AI Has Infiltrated the Fashion Industry. What Does This Mean for the Future of Apparel?
Some in the fashion biz have embraced AI fashion shows, digitally generated models, and what they could mean for retail. Others have not.
By Jodi McCormack
The sixth-floor showroom at Spring Studios in lower Manhattan was eerily quiet. The blinding white walls were decorated with five minimalistic screens, while a semi-spherical metal piece hung in the middle of the room — six screens were attached and hanging downward at eye level. The hum of traffic on Varick Street vibrated off the floor to ceiling windows, while two attendees huddled closely and whispered to one another on the right corner of the room. Another attendee seemingly snapped a couple photos and left the showroom.
Each screen exhibited colorful, mythical, and futuristic clothing designs; one design portraying a male model draped in black, robe-like clothing — an outfit that would fit well in The Matrix — while another screen displayed a gender-neutral all lace collection. The images would methodically transition from one design to the next and a QR code was stationed at each screen for viewers to peek at the rest of the collection.
A closer inspection of the clothes and models in the room reveals what makes these images unique. These designs are all generated by artificial intelligence.
This was New York City’s first AI generated fashion show; an event and competition showcasing the “future of fashion.”
Taking place in late April, the show called for designers to submit a fully AI generated collection consisting of 15 to 30 images of runway style, ready-made designs, including close-up detail shots examining textures and colors used throughout the collection. Having offered free admission to the general public on the 21st, attendees were allowed to vote for their favorite collection, while a panel of judges reviewed the top 3 selections. The declared winner will have their collection made to wear.
The infusion of technology with fashion is not a new phenomenon — augmented reality, digital clothing, and avatar fashion have been resources used by ecommerce and high fashion brands, as a tool to help promote consumer needs, particularly during the Covid pandemic. Yet the rise of AI software, including the sudden takeover of Stable Diffusion, Midjourney, and DALL-E that have been the go-to programs for creatives such as filmmakers, musicians, and designers, have had record breaking numbers in audience growth (Midjourney currently having 14.5 million registered users) due to the algorithm’s capability to produce realistic images. This new frontier in creative industries, particularly the fashion industry, has garnered both excitement and backlash from the community. These generated images are often indistinguishable from human-made work, and sparks issues in terms of how creativity can survive under the guise of technology, as well as ethical concerns regarding who’s profiting off of who’s image with the rise of AI generated models, questioning how creative industries should go forward with the inclusion of AI.
Generative AI & Creativity — Can They Coexist?
Majority of AI generative applications are prompt based, meaning in order to generate a piece of work, the AI program must be provided specific words, images, audio — or a combination of all three — to produce a result. With this prompting, creative ownership becomes blurry; who or what controls the creative product, the person prompting or the AI delivering results? Can the results even be considered creative?
“It is a dance that you have to do between AI and yourself,” said designer and AI Fashion Week participant Tami Gupta. “You’re collaborating with AI.”
Having worked with different programs, including 3D designing on CLO3D, Gupta believes that despite how easy it may seem to give AI a slew of words, sit back, and let it design the next trend in fashion, a designer still needs a background in fashion to get good, professional results. “Creativity is a must,” said Gupta, “If you’re looking at it from a fashion perspective or an artistic perspective, you can create pretty much anything. And it’s your creation… It came from your head. All of these ideas exist within the AI and what you choose to put together is completely a human thought.”
For the Canadian based designer, this was especially important when generating her 1920s inspired collection showcased at Fashion Week. With a theme of fur trimmed coats and delicate fringe — along with Indian inspired patterns to pay tribute to Gupta’s heritage — the one outlier that makes Gupta’s collection unique is the use of a cave-like environment as her runway; a detail Gupta explains could not have been generated if she did not use precise words for the program, Stable Diffusion, to understand.
“I really don’t think a fashion show in a cave would be something that AI could produce on its own… It’s just too abstract,” Gupta said. “That was super challenging because you can’t just type ‘backstage in a cave, on a runway.’ It does help that it’s generating the images for me almost instantly, but if you’re a designer, you obviously have your own ideas, and if you care about your own art and how others perceive it, you’re going to want to make it appear exactly how you want it. So everything that we’re doing with AI does take work and creativity.”
