Want the secondhand scoop?
Oops! Something went wrong while submitting the form. Please try again.
To make fashion mistakes is human, to be creative divine — and AI is taking that away from us, says columnist Genevieve Smith. More alarmingly, it’s prioritizing profitability and predictability over progress, and we need a reboot
Everywhere you look lately, there’s an oddly edited photo, sentence or video that feels a little off, and you have to wonder: did a computer make that?
We are seeing the effects of the artificial intelligence craze permeate almost every industry, and fashion is no exception. AI is everywhere, and it’s not always appreciated.
We see students face backlash for using AI to string together sentences in an educational setting — naturally, because we want them to learn.
But big brands are using AI to generate predictive design for their clothing collections, and no one bats an eyelash. I find that strange. Should they not learn too?
Whatever happened to the humanity of collective effort? Is there no value left in the struggle to improve? Are we slowly programming ourselves out of everything that isn’t easy? Why bother making anything at all at that point?
When we lean on predictive artificial intelligence to influence the choices we make for the future of fashion, we leave little room for inspiration, or growth. To me, that’s dangerous.
Fashion, and style, come with a learning curve. Any well-dressed person can attest to the awkward development phase of their style, often lovingly called “the experimental phase.”
When running decisions through a program, there’s no room to learn the lesson ourselves. If ChatGPT was around when I was getting dressed in grade 10, I may never have paired my paisley scarf with navy stripes and baggy jeans. And I would have lived to regret that missed opportunity to experiment, all these years later.
AI is a tool, and we’re using it wrong. Allow me to contextualize it in computer-friendly terms, though.
Predictive design uses data to predict future trends and forecast what colours, designs and styles are likely to be in demand. This leads designers and retailers to making data-informed decisions about what to produce next.
Predictive design also limits the potential of a collection to escape what’s been done before. That’s not really its fault — it studies what’s been seen recently. It measures how much consumers like something in a specific way: how often we search for it online, how many have been purchased, and how many in that colour can be seen pictured online and in the media.
It can count reviews and cross-compare fabric compositions, it can suggest gentle mimicry and revisitation of classic silhouettes or colour combinations, but it will never be able to create something altogether unique. It can only ever remix the sum of its parts.
If we’re trying to make data-informed decisions, we should be asking AI to revisit the review pages of the previous year’s collection and compile that feedback to improve specific garment concerns.
We should be asking AI to find ways to incorporate local materials and industries into upcoming collections, measuring the impact that using wool from closer farms would have on the supply chain and neighbouring businesses. We should be asking AI to stop trying to predict what we want and give us what we’ve been asking for: progress in the textile world.
If you want to see progress, don’t ask the program to measure metrics from 2008 onwards. Allow technological progress to influence the possibilities we have to create better things through modern development.
I want to see a pair of classic cowboy boots made of mushroom leather. I want to see a 3D-printed corset made from plastic bottles. I want the ’90s striped sweater from the Gap and I want that colourway seen across the collection and I want the wool to be from sheep close by. I want your AI to scour the comments under Pinterest pages and I want to see it draft legible patterns of previously inimitable vintage garments.
Do better. Make better.
We have the technology to study and improve on the processes of ethical fashion across history and you really asked it to design a fall collection for you? Is that not still your job?
In computer code, swap Predictive Design for Consumer-Led Education, Ecological Collaboration and Archival Conservation.
AI tools can be built to identify patterns and relationships within a history of sales data, enabling what they believe to be more accurate predictions of future demand. They can be programmed for up-to-the-minute accuracy in data trends and market conditions, so the second something changes, you can read it.
What they don’t factor in? The whimsy. They always forget about the whimsy. Successful, trendsetting style isn’t about who’s wearing what right now. Authentic style is, and always has been about wearing what we want to wear, at any time.
Watching the sales analytics on any given hot-ticket must-have item (banana yellow Frye boots, anyone?) is like watching the stock market. You can make pretty good guesses, but the next big thing is never a sure thing until it is.
Continued below
Continued from above
Hunter rain boots were huge on all the upcoming style indexes this past spring and I sourced two perfect secondhand pairs to sell. They are both still in my shop. That’s not because the data streams lied — it’s because the people loosely and largely decided they weren’t actually that popular. They were just being seen on Instagram a lot. They make your legs sweaty, too.
Reminder to program a dataset out there to measure ease of comfort or perhaps long-term visibility (worn more than once).
In code, swap Pattern Recognition with The Realization That Fashion Is Ever Evolving and to have your finger on the pulse is to be a half-beat behind. It’s okay to not churn out mass-produced cheap versions of the latest thing. It’s actually significantly better for all of us.
Companies are using AI to prototype garments using artificially intelligent design optimization, creating a fit and style that has been data-informed, but almost never tested on varied, human bodies.
The prototyping step is integral to creating garments that withstand wear and remain flattering across different body types. Prototyping is where a garment is created in real fabric and stress tested: often several versions of one prototype are created before moving onto the next prototype.
