machines, losing any original human touch.

There is the philosopher’s objection: those who outright deny that AI can be truly creative. They criticise that AI models are the mere mimicry of existing patterns, but cannot come up with entirely new ideas altogether. You know, what us humans do so effortlessly on a daily basis…

And so, I wrote this piece as an exploration of this mystery. First, the history of creativity. Then, what AI can actually do with it. Next, the different social dimensions of creativity that all get mistaken for one. And finally, how we can become more creative with AI.

A short history of creativity

What is creativity and where does it come from?

For most of history, creativity was not understood as an ordinary human capability. The very idea that humans can be creative is, in fact, a surprisingly recent one. “Creativity” only became an everyday noun some time between the mid-17th century to 19th century. The word create itself goes back to the Latin creare: to bring forth, produce, even originally to make grow.

We can thus think about creativity as a sort of modern social arrangement. It is arguably a post-industrial compensation: as routine cognitive and physical labor got automated or outsourced, Western culture elevated creativity as the one remaining proof of uniquely human worth.

As a consequence, we now live in a world that has turned creativity into a cult. Being uncreative is a damning insult, right up there with being dumb. Actually, I believe many people may rather be called dumb than uncreative, as being dumb can still be sexy, but being uncreative is just plain boring.

During antiquity, humans were not thought to be creative. Instead, creation was divine territory. According to Aristotle, art is mimesis (imitation) and techne (craft), but humans were not bringing something radically new into existence. They could merely give divine inspiration shape and form. Plato equally saw the artist as a vessel and the Gods as the source. Later on, Thomas Aquinas famously stated that “to create is proper to God alone.”

During the Renaissance and the Enlightenment, the artist as the originator gradually came into focus, and with it, the concept of imagination. Romanticism delivered the decisive shift: creativity moved from the divine into the person, with human genius becoming the source of truly original thoughts and rules rather than merely following them.

By the twentieth century, the arc was complete: creativity had become the distinctly human capacity to bring something genuinely new into the world. And so, as humans claimed more and more credit for their creation, creativity became an expression of what it means to be human at all.

In 1988, the psychologist Mihaly Csikszentmihalyi proposed that creativity isn't so much genius happening in isolation, but a social process. It requires an individual, a domain that transmits existing knowledge, and a field - critics, curators, thought leaders - that decides what counts as genuinely new and what gets included. Creativity only becomes creativity once the field says so. We will come back to that.

The AI moment is, in this sense, a structural reversal of that entire arc. We arrived at claiming creativity as uniquely human just as machines began demonstrating something that looks, from the outside, quite indistinguishable from it. That matters because if creativity had remained merely skill or techne, generative AI would be much less existentially unnerving. Instead, it is so unnerving because modernity taught us to treat creative output as evidence of our personhood.

Can AI be truly creative?

The key scientific debate about AI and creativity asks whether AI can truly create new output, rather than just predict from within the distribution of patterns it has been trained on. AI can discover surprising points in a landscape, but whether it can generate a genuinely new landscape is a different question altogether.

But does creativity really require that? I would argue that many forms of creativity find unusual combinations within an existing space. Think about your average song. Research now suggests that music across genres is becoming less complex and more repetitive over time. We are endlessly making more of the same hit song pattern and yet we still call it creative.

Every now and then, though, creativity reaches its peak and alters the space itself. Rather than rearranging available elements, it introduces new dimensions. The invention of calculus didn’t combine geometry and algebra in a clever way, but introduced an entirely new formal language that let humans reason precisely about motion and change for the first time. Jazz did something similar: it created a new harmonic and rhythmic language rather than recombining existing European forms.

This distinction between exploring a space and expanding it is precisely what a good deal of frontier AI research is now wrestling with. The work on open-endedness seeks to build systems that keep generating structural novelty rather than converging on a solution and stopping.

This is a much more ambitious vision than most people realise. It’s one thing to produce endless variation. It’s another to produce variation that is actually interesting. And interestingness is not some fixed property sitting out there waiting to be measured. It is contextual and relational, much like Csikszentmihalyi’s field deciding what counts as creative.

In other words, surprise may be a better proxy for creativity than quality. A polished output is not necessarily creative, but a surprising one might be. Yet even surprise is not enough if it is merely random. Peak creativity seems to require a rarer combination: novelty that has intention.

This is why the real disagreement about AI and creativity is often not about technology at all. It is about what kind of novelty people think matters.

On that question, the empirical evidence is not kind to AI skeptics. In most creative domains - musical, visual, literary - current AI models already produce outputs that many people experience as creative. They write elegant prose, generate images indistinguishable from real photos, compose viral music, and combine references in ways that feel genuinely surprising. In a blind test, The New York Times recently found that 54% of readers preferred AI-generated content over human-generated content. Current AI models are merely predicting the internet back at us, but it turns out that makes for some very enjoyable output.

