In April, more than 1,800 high-profile tech leaders and researchers signed an open letter declaring that the world is moving too fast in adopting artificial intelligence (AI). The letter states that "AI systems with human-competitive intelligence can pose profound risks to society and humanity," and asks for a six-month moratorium on further training of this technology. While many have begun to question whether AI will develop to a point where it could replace their jobs, the modeling industry is already reckoning with this reality.
While companies like H&M have been using AI for years to predict and analyze trends, generative AI — a technology powered by machine learning that's able to create new images, text, videos and sound based on data it's fed — is now being used to create digital models. In March, Levi's announced a new partnership with Amsterdam startup Lalaland.ai to diversify and supplement its website with AI models, claiming it would use them to increase "the number and diversity of our models for our products in a sustainable way."
The reality is: Companies have been using AI models for some time now without anyone noticing. That's because it's almost impossible to tell the difference between an AI model and a human.
Accessing the technology is as easy as playing around with ChatGPT or trying on the Bold Glamour filter on TikTok. While CGI models — computer-generated images of human-like figures that can be manipulated to look however the creator wants, like virtual influencers Lil Miquela and Shudu Gram — have been around for a few years, there has been a significant shift toward AI models, which use machine learning to adapt and evolve based on user interactions. This means they can become more realistic over time and even learn to mimic human behavior.
Brands are turning to firms Lalaland.ai to get these unclockable models because of how cost- and time-efficient they're proving to be. "With traditional photography, companies need to hire models, work with third parties like model agencies, hair stylists, makeup artists — not to mention undergo reshoots, which happens on average two-to-eight times per collection," Lalaland.ai CEO and Co-Founder Michael Musandu tells me over the phone. "So it makes it very difficult or at least unfeasible for a brand to showcase 20 different models wearing one product for every single collection without drastically having to increase the price on the product."
A year ago, when I first spoke with Musandu, Lalaland.ai had already created AI models for Tommy Hilfiger, Stieglitz, German retailer Zalando and Wehkamp, one of the largest online retail stores in the Netherlands. "A lot has changed since then," he admits.
Today, Lalaland.ai's portfolio has expanded with new clients like Puma, Adidas and of course, Levi's, with more brands inquiring about their AI models than ever before. "We're not here to take away opportunities from people. Our whole mission is to supplement and enrich," Musandu explains. "If a brand really cares about inclusion and body positivity, then they shouldn't scale down on traditional photography."
The public backlash from the Levi's announcement was tough for Musandu. "I took it a bit personally because, as a person of color, I empathize," he says. "I empathize with communities who feel underrepresented when shopping online. We built this company based on empathy."
Lalaland.ai was originally established in 2019 to address frustrations with a homogenous modeling industry and to provide a cost-effective solution. "One thing I wanted to clarify is that we're not here to take away jobs — I found many of the reactions [to Levi's announcement] very useful, so I'm so glad we're having the conversation," he says. "Now we can continue to refine and come up with strong AI principles and figure out how we can deploy the technology in a way that serves humanity."
Lalaland.ai is one of many startups creating AI models nowadays. Deep Agency, a platform that describes itself as an AI photo studio and modeling agency, offers virtual models for as little as $29 per month — a price that no human model can compete with. (The website is currently in closed beta.) Meanwhile, ZMO.ai touts on its website that its AI models can be created in mere minutes, enabling users to "visualize products on ethnically diverse models." (Its subscriptions start at $0 — yes, zero — for three AI models.)
Besides pricing, there are other factors that make the modeling industry particularly vulnerable to being replaced by AI. Firstly, models don't own their work: They don't have ownership of their image, whether in an editorial or commercial shoot. When it comes to AI, there's a grey area when it comes to their data — face, bone structure, poses and more — and who owns that, if it came to those datasets being used to train or be replicated by AI.
According to Sara Ziff, founder and executive director of Model Alliance, the advocacy group has already "received a number of calls from models who found that the rights to their body were signed away to a company after receiving a body scan."
Since 2012, Model Alliance has researched and created policies for models and others employed in the fashion industry. "While the technology may be new, the problem is already an everyday reality for models, many of whom can walk into stores and see their bodies in campaigns they were never paid for," she wrote via e-mail.
To counter these concerns, Zifff is championing legislation with a bill called the Fashion Workers Act to provide clear contracting and transparency between agents and models. "The bottom line is, when your body is your business, having your image sold off without your permission is a violation of your rights," she said.
Then, there's the issue of unionization. In many countries — including the U.S. — models are considered independent contractors under the law, and independent contractors are prohibited from unionizing. So, as AI is deployed in the fashion industry, models won't be able to receive any sort of protection.
The most pressing issue we're seeing today is AI being used in e-commerce and commercial modeling, which is often a model's most reliable source of income.
New York model Ashia Amavè finds the rise in AI unnerving. "I don't like robots," she laughs. "And when I first saw Levi's announcement, I was like, 'Where's the fake model? I almost didn't realize.'"
Having worked as a commercial and e-commerce model for lifestyle brands, Amavè takes pride in what she's achieved. "I know so many people who when they saw someone like me in Target, someone who they know, that made a huge impact on them," she says. "It was bigger than my size and my skin complexion — it was about seeing someone who lives in Brooklyn and lives authentically in Target. Using a computer-generated person and checking a box to display diversity with an AI model doesn't have the same effect."
Since Levi's initial announcement, the company released an additional statement: "AI, more broadly, can potentially assist us by allowing us to publish more images of our products… We are not scaling back our plans for live photo shoots, the use of live models or our commitment to working with diverse models." Amavè is not convinced.
