In a storefront on Ginnekenstraat, a pedestrian street in the Dutch city of Breda, you can answer a short questionnaire about yourself and leave less than an hour later with a perfume that wasn’t there when you arrived.
The questions are not the ones the sales assistant asks. Which color represents you best. Where would you go right now, if you could go anywhere. How would you describe your style.
You answer, a set of algorithms reads your answers, and a machine in the room names a scent to match, the bottle filled and labeled while you wait.
The company that created this room is called Scentronix, and for the better part of a decade they’ve been saying that the way the world buys perfumes is more unusual than we’d like to admit.
💜 for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
Its founders, Dutch artist and filmmaker Frederik Duerinck and fragrance designer Anahita Mekanik, like to pose this unusual question: why should 800 people decide how 8 billion people smell?
They refer to a small group of professional perfumers, noses, who create almost every perfume on almost every shelf.
It is offensive, and like the best offensive it has a real idea within it. For most of its history, perfume has been a closed, elegant and distant art. The software silently informs you that it is open.
That kind of sentence often makes people panic, because we’ve been trained to expect the worst when code comes from work built for human hands.
Fear is often some version of change: the algorithm arrives, the artist is shown the door. In makeup, that’s not what happens, and when you look closely at who created these tools, that’s where another charge appears.
Big names in business have come to the same conclusion over the years. In 2019, German perfume house Symrise paired its perfumers with an artificial intelligence system it built alongside IBM Research and named Philyra, a name taken from Greek mythology.
Philyra was trained in a large archive of formulas and performance data, and could suggest pairings that no one else could access, without the burden of habits or taste.
Working alongside it, Symrise perfumer David Apel created two fragrances for Brazilian brand O Boticário, released as the Egeo line in time for the country’s Valentine’s Day.
By many accounts, it was the first AI-made perfume to be sold anywhere.
Others followed with their machines. Givaudan, the world’s largest fragrance house, has developed Carto, a touchscreen system that places the formula as a visual map and feeds the robot that mixes the visual sample in seconds, so that the perfumer can test the idea as quickly as possible.
Calice Becker, who created J’adore for Dior and runs the perfume school at Givaudan, said that the purpose of this tool is to allow perfumers to dare, to try a combination that would not be an obvious choice.
Firmenich, now part of DSM-Firmenich, aimed on the other hand with Scentmate, a service designed to help small companies and sole entrepreneurs, people who do not have a laboratory and do not have an internal nose, to create a scent at all.
Not everyone is interested, and objections should be taken seriously. Jean-Claude Ellena, former in-house perfumer at Hermès and one of the most popular noses, argued that a machine cannot read the thoughts that guide the perfumer in his creation.
He said, sadly, he feels sorry for the young perfumer who will one day be given the outline of a machine and be asked to perfect it.
Coming from someone who handles perfume as a literary medium, objections do come. There is a real risk that automation makes manual work a workflow, so that irregular, intuitive leaps are developed.
But anxiety takes competition, man against machine, and that’s not what these tools are for. Each of them keeps a perfume in the room.
Symrise calls Philyra a student, not a substitute, and it seems to mean it. Carto puts the formula on the screen, and the person still decides what is good.
Even Scentronix, a highly automated company, refers about one in 50 customers to a human perfumer to fix anything the algorithm has misjudged. The software expands the canvas. It does not sign the painting.
Underneath the trade, there is something really new being created, and it’s an area that should interest anyone who cares about technology like perfume.
Smell is a feeling that always resists the machine. We taught computers to see and hear decades ago, but smell, a tangle of molecules binding receptors in ways we still only partially understand, remained as stubborn as an analogue.
That is changing. Google researchers trained neural networks to predict how a molecule would smell from its structure alone, the first rough sketch of a machine nose.
A European project called Odeuropa used AI to find the lost perfumes of historical Europe from hundreds of years of text. Perfume is the more commercial end of a much larger effort to give software an experience it’s never had before.
The market in which all this resides is large and quietly conserving. Industry estimates put global fragrance sales at around $60bn a year, a business still shaped by a few houses, celebrity licensing deals, and those few hundred noses that decide what the rest of us wear.
Set against that, a system that allows a teenager in a pop-up shop, or a small brand without laboratory capital, to make something that smells like them and only them is not dangerous.
It is an upbringing. The pie doesn’t shrink when more people are allowed to bake.
Which brings us back to the shop in Breda. The machine doesn’t know how your perfume should smell.
It only knows what you’ve told it, and it’s faithful to a fault, which is why a person is kept around for times when you and the algorithm disagree.
What it offers is not a decision but an invitation, an opportunity to treat the oldest and most intimate senses as something you invent rather than pick off the shelf.
You log in as a customer. You come out, an hour later, with a small bottle that smells like the answer to the question that was only asked, and that was nowhere in the world when you woke up in the morning.
