AI can’t smell but it’s still creating your next cologne

It’s said that our sense of smell is the strongest link to our memories, so much so that one whiff can transport us back to different periods of life. Fragrance companies understand this power of scent and aim to convey abstract concepts and moods within each fragrance composition. But this is easier said than done. Professional Noses are said to have memories of more than 4,000 scents, which they’ll then call upon anywhere from 7-100 different ones to create your next perfume or cologne. It’s a tireless process filled with nearly infinite possibilities.

By no means am I a fragrance expert. Heck, I can’t even tell the difference between jasmine and lavender. However, there’s a fair amount of data science research being poured into understanding, recommending, and even creating new fragrances.

AI Gets a Sense for Scents.

It’s hard to imagine that scents can be boiled down to mathematics and numerics. Especially considering it’s such an intangible concept that most of us have trouble actually putting words to. Seriously, take a moment and try to accurately describe what a fresh lemon smells like. It’s hard, isn’t it? Yet, you know instantly when the smell of lemon enters the air.

The complexity of smells is enough to keep only a select number of professional noses in the game of creating new scents.

However, Symrise, one of the global leaders in flavor and fragrance production, teamed up with IBM to see if it was possible to use AI to create new fragrances that were pleasant and marketable. The result of their labor is an AI called Philyra, after the Greek goddess of crafts, paper, healing, and of course, perfumes.

Richard Goodwin, an IBM Research scientist, says Philyra was trained to recognize similarities among the formulas for 1.5 million existing fragrances, which includes fine fragrance (such as women’s perfume and men’s cologne), as well as lesser scents used for shampoo, laundry detergent, and candles.

These formulas were labeled with information about associated human perceptions, as well as “success factors,” such as sales or a client’s initial approval or disapproval of a novel scent. Then a series of algorithms – including deep neural networks, support vector machines, and a random forest — find correlations hidden in the data.

Philyra was able to learn which chemical substances can function as substitutes for others, which substances were good compliments, and which ones weren’t. “We’re trying to ensure that we’re generating novelty,” Goodwin says. “But creating novelty wouldn’t be enough, because there are many things that are novel that don’t smell good.”

While about 99% of the AI-generated fragrances were not good, there were several that stood out in a good way. Symrise would then treat these novel scents as it would any human-created scent: by having a panel of human experts rate them. That feedback was then fed back into the system, which is how the results got better in an iterative manner.

Alex Woodie, datanami

This wasn’t just another way for IBM to change people’s perception of AI. Symrise and their Philyra AI actually created two fragrances that were bought by O Boticário, one of the top global beauty companies, and the two perfumes are scheduled to launch in mid-2019.

To be honest, we presented variations on the AI-generated fragrances that were ‘corrected’ by perfumers, myself and another, and the AI-generated fragrance won. It was the one that was most interesting and innovative.

David Apel, Perfumer

It’s very impressive to see an AI take something creative from zero to completion. Perhaps fragrances will be an area that AI excels at. There’s clearly a monetary incentive.

Already a future thinker?
Then become a friend.

AI Makes A Scent For You.

When you tally up all the different products that fragrances are needed in – ranging from colognes and perfumes to your shampoo and lotions – the global fragrance industry stacks up to a worth of $52.7 billion, and is expected to rise to $72.3 billion by 2024.

All of this money amounts to many others data science initiatives in the fragrance space.

ScentSee, for example, is a company that created a fragrance API which is supposed to help fragrance retailers find the right fragrances for their customers. Normally, this is a process that experienced noses are trained to do. However, ScentSee believes their technology could bridge olfactory and digital.

Along a similar path, one of Estee Lauder’s artisanal perfume brands, Le Labo, is toying with the idea of using AI to create ultra-personalized perfumes, that might be unique to one individual or a small cluster. Pairing data from personal questionnaires and an AI at the level of Philyra could actually make this possible. And more importantly, affordable.

Likewise, Kelly Peng is a data scientist who created a fragrance recommendation system that is designed to help people find the fragrance that fits them best. Although the system was a research project and was live for only a short period of time, it shows the possibility of a future where fragrances are as unique as our fingerprints.