The Myth That AI Has No Creativity of Its Own

It is one of the more comforting things people say about artificial intelligence: sure, it can crunch numbers and write boilerplate, but it cannot truly create. Real creativity, the argument goes, belongs to humans alone, and a machine only remixes what it has already seen. It is a tidy claim, repeated everywhere from op-eds to dinner tables, and the evidence for it is shakier than most people assume.

Consider the most-cited test of the idea. A study titled "Divergent creativity in humans and large language models," published in Scientific Reports, pitted more than 100,000 people against current AI systems on the Divergent Association Task, a standard measure of creative thinking where you name words as unrelated to one another as you can. Researchers from the University of Montreal, the University of Toronto, and Mila ran the comparison. At its most adventurous setting, GPT-4 scored higher than 72 percent of the human participants.

Beating roughly seven in ten people on a recognised creativity test is not the result you would expect if machines were simply incapable of original thought.

The catch the boosters skip

That is not the whole story, and here the opposite myth, that AI is now more creative than we are, falls apart just as quickly. The same research found that the most creative humans, especially the top tenth, still left the models well behind. And on richer creative work, the kind that actually moves people, the gap widened. When the task shifted to poetry, fiction, and other open-ended writing, human-made pieces were judged consistently more creative than the AI versions.

So the honest reading is double-edged. The blanket claim that AI has no creativity does not survive contact with the data, because these systems can produce genuinely novel, surprising combinations and beat the average person at it. But the leap to "AI out-creates humans" is just as wrong. What current models are good at is fluent, statistically plausible novelty. What they still struggle with is the creativity that carries intention, taste, and meaning, the reason a particular poem lands while a competent imitation of it does not.

Why the myth persists

Part of the appeal is reassurance. Creativity has long been the line people draw to keep something uniquely human on the far side of automation, and it is unsettling to watch a model wander into territory we reserved for ourselves. Part of it is a definition problem. The word covers everything from a clever word association to a symphony, and AI's performance varies wildly across that range, which lets people cite whichever end proves their point.

The grounded view is less dramatic than either slogan. Today's AI is creative in narrow, testable ways and unremarkable in the deeper ones we tend to care about most. Treat anyone who tells you it has no creativity at all, or that it has already surpassed us, with the same polite skepticism. The interesting action, as usual, sits in the middle.