"I think it's lazy to use an AI model for diversity purposes because, I mean, just look at any agency in New York City — this isn't a problem of not being able to find the right person to wear their clothes," she argues. "There's more than enough people. There's more than enough body types."
Most companies creating AI models seem to consider diversity and inclusion to be varying skin colors and body types. Still, one group often left out of the conversation are people with disabilities.
Thomas is the founder of Cur8able and a stylist, having worked on major adaptive fashion campaigns for brands like Nike, Zappos and Kohl's. "This technology is developed by mostly non-disabled people who often see disability as something to be fixed or cured," she says. "Being a champion of diversity includes having an understanding of the various demographics you're serving. This includes educating sales associates on dressing people with disabilities."
With any new technology comes a new set problems, with legal protections often taking years to catch up. The proliferation of AI — and the ensuing issues relating to trust — is no different.
"The pendulum swung towards e-commerce during the Covid-19 lockdowns due to necessity, and AI could really just force this swing back into the physical world, right? People may not trust the pictures they're seeing online anymore," Jena Nesbit, a fashion and textile designer with a background in trend forecasting, tells me.
She participated in the inaugural AI Fashion Week in April, and her recent exploration of AI has led her to question the fashion industry's future: "This could really change the way the retail landscape works. People aren't going to have a baseline level of trust in the presentation of the product because they don't know if it's really a physical person that's actually had to put that garment on."
It's difficult to believe that companies are turning to AI models to build trust or showcase diversity. At the crux of what makes AI models so tempting isn't integrity — it's the bottom line.
"AI models can be built and programmed quickly, allowing retailers and brands to showcase their goods more efficiently than they could with conventional photo shoots or fashion shows," explains Andy Ku, co-founder and CEO of Altava, a metaverse platform. "For both brands and customers, this can help cut costs and save time."
When used cautiously, AI could be a powerful tool to harness. However, history has shown us that we have a tendency to abuse technology and run it into the ground. Sometimes it's just a fleeting trend that threatens the environment (hello, crypto) or goes down in flames. (NFTs, anyone?) Other times, it can be sinister, resulting in irreversible damage to the DNA of society (like how the YouTube algorithm lead to unprecedented radicalization). Considering our technological past, what does the future of AI and modeling look like?
Geraldine Wharry, a London-based futurist trend forecaster, tells me that she doesn't like to speculate on what might happen, but that it's time to consider the implications of capitalism when we use technology.
"We're in this constant growth model, where companies want to save money, want to cut costs... But where is this going?," she asks. "Is it going to get to the point where people will no longer have jobs so they can't afford to buy a pair of Levi's jeans anyway?"
Billion-dollar companies touting cost-saving measures isn't a noble cause. But for small designers, it can be the difference between releasing or abandoning a collection. Using AI to propel independent creatives is something Drishti Gangwani, founder of the shopping destination Closr, is excited about.
"I'm honestly a vomit-worthy optimist," she quips over the phone. "And my hope is that it'll really allow emerging designers to propel themselves and beat the system in a way that they haven't really been able to do before. After all, there are so many barriers to success in the industry."
Gangwani explains that the cost of e-commerce shoots and ghost imagery (photographs of clothes that look as if an invisible model is wearing them) can be prohibitive for small brands. She hopes that developments in generative AI and synthetic models will mean that these labels will no longer have to forego this, and that the playing field will be leveled.
But where does this leave human models? Model and sustainability advocate Isabella Charlotta Poppius isn't too worried.
"I don't see models being replaced by AI, especially when a significant part of a model's career involves having an authentic voice, connecting with fans, speaking out on issues and even sharing their personal dating life," she says. "Can an AI model plant trees for a charity or be caught in public canoodling with Harry Styles in the streets of Tokyo? I think not."
However, Shaw Bernard, founder of Strut Models in Brooklyn, is less optimistic: "We should be very concerned about the rise of AI and its use in the fashion world. I was recently asked if Strut would consider getting into the AI space, and I'm not ready for that. I know that it's inevitable. I stand by the fact that modeling is an art form, and there are certain skill sets required like emoting that AI cannot do or imitate in my opinion."
View the original article to see embedded media.
Other models are adopting an "if you can't beat them, join them" attitude. Eva Herzigová revealed on April 25 that she's now a Metahuman — meaning the 50-year-old was scanned and participated in a motion-capture shoot to acquire her iconic runway walk, as well as a multitude of facial expressions, resulting in an avatar created by virtual production studio Dimension Studios, agency Unsigned Group and Metahuman, an open-source tool designed by Epic Games. As a Metahuman, brands will be able to dress and style her in ad campaigns.
"This is just the beginning," says Costas Kazantzis, the lead creative technologist at London College of Fashion's Fashion Innovation Agency. "With increased interest in immersive fashion moments, brands are expected to face the challenge of casting digital twins of diverse models."
A world where models can work remotely by renting out their AI-created selves is a fascinating possibility. However, Kazantzis warns that "we need a legal framework to govern these practices and ensure that models are treated fairly in terms of the usage of their avatar."
"3D assets can be easily imported and manipulated across 3D software and game engines," he continues. "Therefore, it's absolutely necessary to create an infrastructure that deals with reproduction issues and usage rights accordingly."
It's difficult to imagine a fashion industry without a human element. If AI can mimic the failures, complexity, ugliness and unpredictability that has led to some of the biggest fashion moments, then where does that leave us?
Despite confessing that she's worried about an "I, Robot" future, Amavè feels assured that it's her humanity and creativity that will continue to outweigh and outshine anything that technology can do.
"I really believe being me is a superpower," she says. "It's my best asset."