“I view machines as collaborators and push the boundaries of what is possible by utilising data in a poetic way,” explains media artist Refik Anadol at Art Dubai 2023, where he showed his installation ‘Glacier Dreams’ for the first time.

The artwork is based on an enormous dataset of glacier images from around the world, encompassing visuals, mesmerising audio and even an AI-generated scent. The project is part of Julius Baer’s ‘NEXT’ initiative, which is designed to encourage interdisciplinary exploration of megatrends across the arts, science and technology.

Disrupting the artworld and fragrance market

As a pioneer in his field, and the first to use artificial intelligence in a public artwork, Anadol creates art at the intersection of humans and machines. But not only the artworld is being disrupted by AI.

Today’s fragrance market is worth over USD 50 billion and is growing at an average of 5.9% year on year. Scentmate, an innovative, AI-driven fragrance house launched by Swiss perfume manufacturer dsm-firmenich, enables clients to transform their ideas into a signature fragrance not only by pressing, steaming and blending the right ingredients but by mining, crunching and distilling data.

It’s a sweet-smelling demonstration of the theme underpinning the ‘NEXT’ initiative. “Like Refik Anadol, we operate at the intersection of innovation, digitisation and art,” explains Alfonso Alvarez-Prieto, General Manager & Founder of Scentmate.

“After all, perfumery has a touch of art, too. Just as Refik collaborates with a computerised mind to create visualisations of our digitised memories, our perfumers work hand in hand with the algorithm to create scents based on our olfactory experiences.”

Scentmate users are guided step-by-step through an interactive platform to arrive at the distinct fragrance they wish for their brand. Although Refik spoke directly to the computer, the process was largely the same. “Refik entered his vision for the series directly into the computer,” says Alfonso. “The machine played around with the concepts and came up with different formulas, which our perfumers then tested and tweaked to ensure the scents matched the brief. The chosen fragrance was the one he felt evoked his creation most strongly. As a company, we’re very proud of it – it’s a highly sophisticated scent that wearers will feel comfortable wearing,” says Alvarez.

Translating emotions into programming language

Fragrance, the ‘poetry of the Gods’, has always been a subjective experience representing personal tastes and hidden desires. Descriptions of scents like ‘a hint of this’ or ‘a tinge of that’ might stir the imagination but they don’t convey the kind of binary input that computers are used to working with. How can language that remains ambiguous, vague and ‘just out of reach’ for humans – terms like glacial, frozen, fragility – be used to communicate distinct notes within the grasp of an algorithm?

That’s where the beauty of big data comes in. Firmenich has over 50 years of proprietary data about consumer preferences, collected in the many surveys it has conducted over the years as well as the feedback on its samples. “This rich data base, combined with the possibilities offered by AI, means the company can predict with a high degree of accuracy which fragrance someone of a particular age, location and gender will prefer based on the concepts they express,” explains Alfonso. “We have data that indicates what scent would appeal more, say, to a middle-aged woman in Sweden or a teenage boy in the US.”

The company is also aided in its efforts to pick up the scent of consumer preferences by the proprietary technology it has developed to map user’s emotional response to specific patented ingredients – based on magnetic resonance images. “We can tell which emotion, for example happiness, a specific fragrance will elicit by mapping millions of volunteers’ brains across different countries and demographics. By using machine learning to examine data in huge quantities like this, we catch an accurate whiff of the emotions people want to elicit when they describe a scent in a certain way.”

Crunching the numbers for greater customisation and sustainability

Alfonso and his colleagues realised that in a market where fragrance is the biggest driver of repurchase for scented products, they needed to find a way to create a distinct connection with their customers. By exploiting the efficiencies of AI, they sought to extend the level of services and customisation available. “Our aim was to offer top-notch expertise, craftmanship, ingredients and security for entrepreneurs and less well-established brands,” he says. “This meant, for example, that we could go to a company that produces lemon-scented products – just because that’s the fragrance all their competitors use – and tell them for the first time that they could have a much more sophisticated and distinct scent for their product at an affordable price.”

In addition to enriching the customisability of the scents, artificial intelligence is also boosting the fragrance house’s sustainability credentials. “To get one kilogramme of sandalwood oil for use in a fragrance, you’d need to cut down thousands of trees,” says Alfonso. “The result might be natural, but it’s hardly sustainable. AI helped us recreate the same sandalwood molecule using sugar cane and bacteria. The end product is 100% natural and not a single tree was cut down. Even better, it doesn’t contain any of the allergens in natural sandalwood so it’s tolerated by more wearers.”

Evoking emotions through the use of AI

With literally billions of combinations of ingredients possible, each yielding a different note, this time-honoured industry seems custom-made to benefit from the opportunities presented by machine learning. However, Alfonso is confident that the use of AI will not dilute the importance of human perfumers. “Perfumery is a highly skilled craft, involving 15-20 years of training. Experienced perfumers can predict what a fragrance is going to smell like simply by glancing at the combination of ingredients,” says Alfonso. “AI can help us create efficiencies of scale and handle some of the complexities of production, but the success of our products depends on the emotions and sensations they evoke. That’s something which will always call for the human touch.”