What is perfume AI, and how technology can create fragrance a human will love?

Originally, perfumes were used to seduce partners, display wealth and communicate with the Gods. While modern perfumery developed in the 19th century, it was not until the 1950s that perfume transformed into an everyday essential. Today, individuals buy and wear perfumes to express themselves, attract partners, indicate status and even evoke certain feelings and moods.

Globally, the fragrance market is worth $48 billion per Euromonitor. This market has been predicted to grow at a CAGR of 5.5% between 2020 and 2025. While the perfume industry has taken an inevitable hit during the 2020 pandemic, the market has started to re-open again. It is important now more than ever for retailers across all sectors to resist becoming too comfortable, and instead focus on innovating and creating new products that keep up to date with changing market trends. Perfume AI has been the answer to this call to innovation.

The first AI perfume

In 2019, the cosmetics company O Boticàrio became the first business to sell a fragrance developed perfume AI. O Boticàrio achieved this by working alongside IBM Research and Symrise to create the first AI apprentice, Philyria.

Firstly, Philyria operates by drawing on scents from Symrise’s database of 1.7 million scent formulas. It processes these scents alongside information about the popularity of each formula among different demographic groups. For example, this might include predicting popular scents for people living in specific countries, or from particular age groups. Lastly, Philyria applies an advanced learning algorithm to predict fragrances that would perform well for a given target group. For the purposes of this perfume, Philyria was instructed to create a fragrance that would appeal to millennials living in Brazil. They called this scent ‘Eggeo ON’.

Different variations of the scent were released, including fragrances adjusted by the esteemed perfumer David Apel. The AI scent, however, definitively gained the most popularity.

ScenTronix

In 2020, ScenTronix became another important name in the perfume Al industry. They are a Netherlands based company that set out to create personalised scented products for consumers. After stepping into the ‘Algorithmic Perfumery’ shop, consumers are asked to complete a survey about their personality and lifestyle. They then watch an AI machine create an array of scents based on their survey responses. Customers evaluate the perfume trials and make changes if they desire before they decide on the final product. The AI program learns from customer feedback and uses these lessons to fine-tune suggestions for future customers. The founder of ScenTronix, Frederik Duerinck, describes how “What we offer is, at its most simple, a platform for people to co-create their own perfume”. He envisions ScenTronix as a means for customers to explore their own imagination, and imbed the perfume of their choosing with a meaningful and memorable experience.

In the long term, ScenTronix hopes that their AI system will develop to the point of being able to create the perfect perfume for any given person in the world.

Carto – perfume AI system developed by Givaudan Fragrances

Carto is a perfume AI system developed by Givaudan Fragrances. It has been designed to assist perfumers by suggesting combinations of scents. Perfumers use a simple touchscreen to take advantage of this system, combining different scents from a vast digital scent library holding more than 1500 ingredients. A robot then processes these scents and produces a fragrance based on them that perfumers can test out before deciding on the final product. This helps perfumers by saving them time and easing their manual efforts.

Maison 21g

Maison 21g is a highly personalised perfume designed to capture the essence of the consumer’s soul. ‘21g’ refers to the estimated weight of the human soul, calculated by the physician Duncan MacDougall. It also refers to the percentage of perfume concentrate present in the fragrance. The perfume achieves this high level of personalisation with the use of an AI-powered quiz. This quiz leads the user to a fragrance that is suited to their personality, lifestyle and taste. Perfume is then produced based on this fragrance, with the use of a machine named ‘La Source’. The details of how this machine works has been kept secret and is patented by Johanna Monangue, a French perfumer.

Reactions to perfume AI

1. Machines can’t ‘feel’

Unsurprisingly, there have been mixed reactions to the infiltration of AI into the perfume industry. Many point out that scents are detected and experienced by humans, not machines. Humans therefore will always be in the best position to develop the best fragrances. This is especially as, according to scientists, our sense of smell is the most powerful sense for triggering memories, emotions and moods; a profoundly human reaction that machines cannot begin to comprehend. This highly emotive experience is why perfumery is often classed as an ‘art’. Those that wish to study the art of perfumery must train as ‘noses’ for years before they have the chance of developing fragrances and evoking emotions through scents. Therefore, while AI can understand that different scents appeal to different people, it cannot understand why. This is reflected in the mistakes that AI machines often make. Claire Viola, vice president of digital strategy fragrance at Symrise- describes how AI produced perfumes rely on “… machine-learning and sometimes the results have been wrong… It constantly needs training. You have to qualify every new material, so it understands the difference between different florals and oriental scents, for example”. Following this line of argument, there is no place in the perfume industry for unthinking and unfeeling machines.

2. Machines may not be able to feel, but they can learn how to evoke feelings

AI operates by mimicking a human-style learning of the art of perfumery. Richard Goodwin, a research scientist at IBM describes how “Just like an apprentice would learn from a master what combinations of ingredients would work well… machine learning will create a fragrance based off which formula worked best”. This suggests that perhaps we don’t need humans to implement the human process of learning. Margaux Caron, global beauty analyst for colour cosmetics and fragrance at Mintel, supports this viewpoint. She describes how “Technology and science are sometimes pictured and perceived as cold and rational, but the fragrance category is displaying a warm, emotional, human approach to it. The partnership between AI and perfumers is anchored in this philosophy.” Further, although AI often makes mistakes at the beginning, the more it is used the more accurate and helpful it becomes.

3. A partnership instead of a replacement

AI is useful for building the skeleton and main body of a fragrance,” describes Julia Zangrilli, the founder of the customer fragrance company ‘Nova’. While the real creative process should lie at the hands of the perfumer, AI is useful for large companies to identify new combinations of scents. Richard Goodwin also points out that AI can analyse formulas and data at a far greater speed than any perfumer. It should be taken advantage of for these purposes. This is especially as new scents are needed at a faster rate to keep up to date with changing market trends. AI could even help companies to use protected scents without exposing trade secrets, by tweaking formulas slightly. Claire Viola, Vice President of Digital Strategy Fragrance at Symrise describes how AI should be used as a partner to the perfumer, instead of a replacement. She describes how “It’s a man-machine collaboration…It’s helping [the perfumer] to be better faster and creative, and freeing them from boring tasks. It still starts and ends with the perfumer. They’re the ones with the intuition, emotion and feeling and guiding the machine to better results”.

In conclusion, AI has infiltrated the perfume sector in a number of ways; spurred on by global brands such as IBM and ScenTronix. AI has helped cosmetics companies to develop innovative and personalised fragrances. Additionally, it has eased the manual labour of creating perfumes and has significantly sped up the process of perfume creation. There is a diversity of opinion within the perfume industry about the appropriate role of AI in the creation of perfumes. The greatest concern is that it should not or cannot replace the role of the perfumer. Perhaps, however, AI might be better seen as a partner to the perfumer, instead of a replacement.