AI is making the fashion industry more intelligent — and expressions of style less individualistic
No amount of AI can create a Frida Kahlo or Lady Gaga; fashion delights us with its element of surprise, or when it cracks open a space for people’s own quirks to shine. An algorithm can do many things, but it can’t be you.
Artificial intelligence is changing the fashion industry in many ways.
AI innovations will serve mass retail and mainstream looks, so we have a whole population of stylish people.
But is the point of engaging in fashion to look like a more authentic version of yourself or to look dapper — but like everyone else?
'Curious Fashion' is a monthly column by feminist researcher, writer and activist Manjima Bhattacharjya. Read more from the series here.
Once a year I have my annual AI conversation when my research scientist brother comes to visit. Words like machine learning and algorithms enter my world, as I learn of the many amazing and alarming ways that AI is being applied around us. That’s when I first heard of Stitch Fix from my sister-in-law. If like me you’ve apparently been living under a rock and don’t know what that is, Stitch Fix is a US-based online personal styling service.
Every month Stitch Fix will send you a box (a “fix”) with up to five clothing or accessory items based on the information you’ve provided about yourself. An algorithm, along with a human stylist, comes up with a personalised style for you and picks out what to send. You keep (buy) items you like and send back what you don’t.
As the algorithm learns your preferences every month based on what you keep or return (+ scouring your Pinterest, Instagram for your aesthetics + the choices of other clients profiled as being just like you), it becomes more intelligent — and the contents of the box become more and more attuned to your taste. With 2.7 million clients and 3,000 stylists working for them hand-in-glove with their algorithms, the brand claims to have helped people “from plus-sized to pregnant” discover their sense of style.
The pictures of the Stitch Fix warehouse look like a different kind of factory floor. Hundreds of stylists standing one behind the other on the shop floor in front of their tables, folding and packing clothes. Getting “fixes” out the door.
There are several other similar box services in the US — Le Tote, Wantable, or Trunk Club from the Nordstrom departmental store, but Stitch Fix has become the poster girl of the confluence of fashion and AI. Partly this is because of its financial success — the kind of start-up story that Silicon Valley loves. Co-founded and led by a millennial woman, going from modest beginnings in 2011 to being a $2 billion fashion start up, helped along by millions of dollars in funding from VCs, finally going public in 2017.
And in part, it is their loud and proud use of AI that makes them seem cutting edge. Its algorithms can track 30 trillion possible combinations of fashion attributes (sleeves, necklines, styles, silhouettes) to come up with nine recommendations for a client. Data science is “woven into the fabric of Stitch Fix”. They even have a Chief Algorithms Officer, albeit a human one.
Artificial intelligence is changing the fashion industry in more ways than this. From manufacturing processes to planning for retail, to projecting sales or managing inventory, it is applied across the sector. AI helps brands or designers comb millions of social media images and crunch it to spit out trends or gaps in the market, that they can rush in and fill.
You’ll find AI being used to recommend the right size of jeans to you (ASOS’s Fit Assistant), or the right style (Levis’ Virtual Stylist), or to create human-like avatars on social media to act as virtual “influencers” (@LilMiquela and @Noonoouri) for brands.
AI is even becoming the fashion designer in a fast paced industry that has to design for up to 52 seasons in the global market. Amazon and Myntra are two such retail giants using AI to create its designs, which have directly contributed to sharp rise in their sales.
It’s clear this powerful rise of AI has created some tension: namely, the fear robots will be replacing designers, models, and AI will take over people’s jobs. This is not an unreal fear, given that AI is actually doing the job of a designer in several cases, a digital “supermodel” called Shudu is the face of important fashion campaigns, and Sophia the Robot is on the cover of magazines.
On the other hand, there are many problems with algorithms that we are just beginning to discover. Racial bias, for one. Or a tendency to operate within rigid gender norms, all human biases built into programmes. We’re already seeing evidence of this: Amazon Echo Look (a personal assistant that scores an outfit) apparently gave higher scores to navy or muted colours rather than brighter shades, revealing a corporate Western bias in defining what it means to “look good”.
