Creating perfumes is often considered an art exclusive to humans and a gift that requires years of practice. Now, this task of evoking olfactory senses is now being done by artificial intelligence. Traditionally, professional scent masters or ‘noses’ spend decades learning the craft, demanding hefty pay cheques from companies for their meticulous work — but all this may change as AI intervenes.
In a recent development, Symrise, a global player in flavours and fragrance which boasts clients like Estée Lauder, has collaborated with IBM to test how machine learning and AI can be applied in the world of fragrance. This olfactory world includes detergents, air fresheners, candles and soaps among others.
Now, the AI called Philyra will help the company in augmenting the process of fragrance-making, giving a technological twist to the traditional method. The company has reportedly developed two AI-based perfumes already, which have been sold to O Boticário and will be launched in 2019 across 4,000 locations.
The Science Behind Philyra
The ML model can quite interestingly understand sensual, transformative and personal components of making perfume with ease. IBM researchers, led by Richard Goodwin, have developed an AI that could learn these intricate details from professionals like chemical formulas, raw materials or industry trends, among others.
Philyra has been built on IBM’s previous research about using AI for recipe creation and a new AI for product composition. The advanced ML algorithm can study thousands of existing formulas for fragrance and compare the ingredients to other data sets such as geography and customer age, helping identify patterns and novel combination.
To explain in detail Symrise shared a list of fragrance formulas, data on sales performance, etc. which was compared to additional customer data. It included data such as geographies served, best seller products, countries where it performed well, age demography, among others. Based on these data the new ML algorithm was made that could be used to create new formulas targeted to a specific audience.
“Creating a fine fragrance is an art as well as a science that requires precision, as even the smallest change in the amount of a material can make or break a new perfume. As a result, our team took a data-driven approach, relying on data about hundreds of thousands of fragrance-formulas, fragrance families (e.g. fruity or floral), a fragrance’s raw materials and historical information that captures the success of previously designed perfumes and formulas, among other data points”, Goodwin explains in a blog post.
How Is The AI-Created Perfume Unique?
With the help of ML, the new algorithm could predict alternative raw material or substitutes to be used, human response, the novelty of fragrance as well as the appropriate dosing of raw material, among other components. However, to bring uniqueness in fragrance is a key challenge in the perfume industry. “Philyra learns a distance model to identify fragrances that are close in smell to existing fragrances. The larger the distance between a fragrance and its neighbours, the more novel the perfume is predicted to be,” Goodwin explained.
It also helps in understanding consumer preferences which in turn helps the company to focus on perfecting the final product rather than spending time searching for new fragrance combinations.
In the coming days, Symrise plans to introduce the technology to their master perfumes across the globe and also design fragrances for personal care and home care products. Symrise also plans to introduce Philyra into their Perfumery School to help train the next generation of perfumers, firmly embedding AI into the heart of its organisation.
Before this, a team of computer scientists had unveiled a set of algorithms to predict the odour of different molecules based on their chemical structure. They labelled the smell with more than 19 descriptors, including “fish,” “garlic,” “sweet,” or “burnt.” They also created a massive database based on pleasantness and intensity of odour. While the immediate use of these programs was not sure, it can now find a way into the fragrance or perfume industry.
Can AI Create Better Perfumes Than Humans?
The results have been quite exciting so far, but whether AI can completely replace humans, is a question that still requires time to answer. The aim right now is augmenting human expertise and demonstrating how AI can assist in domains where creativity is the key. As the researchers also believe, the science of designing perfume is quite old and for AI to explore all aspects of it might take a lot of time. There are a lot of formulas and historical data to identify patterns, AI has a lot of perfection to achieve before it could surpass humans.