Many skeptics who still insist AI cannot be truly creative often do so less from deep scientific conviction than from a place of arrogance. The reality is that most people are average at most things. That is, definitionally, what average means. Yet we hold AI to a superhuman standard while failing to produce so much as a recognisable insect when asked to paint one in watercolour. We are remarkably generous critics of everyone but ourselves.

Recombination embedded in a life

One reason I find this debate so interesting is that it has an old predecessor. In the late eighteenth century, the English poet Samuel Taylor Coleridge drew a distinction between fancy and imagination. Fancy, for him, was associative, mechanical, and recombinatory. It reassembles existing material. On the contrary, he believed that imagination does something more profound. It transforms the nature of what it touches. Rather than merely rearranging existing materials, it alters them in the act of creation.

It is difficult to read that distinction now and not notice how closely it maps onto the contemporary debate about AI. Fancy sounds suspiciously like what current generative systems do extraordinarily well: combining, varying, and reconfiguring. Imagination sounds like the more sacred thing people are still trying to protect: transformation.

I believe that Coleridge’s distinction, while framed as a cognitive one, was really a social one. Because perhaps the reason imagination feels different from fancy is not merely that the mechanism itself is different. Perhaps it feels different because a person stands behind it. A person with a history. A person with something at stake. A person who risks failure, exposure, ridicule and misunderstanding. A person who is not just producing an output, but being altered by the act of making it. Imagination, in this view, is not simply superior recombination. It is recombination embedded in a life, and a life embedded in society, and a society embedded in a world.

That would explain why a wedding vow written by hand matters differently from a beautiful sentence generated by AI. Why a love song feels different when someone has actually loved. Why a speech matters differently when something is at stake for the speaker. Why even mediocre human work can sometimes feel more alive than technically superior synthetic work. The difference is not always in the artefact itself, but in the relation between the artefact and the being who made it.

This is where I think many supposedly technical arguments about AI and creativity really become arguments about the social dimensions of creativity and the conditions under which meaning is made.

And that is where things get really interesting.

The three social dimensions of creativity

When considering the different social dimension of creativity, we realise that we use one word for different things. Just like with the varying degrees to which creativity requires novelty, we are suffering yet another sort of contemporary confusion. I believe that when we say “creative” in the social context, we typically mean either of the below:

Sometimes what we mean by creativity is really productivity - making something. Sometimes we mean authorship - expressing who you are. And sometimes we mean psychology - discovering who you are. All three can be found in the word’s etymology: to produce and to make grow, to be born and to arise, but they are not the same thing. They follow different social expectations, have different kinds of value, and are affected by AI in very different ways. A great deal of the current debate consists of people discussing one of these meanings while assuming everyone else is talking about the same one.

1. Creativity as productivity

In the first sense, creativity is about output.

Here, to be creative means to generate options, drafts, concepts, combinations, campaigns, images, names, slogans, layouts and videos. Quantity, speed and iteration matter. The economic logic is abundance. The implicit contract is simple: make more, faster, cheaper and better.

In this domain, AI is an extraordinary multiplier. It reduces scarcity, making first drafts cheap and allowing people to explore a much wider range of possibilities than they otherwise could. A marketer can test ten campaign directions instead of two. A founder can prototype an entire landing page in an afternoon. Even a person with relatively little craft can suddenly produce output that looks polished.

This does not mean taste stops mattering - quite the opposite. As generation becomes easier, the bottleneck, and therefore much of the value, shifts elsewhere. The scarce things are no longer raw production, but selection and judgement. What matters is not only whether you can make something, but whether you know what is worth making in the first place because of your unique understanding of the social context in which the output will be consumed.

2. Creativity as authorship

In other instances, creativity means something else entirely.

Sometimes what people mean when they call something creative is not simply that it exists, but that it comes from someone. Here, creativity is tied to a dedicated point of view or a specific biography. The social contract is no longer efficiency, but origin and authorship. Show me who you are by showing me what is distinctly yours.

This is where AI enters shakier territory because it muddies the relation between the work and the self behind it. It changes, or at least complicates, the work’s origin.

Authorship matters enormously in domains like publishing, art, music and public speaking - really wherever the perceived relation between creator and creation defines the meaning of an output. A handwritten note by a known artist is valuable not only because of the words on the page, but because they represent a trace of the person behind them.

In the age of infinite generation, creativity is therefore no longer only about making more. It is increasingly about meaning more by maintaining tangible traces between the work and its originator.

3. Creativity as psychology

Lastly, there is a third dimension in which creativity has very little to do with public output at all.

Sometimes creativity is about what happens inside the person making. It is a process of experimentation with identity, desire, fantasy, shame, courage, permission, play and voice. Instead of external production, creativity is about inner transformation. Not “I made something”, nor even “this is me”, but something much more intimate: “I am discovering who I am”.

Think about the act of journaling, doodling and fantasising. All of these are hugely creative processes that fall entirely outside the realm of both productivity and authorship. Instead, they are expressions of creativity as psychology.

