The Machine That Learned to Smell

Of all the senses, smell was the last to surrender to the machine. We taught computers to read with language, to see with vision, to hear and to speak — and for decades scent held out, too ancient and too animal to be reduced to code. It lives in the oldest quarter of the brain, wired straight to memory and desire, bypassing reason entirely. To digitize it seemed almost impertinent. And then, quietly, it happened.

The house at the centre of this shift is Osmo, founded by Alex Wiltschko, a neuroscientist who left Google's research labs convinced that smell could be mapped the way colour once was. His team built what they call a Principal Odor Map — a model that studies the shape of a molecule and predicts, with unsettling accuracy, how it will smell to a human nose. Feed it a structure it has never encountered, and it will tell you: this one is jasmine, this one is diesel, this one is the green snap of a cucumber. For the first time, a machine could read the language of scent rather than merely catalogue it.

Reading, though, was only the beginning. The more astonishing feat is what Osmo calls scent teleportation. A device inhales the air around a peach, a rose, a rain-soaked street, and a spectrometer takes it apart molecule by molecule into a digital encoding — a kind of recipe written in data. That file can travel anywhere on earth, and at the other end a second machine reassembles the smell from a palette of raw materials. In 2024 the team announced its first complete success: a fresh summer plum, captured in one room and rebuilt, faithfully, in another. Fragrance had learned to move at the speed of information.

For perfumery, the implications ripple outward in every direction. Osmo has already opened a creative arm — a fragrance house built to place these tools in the hands of brands — offering AI-assisted formulation that can propose accords, hunt for novel molecules, and sketch a composition in an afternoon that once took a perfumer months. More quietly, the same intelligence promises to design entirely new synthetic notes: replacements for the sandalwoods and musks and orchids that the natural world can no longer supply in good conscience. The machine, it turns out, may become an unlikely ally of sustainability.

And yet, standing in front of the bottle on the dresser, the connoisseur feels a familiar unease. If a formula can be written by a map, is it still a work of art? If a smell can be printed, is it still precious? These are not idle worries. They are the same questions photography once asked of painting, and the synthesizer of the orchestra — and, as ever, the answer is less a verdict than a redrawing of the line.

Here is the reassurance the map cannot take from us. A model can tell you, flawlessly, what a molecule smells like. It cannot tell you what it means. It does not know that a certain powdery iris is your grandmother's dressing table, that a thread of orange blossom is a wedding you were too young to remember, that the reason you reach for smoke and leather on grey mornings has nothing to do with chemistry. Scent is not information; it is the feeling that information triggers. The machine has learned to read the notes. It has not learned to be moved by the music.

The great perfumers understand this instinctively, which is why the best of them are not frightened. To them, artificial intelligence is simply the newest instrument in a very old atelier — no more threatening than the gas chromatograph that let their predecessors see, for the first time, exactly what a natural essence was made of. The GC-MS did not replace the nose; it gave the nose a new pair of eyes. The map may do the same: a tireless collaborator that proposes a thousand paths, leaving the human to choose the one that carries meaning.

So let the machines learn to smell. Let them teleport the plum and map the rose and dream up woods that no forest has to die for. The bottle on your table will still be a decision — yours — about who you are today and to whom you will lean close tonight. The intelligence can compose the note. Only you can decide it is beautiful.