Portuguese AI fashion designer José Sobral shares similar sentiments, “I don’t think it’ll replace creatives,” said Sobral. “It’ll be used as a tool because you still need to have the background to have the knowledge, the references. I use a lot of knowledge about photography… I choose ISO, the lenses, the shutter speed…everything is taken into account. Everyone has their space and everyone can bring more to the scene, because backgrounds help.”
Sobral’s collection titled “Futuristic Old Soul” and described as sustainability with a futuristic twist, is currently in the top 10 leader board for winner of Fashion Week, along with his partner Matilde Mariano, who’s collection focuses on the infusion of nature and comfort, inspired by the Jardin des Tuileries.
Both having a background in architecture, the couple decided to shift their attention to fashion, especially once AI programs offered the pair a way to bypass extra “buffering” work that usually precedes the design portion –such as sketching or acquiring materials and fabrics — skipping straight to creating the look.
“AI makes that bridge between our imagination and the good story coming to life,” said Sobral. “Just embrace and have your creative view…tell a good story and help humanity with your story. There’s a lot of good people with good ideas that just can’t make the images. Let’s cut the buffer…the buffering will kill us all. Life is short. So let’s hurry up.”
The fast pace work flow AI programs offer in terms of cutting out the material phase not only allows creators to get down to the nitty-gritty of designing, but it also ushers in a new insight into how creatives can promote eco-friendly fashion and sustainability in their work; a responsibility that individual designers can now prompt in their designing stage.
“We’re using AI to calculate exactly the amount of garments and the amount of fabric for each piece…so no scraps,” said Mariano. “It’s fascinating that a machine can calculate exactly how much thread is going to be needed to sew a pair of pants…it’s a tool that can be used in our favor.”
“When we design our pieces, we think about the environment,” said designer Alena Stepanova. “It’s a responsibility and if we change the world with AI Fashion Week or in the future, we absolutely have to think about the environment.”
Stepanova’s collection “Entanglement” displayed at AI Fashion Week was a tribute to her homeland in Eastern Europe. Yet, designing was a fraught experience for Stepanova, especially when it came to communicating with Midjourney to deliver her desired outcome.
“I started using different words, different order of words and, suddenly the AI started listening to me, and offered me something that looked like what I just imagined,” Stepanova. “I was changing prompts until we found our unique language. Just me and AI to understand…AI is a tool that we train by using our words, that’s why it is important.”
Yet, with AI software relying on human word choice and communication, there is the question of how an AI is pre-programmed, including what possible pre-existing stigmas or stereotypes can influence AI in the generative process, particularly when faced with the issue of AI generated models.
Digital Black Face, Job Security, & AI Models
Designer Mabu Yussif despises generating facial features. As a requirement for Fashion Week, every detail in each submitted image must be completely AI generated — including the models. Yussif, who gets most of her inspiration from West African textiles, generally casts a diverse group of models in her work, yet when it comes to generating specific features on Black models, Yussif noticed there tends to be an exaggeration on certain aspects.
“The reality of it is that a lot of these AI apps botch Black faces at the moment,” said Yussif. “It really over exaggerates them or it’s just a blurry mess.”
Instead of trying to work with an AI system that lacked understanding for certain facial features, Yussif opted to cover her model’s faces with balaclava masks or blur faces out entirely, allowing viewers to focus on the clothing designs rather than the models.
“It just really makes you question what other people are typing now,” Yussif said. “I type in something like a ‘West African male model,’ but I’m scared that people are actually typing in prompts that have a stereotype attached…like big lips or Afro fluffy kind of hair.”
Representation and the use of AI generated models is a particular issue regarding job insecurity and “digital blackface” in the fashion industry. Most recently sparked by Levi’s announcement in March that the brand will be using AI models as a way to add more “inclusivity” to their e-commerce website, models and others in the industry took to social media to discuss the negative implications this will have on the industry.
