The lack of diversity in fashion is a perennial problem – especially for women and men whose body shapes, age, and ethnicity don’t fit the catwalk image of young, statuesque, skinny, and white.
In recent years, many clothing retailers have made bigger efforts to show black, Asian, Latina/Latino, and other minority ethnic models on their sites, though fashion remains the preserve of the young and slim, of course – especially on the catwalk.
But now AI is here to help. Or is it?
Last week, it was reported that Gold Rush-era clothing giant Levi Strauss is experimenting with AI-generated fashion models – from a start-up called LaLaLand.ai – in order to “increase the diversity of images” that different visitors see on its websites, stores, and apps.
But why not just employ a diverse mix of real models?
That a company with 2021 revenues of $5.7 billion thinks denying opportunities to real humans of “every body type, age, size, and skin tone” by using AI to increase perceived diversity is yet another example of the convoluted thinking that already characterizes the AI gold rush.
File it alongside those publishers who are employing ChatGPT to write articles that are supposedly for human eyes, but are really designed to increase machine visibility and clicks.
The concept is that if you are black, Asian, or bigger than a size 12 yourself, you will see a face and body that looks like you online: a mirror for inclusion; an interesting definition of diversity. Amsterdam, Netherlands-based LaLaLand.ai claims:
We’re here to diversify the fashion industry and challenge the status quo when it comes to inclusivity, sustainability and innovation.
There’s no doubt it believes that. But if you don’t think this is all about saving money for the client, think again. The Retail Week report says, “virtual models could cut costs by reducing the need to hire models and pay for shoots”. Ker-ching!
Even Walmart, the world’s number-one company by revenue ($611 billion in 2023), is reportedly exploring similar solutions. So what hope is there that smaller enterprises will keep investing in humans?
Levi Strauss added, “AI will likely never fully replace human models”. But ‘likely’ and ‘fully’ hardly represent a ringing endorsement of flesh-and-blood, do they?
So, where are we now? AI simulations, but sourced from and trained on millions of real images of people who once had a career: an ever finer and more exclusive version of who is and isn’t beautiful. All selected by LaLaLand coders and turned into photo-realistic avatars for an Enterprise subscription of €720 ($775) a month.
Those coders are fast becoming the world’s arbiters of style and beauty, with other clients already including fashion behemoths Calvin Klein and Tommy Hilfiger, plus German ecommerce provider Otto Group.
That money gets you a minimum of two seats, and models with four hairstyle options, S/M/L/XL body sizes, and light brown, dark brown, and fair complexions. The ‘Freemium’ free account? One pose, and one M body with a ‘light brown’ complexion. So, real diversity comes with a hefty price tag, it seems.
But no income – even a micropayment – goes to real human models or to whoever’s face and body provided the training data for your avatar; instead, it goes to two Dutch entrepreneurs and their team. Ker-ching!
Now, perhaps such a solution will increase the visibility of black, Asian, and other minority ethnic faces, and of plus-size models. And in itself, that would be a good thing. And perhaps you don’t care about fashion models’ problems in a world of war, poverty, exclusion, and disease; indeed, you might think LaLaLand.AI is saving them from a life of exploitation and artifice. Fair enough: it’s a good point.
But one thing is undeniable: innovations like this are slamming the door of opportunity in the faces of real people, especially from minority backgrounds, who struggle to be visible in a tough industry.
Those humans will now see faces like theirs gazing at them from websites, apps, and billboards, but they’ll be avatars, not opportunities for real people like them. Indeed, thanks to AI, those opportunities are vanishing. Fast.
Diverting money from humans to platforms
Another thing is undeniable, too: this broad trend is snowballing through every creative sector, and every job that used to require human knowledge, skill, talent, and experience.
Once again, wealthy AI providers and their rich clients seem to think that having to pay creative professionals and experts is a problem in need of urgent solution in the era of ChatGPT – with GPT-4 itself now morphing into a platform with vertical and horizontal plugins, of course.
The new AI-fuelled imperative ‘never pay for talent when you can fake it’ is seemingly more important than preventing climate change, pandemics, nuclear Armageddon, and rogue asteroids, judging from the tactical rush to adopt these derivative work generators that masquerade as artificial intelligences.
The network effect has already denied many creatives a living wage from their work, by persuading the public that music, writing, academic expertise, photography, design, illustration, films, and more, are something you get free with a phone, with fractions of a cent per stream going to the rights holder (who may not be the artist), but billions of dollars to the tech owner and platform.
And now AI is here to tell us that human creativity and talent have no intrinsic value at all. And that diversity and beauty are things to be faked online – on demand, based on different viewers’ profiles – rather than encouraged in the real world with investment and opportunity for human beings. Many of whom will struggle to be seen and heard amid the techno noise.
So, fashion is merely the latest, most beautiful victim. In a radio interview, Canadian model and futurist Sinead Bovell said:
A lot of models have put their entire careers on the line to advocate for a more inclusive world. Are those gains just going to go to a coder somewhere in Silicon Valley?
Apparently so – or at least to a coder in Amsterdam. But more and more such solutions will follow. According to Retail Week, another critic tweeted:
They could have chosen a woman in real life to be their ‘female advocate’, but instead chose a CGI personality who is probably run by a team of 10 guys.
Indeed, and this is the core point: roughly 85 percent of people in tech are male, and over 91% of them are white, according to techUK statistics. And those are the white guys who now get to help you fake beauty and diversity for cash.
Meanwhile, anyone who is familiar with online communities where photographers, illustrators, and painters share their work knows that they are already being populated by AI simulacrums and phantoms: most of them eerie, skinny, flawless creations, frequently young, blonde, and – above all – white.
That said, there are signs that some tech companies at least understand the ethical pressure to ensure that human beings – other than coders – get paid. Especially those whose work ends up as training data in an AI system. Adobe and Bria AI are two visual AI providers who are differentiating themselves by stressing the need for licensed content and to pay the original copyright owners.
But I can imagine we might see lawsuits against various generative AI vendors alleging scraping of reams of data off the internet without permission. Assuming that by then, it’s not too late.
My take
So, is AI enabling systemic racism, reinforcing and automating societal problems and biases?
Black American coder and IBM employee Calvin Lawrence certainly thinks so. He’s just written a 200-page book on the subject, called ‘Hidden in White Sight’, which is published next month. The introduction says:
It’s a painful reality that AI doesn’t fail everyone equally. When tech goes wrong, it often goes terribly for people of color.
Read my exclusive interview with him later this week.
