From Siri on our iPhones to even the junk mailbox (yes, it helps to filter out the spam mail we receive) in our e-mail folders, artificial intelligence plays a huge part in our daily lives – even if we’re unaware of it. And now, it’s making its way into the way our skincare is made. Cue: Dior’s refreshed Capture Totale C.E.L.L Energy line.
Slated to hit stores come Jan 1, this five-piece beauty regime consists of a Super Potent Serum, a Firming & Wrinkle-Correcting Creme, a Firming & Wrinkle-Correcting Eye Cream, a High-Performance Treatment Serum-Lotion and a High-Performance Gentle Cleanser. The reformulated range focuses on skin stem cell technology. Skin stem cells are what produces new skin cells, which helps keep skin looking youthful and healthy. Harvesting the prowess of a potent blend of four floral extracts – namely madagascan longoza, Chinese peony, Chinese Jasmine and white lily – the products re-energise skin stem cells to produce fresh ones. Read: your skin will look youthful and taut.
We sit down in an exclusive interview with the French beauty giant’s Executive Vice‐President of Product, Michael Pressigout, and its Scientific Director, Edouard Mauvais Jarvis, where they break down how this new line taps in on the power of Deep Learning (a subset of artificial intelligence that Dior uses) that powers the brand’s upcoming age-defying skincare line.
Can you explain how Deep Learning works for us in a few sentences?
Michael Pressigout (MP): “The term dates back to 2012, a time when artificial intelligence was
making enormous progress, both theoretical and also linked to the progress seen in
computers. We realised that a machine could benefit from automatic performance learning,
to recognise shapes for example: if you show it a lot of images of cats and dogs, it will
eventually be able to identify a dog and a cat. In fact, the software is a bit like the human
brain, with different levels of neurons. Deep Learning uses a type of neurons network and
adapts the connections between them in order to improve shapes recognition.”
What are the classical uses for this technology? Which types of organisations and sectors does Owkin (the AI startup) usually work with?
MP: “Anything that concerns sound, image and text analysis, for example, iPhone’s Siri, or Google’s Alexa. On social media, Facebook and Google use Deep Learning to identify your face; chatbots decipher words to understand sentences; automatic cars also use it in order to distinguish a pedestrian from a car or an animal. We work especially in the field of health, with hospitals, for medical imagery, looking at how you can predict a patient’s response to a treatment from looking at an image. Moreover, Owkin was created in 2016 by a doctor and a researcher in Artificial Intelligence. We’re based in France (Paris, Nantes) and in the US.”
How did Dior and Owkin come together?
Edouard Mauvais‐Jarvis (EMJ): “We have been working on neuroscience and the criteria for
perceiving beauty in particular, for Capture Totale. We were looking for another, different
way to approach the subject, and thought why not look at the latest technological methods
available? Today, the only way of measuring these criteria is to sit 30 people in front of a
face and to record their impressions – which we did, for many years. With Deep Learning,
we are taking it up a notch, because after all, behind of the subjective nature of perception,
a logic exists, that we don’t know how to explain.”
Do you mean that thanks to artificial intelligence we are capable of objectivising beauty?
EMJ: “Not beauty, no, which is linked to judgement and emotion. However when we look at someone, we immediately know if their face looks healthy or not – and that perception goes far beyond lines, pigmentation spots etc. In fact, our general impression of a face is linked to a database that has been recorded in our brains throughout our lives, and during our childhoods in particular.”
MP: “That’s why it is so difficult to guess how old someone from a different culture is! Man is a social animal, and the same goes for the software ‐ if you only show it Asian or Western faces, it won’t be able to evaluate other cultures and ethnicities.”
In concrete terms, how did you set about it?
EMJ: “We started by showing about twenty people some French, Chinese and Japanese women’s faces. We then asked them to rank, on a scale of 1 to 5, the fundamental biological parameters, identified as health, radiance, vitality, energy, strength, tone, attractiveness, and we asked them to evaluate age too. After calculating an average of the data obtained from the panellists, each face contained several pieces of information: real age, perceived age, and 7 marks (for the 7 criteria).”
MP: “This data served as a base for the software, which we then “trained” with almost 600 photos. Indeed, generally, the more data you have, the better the results. Information quality, a recurring subject in Deep Learning, is also essential, and in this case, the photos of faces supplied by Dior were excellent quality, taken in different lighting, from different angles, without any variation in the quality.”
EMJ: “That is also a fundamental piece of data in cosmetic research. All our photos were taken using the Dior Skin Scanner, our skin diagnostic tool that stems from medical imagery. You measured faces before and after applying the Super Potent Serum Capture Totale C.E.L.L Energy.”
Did the results surpass your expectations?
EMJ: “Technology doesn’t mean that we can do away with the human evaluation system, but it enables us to confirm and stabilize it over time. And we were in fact particularly pleased with the effects of our new serum!”
MP: “As for Owkin, we were very impressed. First of all, you realise that the perception criteria for faces are not random and that there is an existing consensus that the computer manages to analyse and reproduce. Next, the before/after results were quite astonishing. This serum claims to “create new cellular matter and correct all visible signs of the passage of time.” It made me think of emerging technologies stemming from facial recognition that generate matter, new images such as Faceapp or style transfers (Van Gogh‐style photos).”
In the future, can we expect to see other innovations in the cosmetic industry thanks to Deep Learning?
MP: “Definitely. I’m thinking in particular of the construction of new active molecules and product design based on these ingredients. Also, the product evaluation system – anticipating performance and tolerance. It won’t just be about processing images, but also clinical data. Options are opening up.”
EMJ: “Today, we systematically evaluate hydration with numerous measurements, but what if visible data existed, indecipherable by mankind, but that a machine could identify? We’re moving further away from beauty, but just look at profiling. In the United States, the secret services are already using Deep Learning to study criminals. What did the victims have in common? It takes hours for the human brain, and sometime luck, to make the connection. A machine can find it instantly.”