Digital blackface, defined by CNN as “a practice where White people co-opt online expressions of Black imagery, slang, catchphrases or culture to convey comic relief or express emotions,” brings Yussif’s questioning to the forefront — who exactly is profiting off of a someone else’s image, particularly when marginalized communities already suffer from lack of representation and compensation from an industry like fashion.
“You don’t know if it is an old White man in a desk chair with a computer saying ‘I’m the, the voice of diversity’ when he’s designing Black models, curve models for this historically, white institution,” said model Alyssa Labrie.
The use of influential “avatars’’ on social media brings legitimacy to Labrie’s statement, particularly with Instagram avatars such as Shudu Gram — a 3D generated model that attracted a large number of followers with her symmetrical features and South African heritage. It was later revealed that Shudu was actually the creation of a 28-year-old White man.
With AI being a tool that is prompted based and often, as Yussif describes, “scraps the internet” or public domain for reference, this brings up a new level of appropriation in fashion.
As mentioned in Why We Can’t Have Nice Things, author Minh-Ha T. Pham explains the connection between social media and the tendency for brands to “steal” looks, propelling an ongoing appropriation within the fashion industry. Pham explains that, “In the fashion context, the structural flaws and biases of the public domain are most clearly illustrated in the kinds of fashion copying that goes by the name of cultural appropriation…Such entitlement creates an informal but valuable license to copy racially marked and indigenous designs that doesn’t extend to the copying of white Euro-American designs.”
Digital blackface and appropriation are nothing new. But it does pose a risk for the new emerging economy with AI.
“I’ve seen the digital blackface,” said Sierra Imari, founder of a tech-based influential social media blog Genzexec. “I’ve seen the ignorance that comes with people that are coming from a very privileged space and are expecting to bring that in this new era of technology. I just hope that creatives see that they are still in the power seat.”
With a background in the intersection of entertainment and technology, Imari believes that regardless of the issues rising from digital blackface, AI holds the power for POC designers and creatives to reclaim ownership over their own image and designs, instigating what she labels as the “Harlem Renaissance reincarnated.”
“If we look historically, the Harlem Renaissance contributed a lot to economic growth, especially in America. I think that this is a really great way for Black people and people of color to get their reparations,” said Imari. “AI gives accessibility to creatives that know how to utilize it in a business way…it’s an opportunity to start-up whatever they envision for themselves without the middleman.”
Yet there is still a looming threat when dealing with AI in creative fields: job security.
Programs such as ChatGPT have already shown the potential to take over certain occupations; According to a recent survey conducted by Resume Builder, 48% of US based companies plan to lay off employees in the hopes of replacing them with ChatGPT, leaving speculation as to how softwares such as Midjoureny and Stable Diffusion will impact jobs within creative industries.
When hearing about Levi’s statement, Labrie immediately took to TikTok addressing the brand’s use of POC models and curvy models for their own profit, while avoiding the need to compensate models for their image.
“As a working model, you worry about your jobs obviously,” said Laibre “Tech is the future… AI is already integrating into our lives, but seeing a huge household brand name not only embrace it, which we’ve seen with many other brands, but specifically say this is for diversity and we’re getting rid of our actual human workers… I know most models don’t support it.”
Fellow model Sasha Payton also discussed Levi’s decision in the same vein as Labrie, breaking down her initial reaction on TikTok. “It’s just really frustrating,” said Payton. “You have actual people that do this for a living…we are supposed to be creative and then you’re automating creativity, which is strange in itself.”
Working as a part-time model, Sobral is not necessarily worried about a potential risk of AI generative work replacing jobs in the fashion industry. “We’re going to lose the image model,” said Sobral. “But people can be so much more than just an image…we need personalities.” Mariano injected saying, “The human presence, the interactions between them, AI still lacks that a bit.”
But models are not the only ones at risk.
