Technology
How to stay creative in the AI Era
Production costs almost nothing anymore. What stays precious is choosing: the hand, the eye, taste. Nine habits to keep them alive, while a machine takes care of the rest.
For the first time in history, producing an image, a page, a text or a piece of code costs next to nothing: an idea sketched at nine becomes a prototype at half past nine. Anyone who works in front of a computer feels it every day: the manufacturing of the product, which once took weeks, months, has been compressed into minutes, hours. It is good news, and an uncomfortable question at the same time. If making costs nothing anymore, what is still precious?
The short answer: everything that came before and after the manufacturing. Understanding what to make, deciding how, recognising when it is right. Creativity was never the ability to produce, but the ability to choose. AI has not made it useless: it has laid it bare. This piece collects nine habits, borrowed from people who have reflected on the question, for remaining “people who choose”.
Creating is choosing
Ted Chiang, who writes science fiction and reasons about machines better than many engineers, put the matter in a sentence: a work is the result of thousands of choices, a prompt contains a few dozen. When you ask a model for an image with fifty words, it makes all the other decisions for you, with the average of everything it has seen. The result is competent, with nobody behind it. Staying creative means taking the decisions back: the palette, the crop, the rhythm of the sentences, the white around things. Not all of them, but the ones that count.
Taste becomes the job
Ira Glass, the voice of This American Life, once described the gap that torments beginners: you have enough taste to know that what you make is not good yet, and not yet the hand to close the distance. For decades the hand was paid for dearly, in hours. Today the hand can be rented, but what cannot be rented is taste.
A study published in Science Advances in 2024, by Anil Doshi and Oliver Hauser, measured what happens when short-story writers receive ideas from a model. The individual stories get better, especially for those who started further back; but the stories, taken together, resemble each other more. The average rises and the variety falls. It is the most precise description of the era we are in: the average goes up and flattens at the same moment. Being recognisable is worth more than it was yesterday.
Keep a practice by hand
Not out of nostalgia. The body learns things the head does not: how much a line weighs, when a pot is about to give on the wheel, how much ink the paper will take. People who work on screens need a practice that cannot be undone with a key. A sketchbook, a roll of film, a lump of clay, a line of type set by hand: one thing a week you can genuinely get wrong.
Rick Rubin, in his book on the creative act, insists on a point that seems obvious and is not: the artist works in order to notice, not in order to produce. Manual practice is training in noticing. Then you go back to the machine with different eyes.
Feed yourself slowly
A model has read everything and lived nothing. Anyone who creates has to do the opposite: read less and live more, and live attentively. Whole books instead of summaries, exhibitions instead of scrolling, a wall of torn posters in an unknown city instead of a moodboard of images already filtered by someone else. Austin Kleon, in Steal Like an Artist, says it plainly: you are the sum of your influences, so choose them with care.
And walk. A Stanford study from 2014, by Marily Oppezzo and Daniel Schwartz, measured that walking increases the production of original ideas by sixty per cent on average compared with sitting, and that the effect lasts a little after sitting down again. You do not need a park. You only need to go out.
Know the history of the craft
A model has seen everything and does not know where anything comes from: to it, Swiss typography, a poster by Fortunato Depero and a template from 2019 are the same material, blended into the same average. Whoever knows the history of their craft has an advantage that cannot be bought: they know which line a choice comes from, and can choose a line. Reading Müller-Brockmann before setting up a grid, looking at how Vignelli put a subway in order, knowing what the web was like in 2003, when pages were slow and personal: it is not erudition, it is what lets you recognise an old idea dressed up as new, and pick a genuinely old one, with your own hands.
The simplest way is to keep a small library, on paper, and use it: ten craft books read twice beat a thousand bookmarks never reopened.
Defend boredom
Boredom is the moment when the mind, having nothing to do, starts doing things on its own. Sandi Mann and Rebekah Cadman, in a 2014 study, had a group of people copy numbers out of a phone book and then put them in front of a creative task: those who had been bored produced more ideas than those who had not. The problem is not that AI takes away our creative work. It is that, together with the phone, it takes away the waiting: the queue, the train, the ten minutes before a meeting. That is where ideas used to be born.
Defending boredom is a practical decision: a walk without headphones, an afternoon without screens, a sofa and a ceiling.
Work with the machine, not inside it
Ethan Mollick, who teaches at Wharton and has studied how models are used in real work, describes a “jagged frontier”: there are tasks at which AI is surprisingly good and tasks, apparently similar, at which it is terrible, and the border cannot be seen by eye. In a 2023 experiment with hundreds of consultants, those using GPT-4 worked better and faster on tasks inside the frontier, and worse on those outside it, because they trusted it. The rule that comes out is simple: use the machine as a sparring partner, not as an oracle. Ask it to criticise, to list alternatives, to find the hole; not to decide.
And write the brief before the prompt. If you cannot explain to a person what you want and why, you will not be able to explain it to a model, and it will fill the gap with the average.
Three questions that work, in this order: what is missing here; what are three alternatives genuinely different from each other, not three variants of the same one; and what would the harshest critic you know say about this work. The answers are the material to decide on.
Make things for nobody
The last habit is the most fragile. Make something with no client, no metrics, no audience: a website for a friend, a typeface you will never sell, a series of photographs of your own street. That is where you try without fear, fail at no cost, and find out what you really like when nobody is watching. And finish it: the difference between having taste and having a style is a certain number of things carried all the way to the end.
What remains
The hand remains, which knows when something is right before the head does. The eye remains, trained on real, slow things. Taste remains, which is the sum of what you chose to look at. And the choice remains. The machine produces, but choosing is still a craft, and for a while at least, it will be the real craft.
Photo credits: all images are under the Unsplash License. Photographers: Antonio Francisco (cover), Abuzar Xheikh, Goashape Studio, Bundo Kim, Julia Kutsenko, Yanhao Fang, ritesh singh, Courtney Cook, Declan Sun, Manolo Chrétien, NordWood Themes.