And there are murmurs of other concerns too. Is fashion too now in the hands of start-ups and entrepreneurs fed by VCs? Being stylish can be outsourced to something like AI — but can taste? Can AI account for change in people’s preferences as they age, as they experiment and evolve, or take personal and political stands?
Exciting ideas are emerging — new kinds of fabric that integrate technology, smart mirrors that can answer your “how am I looking” question, apparel that can be re-programmed rather than thrown. But there’s much more thinking to be done on how to apply AI that solves the real problems of the fashion industry — labour issues, diversity, the waste created, sustainability — and integrate it with human design in collaborative ways. At this moment, the problem AI is solving for is limited to the retail experience.
No irritating sales staff standing too close to you. No racial profiling of customers. No waiting in queues at Zara sales for the trial room. No inefficient check-outs. Fashion is stifled by terrible shopping experiences. In large part, AI solves this. The other problem it’s trying to solve is a time deficit. People don’t have time, they assume, to spend in a mall going through hundreds of racks of clothes. They now have online fatigue — clicking through hundreds of photos on websites essentially is the same thing. As Stitch Fix co-founder Katrina Lake says, “Having millions of options for jeans isn’t actually helpful when you want a pair of jeans that fits.”
This is all good in theory. But in practice? Fashion bloggers and YouTubers have dozens of videos online “unboxing” their fixes from Stitch Fix and other subscription services as their dog/cat/children wander around behind them. It seems like the problems aren’t always being solved after all. One well-known YouTuber (with 10 million subscribers) documents how it took her 20 minutes just to answer the 51 question survey of Stitch Fix, as compared to the other services (which also took a fair bit of time!). Then was all the time not saved in returning things. Several bloggers complained about the material, the quality, how they were getting it wrong, or were out of sync with the current season, or had issues with their pricing.
After watching at least two dozen of these unboxing videos (and there were some good ones too, lots of appreciation for their no-plastic no-bubble wrap packaging) there seems to be a clear trend in what Stitch Fix sends to the average female client — a floral strappy maxi dress, a pair of jeans, at least two polyester tops with some detail element or a plaid shirt and a piece of jewellery.
“I felt unheard,” said one user. “This is random crap,” said another. “It's almost ridiculous how well my stylist knows me,” said a third. “This looks super cute,” they all say of the items they keep. Everything looks similar!
And this is probably the biggest problem. Ironically, at the heart of fashion is the potential for a person to create their individual expression of style. But for a brand like Stitch Fix, a good client is often determined by how much they look like other clients — so that the algorithm or machine learning can serve a wider net of populations. AI innovations will serve mass retail, and mainstream looks so we have a whole population of stylish people. But is the point of engaging in fashion to look like a more authentic version of yourself or to look dapper — but like everyone else? No amount of AI can create a Frida Kahlo or Lady Gaga; fashion delights us with its element of surprise, or when it cracks open a space for people’s own quirks to shine. An algorithm can do many things, but it can’t be you.
Manjima Bhattacharjya is the author of Mannequin: Working Women in India's Glamour Industry (Zubaan, 2018)
Find latest and upcoming tech gadgets online on Tech2 Gadgets. Get technology news, gadgets reviews & ratings. Popular gadgets including laptop, tablet and mobile specifications, features, prices, comparison.
As Guftagoo goes off air on Rajya Sabha TV, host Irfan weighs in on its historic run and state of film journalism today
Host Irfan looks back at his nine-year-long journey on the show, the celebrities who continue to evade him, and why a Hindi show on a public broadcast channel did not become a celebrity destination in the era of public relations and corporatisation
Hamlet, Shakespeare's greatest villain: A charismatic manipulator who persuaded the world of his heroic nature
Hamlet is a self-centred, entitled, manipulative, callous bully. However, he is also intensely charismatic, so much so that he has persuaded the world to share his Hamlet-centric view. That is what makes him a villain of genius.
Facebook introduces an AI-based software, M2M-100, that can translate 100 languages without relying on English data
Facebook used a number of scaling techniques to build a universal model with 15 billion parameters.