“If I’m doing e-commerce, maybe in one day there’s a room with five models, that means there’s a hair person, a makeup person, the stylist, the photographer, and the assistants there all fixing on this one person,” said Payton. “You’ve slashed all of those jobs.”
“You also don’t have the top models doing this kind of work when you go shopping online on any platform. It’s the smaller models, so to me, that’s a major issue,” added Payton.
This back and forth in regards to job insecurity and racial biases comes down to one major regulatory issue, “The only thing that we need to work is in ethics,” said Sobral “How is AI going to be used with human ethics?”
What Are the Ethics?
Harvard law graduate and AI ethicist Paige Lord defines AI ethics as a complicated set of moral principles. “The thing that makes that challenging is that moral principles vary from person to person,” said Lord. This subjective ethical standing is especially important for those working in creative industries, including acknowledging what “responsible AI” looks like for each individual designer.
At its core, ethics ensures a mitigation system that will limit any negative impact on society, and in regards to AI, this requires what Lord refers to as “human in the loop” — meaning a human presence is necessary to monitor AI generative work in order to check if that work is harmful in any capacity. “The place we find ourselves in right now is that there are no rules…there’s almost no protection when it comes to the use of AI,” said Lord. “That human in the loop is the best first step…where there will always be a human that’s a part of every process that AI is a part of, so that humans can catch where there might be some sort of stereotype or bias that has been coded into the artificial intelligence.”
Yet, this also brings up the issue of who exactly gets to be this “human in the loop,” and how will they understand what should be deemed harmful or not? To combat this, Lord’s second step towards a relationship between AI and humans is incorporating multiple perspectives, creating a hub of regulation in regards to what AI produces.
“For a long time, people would say, ‘we have a person here and they can give this one perspective…but, diversity in gender, race, culture, and socioeconomic diversity… all of these things really matter,” Lord said. “If you’re going to have a human in the loop, you should have several humans in the loop.”
On the basis of racial biases within generated AI work, there is also creator-based accountability surrounding what references or words are used for prompting. Yet, taking responsibility for mal-use in word choice during prompting might create another issue.
“There is going to be a time when we’re going to see companies blaming AI for what it is putting out, instead of actually putting that burden of blame on the individuals who were creating the prompts,” said Lord. “They’re going to have this sort of middleman of AI to be able to say, ‘it was the AI’s fault, it was the computer’s fault that it generated this,’ instead of saying, maybe this AI wasn’t ready.”
With this regulation against word choice, there is also a need for transparency towards the general public, particularly when dealing with software that can mimic human work. “Transparency is really important, but we can’t hold every citizen to a standard of transparency like that,” Lord said. “But we can hold corporations to a standard of transparency, and right now, companies are opting in. They have the choice of whether or not they’re going to define and create their own ethical principles that they’re going to abide by how they’re going to embody responsible AI.”
There is still a blind spot for many designers and brands when it comes to the use of AI and what ethical concerns it could potentially raise. But Lord aims to fix that.
“My goal is to be able to be that consultant for companies across industries to be able to consider how they’re using AI, and what the potential risks could be,” said Lord. “By doing that, they can protect their brand, they can help to elevate humanity, and they can be considerate about how they’re putting their messages out into the world.”
Looking Forward
Regardless of AI Fashion Week’s small turnout — and lacking the luster of a fashion show– the event still collected over 350 submissions from designers across the globe at varying ages. This access that AI has provided for up-and-coming designers is what makes AI an appealing tool for creativity. Done ethically, providing access to new tech to independent designers has the chance to adjust power imbalances within creative industries, and avoid a potential monetization of AI programs.
“Everything from TV, fashion, music, the creative side is very toxic… it’s outdated,”
said Imari “Artists utilizing these tools to enhance what they’re already doing creates a re-shift in the industry infrastructures.”
Moving forward in the early stages of AI tech will prove to be a long-drawn battle over ethical codes, creative rights, and the desire to simply advocate for unique storytelling. Yet individual morals need to be at the forefront of this new era.
“It’s always important for every brand, every designer, every creator to consider,” said Lord “Is this reflective of the world that we’re in?”
