Is There A Formula To Fashion?
With the rise of Artificial Intelligent stylists that can analyse and recommend fits, can we now assume that robots have acquired a human taste in fashion? Jacinth Chia and Nicole Fong investigate.
In the opening scenes of 1995 cult classic chick flick “Clueless”, Cher Horowitz, played by Alicia Silverstone, flicks through various outfits on her computer. “MIS-MATCH”, the screen would read if she ever picked a bad combination. Eventually, with the help of the app, she picks out her iconic yellow plaid suit.
Fast forward to London’s Fall/Winter Fashion Week 2019, where London-based Artificial Intelligence (A.I.) styling company, Intelistyle, pit their virtual stylist against human ones in an experiment to see if people could tell which outfit was created by a human. Not only were the differences virtually undetectable, 70% of respondents indicated that they preferred the outfits created by the machine.
It seems like A.I stylists, who were once nothing but a movie gimmick are becoming a reality, not just for runway fashion, but for the masses too.
Companies around the world are also making use of A.I. technology to help recommend outfits. In Kuala Lumpur, Malaysia, fashion styling business Ombré has A.I. that uses details about shoppers, such as age, skin tone and body shape, to match an outfit to them. According to Fildzah Zulkilfi, chief marketing officer, the A.I. also takes into account the fit and look of the garments, such as the fit, sleeve length and colour, so that the recommended combination will look good on the shopper - much like what a human stylist does.
That said, human stylists might not accept robots like these into their ranks so quickly.
“I feel the point of a stylist is, it's not only just the clothes, it’s about a form of expression. It’s reading a person and knowing exactly what fits best for them and how they can be elevated. And a lot of this is based on your intuition.” says stylist Mae Tan, 26. “I would like to think that computers don't have intuition.”
“The whole process is like a machine learning from a fashion expert.”
-
Yuwen Xu, product lead for Visenze
And she’s right - experts predict that we still have 40 years to go before we get close to seeing A.I. that can pass off as a human, let alone one that has the ability of self-expression. But while fashionable sentient robots might be far out in the future, the A.I. stylists of today make use of data to recommend great fits and tap on patterns in fashion.
“The whole process is like a machine learning from a fashion expert,” says Yuwen Xu, product lead for visual A.I. company Visenze. The Singapore-based company creates A.I. stylists for stores like Uniqlo and Zalora.
The A.I. learns from an in-house stylist how to read an image of various apparel, identifying the type of garment, colour and neckline, alongside 24 other categories of attributes. The stylist also teaches A.I. some basic fashion rules, so that shoppers aren’t offered a combination that’s a walking fashion disaster. Put these all together, and their A.I. stylist can process millions of products in seconds, according to Xu.
Apart from making recommendations with advice from a human stylist, A.I. can also trawl the Internet for information on trends that can help with styling.
“What we do is we train the AI in crawling the web,” says Ombré’s Zulkilfi. “From articles, to images, to videos, to sizing charts, to live streams of Fashion Week events. Anything and everything on the internet that's related to styling, our A.I. will capture this data. That's where we try to generate new styling parts.”
“There are some hidden patterns in the data that the machine sees, and the machine tries to simulate how that data is generated.”
-
Yuwen Xu, product lead for Visenze
Another way A.I. recommends styles are to tap on fashion’s past. Not only does it provide good training data, but it also allows the A.I. to predict what’s about to make a comeback. As the saying goes, fashion repeats itself every twenty years.
It’s not just empty words. According to LASALLE fashion lecturer and researcher Daniela Monasterios-Tan, fashion designers tend to reference their designs to styles or things they grew up with.
To illustrate, referring to the current trend of the Y2K aesthetic (fashion from the 1990s to early 2000s), she explains how styles from that decade would have been referenced from fashion in the past. “So, in 1999, they would have been referencing the 70s. And in the 70s, they referenced the 50s. And it’s also because the people who design clothing tend to look back into the past for inspiration,” she says.
“In the 90s, John Galliano was doing bias-cut dresses,” she adds, “and these bias-cut dresses were very popular in the 30s. He might have seen them in the 50s, or 60s, for example, from his mom, or something like that. So there's always this language in fashion, where there's quite a few decades being referenced at the same time.”
For the A.I. stylists, this means that they can not only learn styling from past fashion, but also attempt to predict what trends will be next. “There are some hidden patterns in the data that the machine sees,” says Xu, “and the machine tries to simulate how that data is generated. So over the years, it can simulate a very complicated, very detailed formula.”
But does recommending styles based on past fashion decisions certify the A.I. to be a stylist? According to Tan, the role of a stylist goes beyond just suggesting outfits to wear.
“At the end of the day, I feel like it's easy to give someone clothes to wear or give someone a direction and tell them ‘this is what you should do.’ But it's so different when there's a human personal touch, you see,” she says.
Admittedly, churning out styling recommendations with information from the Internet doesn’t take into account the shopper’s own personality. At the very most, these A.I. technologies can make reference to a shopper’s purchase history, but that doesn’t come close to the authenticity that a human stylist can provide.
“It’s important that it reflects their personality also through styling,” Tan says. “I believe in authenticity, and I believe that the planning always has to reflect their personality and how they're feeling.”
At the end of the day, A.I. simply learns from millions of people's preferences and their behaviors, and recommends style based on that, says Xu.
Other than the human touch, it’s also impossible for A.I. to always predict what’s next for fashion, such as in the case of avant-garde designers - designers who design not based on trends, but on a theme that they come up with, says Monasterios-Tan. “For example, someone like Martin Margiela who starts creating things that have no reference to trends or to data,” she says, referring to the founder of French fashion house Maison Margiela who is known for revolting against luxury fashion in the 80s with “ugly” garments.
-
Daniela Monasterios-Tan, LASALLE Fashion Lecturer & Researcher
“The purpose sometimes of certain brands is just to express themselves, and it's the same with style.”
But while A.I. stylists cannot possibly predict the future of fashion, humans cannot either.
And there’s nothing wrong with that, says Monasterios-Tan. “The purpose sometimes of certain brands is just to express themselves, and it's the same with style. People will always wear clothes. Some people wear it to express themselves, some people wear it to express how they are not like others, some people wear them to fit in.”
“Because even though you have mass data, and you might be catering to the mass taste, there will always be a group of people who refuse to follow the masses.”