This is where AI becomes especially interesting. It can function as a uniquely gifted sparring partner, rehearsal space, confessor, ideation companion, therapeutic mirror, fantasy multiplier, or identity simulator. It allows people to try on selves, aesthetics, tones, arguments and even simulate different futures with remarkably low social risk. A difficult conversation can be rehearsed before it is had. A private fantasy can be explored without immediate exposure. A person can move through shame or uncertainty faster because the first audience is not a room full of humans, but your most intimate machine.

That can be profoundly generative. In this inner sense, AI can make people significantly more creative because it helps them access parts of themselves they otherwise could not reach so easily.

But with this possibility also comes a big ambiguity. AI can enlarge a self, as much as it can replace one. It can help someone become more substantial, more daring, and more articulate. Or they can become a kind of hollow shell, endlessly sifting through styles, thoughts and identities that the person never quite metabolises into a point of view of their own.

So in the psychological dimension, the question is not simply whether AI helps people make, but whether it helps them become.

Where these dimensions overlap

Importantly, these three dimensions are not mutually exclusive. Most creative acts sit somewhere between them.

Writing your wedding speech, for example, is about authorship (publicly expressing what you feel for your partner), but also deeply psychological. Writing a public blog post may be partly about productivity (regularly producing work of great quality), authorship (expressing who you are) and psychology (working towards who you want to become). Deciding the creative direction for a campaign sits between productivity and authorship. Rehearsing a difficult work conversation sits between productivity and psychology.

This overlap - and the ensuing complexity - is precisely why the debate about creativity and AI becomes so muddled. We constantly reach for one word to describe situations that involve very different expectations and norms. In one case, AI assistance feels efficient and welcome, even very much mandated. In another, it feels like contamination. In yet another, it enables liberation.

This means several apparently contradictory statements can all be true at once. “AI makes people more creative” may be true in the productive and psychological sense. “AI makes people less creative” may be true in the authorial sense.

I don’t believe there is a hierarchy of the social dimensions of creativity. Authorship isn’t more important than psychology or productivity. They simply matter at different times and in different contexts.

What we cannot negate, however, is that in the psychological and authorial dimension of creativity, the human matters enormously. Even identical output may be judged differently if the reader knows a human made it. Meaning is not merely the artefact, but sits inside the relation between artefact and maker.

Yet I believe AI-generated output is often unfairly stigmatised. Productivity is a hugely important dimension of creativity, and in many contexts we care far more about speed and quality than about the personal relationship between maker and object. Humans care a great deal about reducing friction. Meeting note summaries, ad copy variations, lesson plan drafts, even a decent bedtime story on demand: many creative acts are judged less by who made them than by whether they are frictionless and quite simply: good.

Becoming more creative with AI

Lastly, you may wonder how we can leverage AI to become more, rather than less creative?

Here’s what I have been observing. AI is making some people feel almost superhumanly creative. But rather than giving the creative have-nots superpowers, it often seems to give already creatively inclined people something closer to ultrapowers.

Highly creative people differ not only in idea generation, but in how they relate to possibility itself. They often notice more of what goes on around them, connect distant concepts, tolerate ambiguity longer and risk public failure more often. AI elevates these qualities.

I don’t believe that image, video and text models make everybody equally creative. I believe that AI turns pre-existing asymmetries in creativity into larger asymmetries in output. AI increases the returns to taste, selection, intrinsic motivation and the courage to explore before the result is polished.

In that sense, there is even a case to be made in defence of slop. Not because slop is good quality, but because the path to anything distinctive is often littered with a great deal of awkward, excessive and half-formed work. You need to build a muscle of trying things by…trying them.

Like with any creative tool - the use of watercolour, clay or photography - the first tries and explorations likely won’t be masterpieces yet. In some cases, they will look like an actual heap of shit. Chances are that they also won’t yet feel distinctly yours. They’ll be playful explorations. In the beginning, what you produce is often less a finished statement than a form of searching. By adopting tools before they are perfect and wrestling with the jagged edges of their capability, we can hope to eventually produce something that genuinely has our signature on it.

That early mess is not just about producing artefacts. It is about learning the contours of your taste. It is about discovering what draws you in, what repels you, and what begins to actually sound like you.

Over time, the people who treat AI as a serious medium for exploration will, eventually, generate their own signature. Turning the logic of productivity on its head, their work will begin in abundance, pass through inner transformation and selection, and only then arrive at authorship.

Asking better questions

AI is not killing creativity, but disaggregating a concept we had treated as one thing and thereby flattened into something thinner than it really is. Creativity is a kaleidoscope of very different human experiences and intentions. Once the coloured stones in the kaleidoscope split apart, we can stop having one giant undifferentiated argument and start asking better questions.

Where do we most value novelty?

Where do we most value abundance?

Where do we most value quality?

Where do we most value origin and authenticity?

Where do we most value personal transformation?

Where does AI reduce friction?

Where does human presence add meaning?

Those are harder questions. They are also, finally, the interesting ones.