It’s an expensive part of creating clothing, so it’s no wonder we have seen a steady decrease in the number of physical prototypes being tested by designers before moving onto final production lines for distribution. It’s become too expensive to make sure the clothing fits properly before they make it.
Enter the program.
When you program AI to measure and optimize designs based on criteria specified by the business, you’re not getting data-informed design. You’re getting profit-informed design. These AI programs are tailored to make recommendations on weight reduction, structural integrity and material usage. These programs are of course marketed to us as resulting in “efficient and effective” production, but for who? The retailer, of course.
As far as weight reduction goes, I know that I don’t want my clothes to be lighter (read: cheaper, thinner fabrics). I know the company shipping them for free sure does, though. As a consumer, I want quality, longevity. If it’s a running shoe, sure, measure the weight of the lace tips if that makes a difference performance-wise. I’m just saying, maybe stop reducing the weight of things we expect and pay for to be heavy.
And “structural integrity”? That has yet to be seen in any of the modern garments I’ve inspected at the mall: lazily joined corners, crooked darts, exposed zippers, dropped stitches…this is the equivalent of letting a computer build a bridge without providing them the weight of the building materials.
For as long as we have had bodies to dress, we have had humans to tailor. The unique and meticulous nature of garment fitting is so finely tuned to details that an automated program could never think to account for, and we see it in the cheap construction of machine-tailored garments.
There is a huge gap between AI-powered textile digitization and real material performance, which highlights how much work is needed to fully recreate and represent the mechanical behaviours of each fabric, especially under stress or strain, for which the garments are tested.
As long as we have AI making choices on “streamlining” structural integrity, the more we are going to see those poorly made garments show up further down the chain: cluttering our thrift store racks before they are ultimately discarded.
“Material usage” really gets me. I have been beating a very dead horse for some time now about the way that brands are straight-cutting their fabrics to maximize production, and expect for those fabrics to perform as though bias-cut in the same material.
A good example for this is the Aritzia Midi shift dress in satin or silk. It used to be cut on the bias, which added a dimension of fluid movement to the way the silk or satin weave fell along the bodice, creating a flattering, body-skimming shape to the mid section.
When you simply sew two straight cut pieces of satin together and add spaghetti straps, as the company has done now, you get a tubular shiny sack that embarrasses everyone who walks out of the horribly lit, humiliation ritual of a change room to witness in the communal disappointment mirror. And off the unpurchased garments go to landfill.
Unfortunately, that’s a very “Human Experience,” and the program that makes recommendations for the price of fabric used is only thinking about, well, the price of fabric.
Swap Design Optimization for Integration of Garment Composition and Accurate Pattern Drafting (read: do your job).
I don’t want AI designing my clothes to be the cheapest possible to produce. I also don’t want AI designing my clothes in general.
Maybe I’m old school, but I think it’s kind of neat to wear clothes designed by people, because people are who fashion should really serve. We get dressed to feel like ourselves, to feel like someone else, to belong, to stand out. That’s a very human thing to do.
I take pride in the human experience, which is why AI will never replace my job as a thrifter.
You can program a computer to run datasets to find out whether vintage Levi’s or Wranglers are performing better on Etsy; you can even run a program that filters the cheapest bulk lots of either brand of denim and automatically bids on them.
But you can’t program a computer to scour the city for the low-rise hip huggers with a grass stain on the knee, and you can’t get a computer to scrub the grass stain out, and you can’t get a computer to sew on the torn belt loop by hand and leave a little note in the back pocket for their next owners. That program is run entirely by hand, and for the love of the find.
There’s nothing data-driven about thrifting, except for the data that shows the more you go, the better you get, the more you find and the more you grow.
Finding your style is a skill, not an equation or a set of keywords you can run through. Producing quality garments is a skill, and in a world that values profit over progress, the rise of AI is threatening to take over another key sector of the human experience: bad fashion choices.
Only these bad fashion choices are ones that go beyond making into a high school yearbook. These choices threaten to overtake our place in the artistry of design, they drain our resources, and they pollute our delicate ecosystems.
A quick Google search will tell you that AI data centres are one of the top 10 water-consuming commercial industries in the U.S., with 20 per cent of those centres pulling water from highly stressed watersheds, leading to water scarcity in the region. Fast fashion isn’t just polluting rivers overseas. It takes from everywhere.
Artificial Intelligence is a tool, and we’re using it wrong. Making bad choices and learning from them is part of the human experience, too.
Just don’t let the robots tell you what to wear or how to make stuff. It’s weird, and it’s really bad for the planet.
_____
Genevieve Smith is a fashion stylist, writer and founder of Gifts of Thrift. As a yard sale enthusiast, thrift store supporter, and die-hard environmental entrepreneur, she has spent the last two decades trying to figure out how to convince people it is, in fact, cooler to care. Her column for The Vintage Seeker, ThreadFul, covers the intersection of thrifting, secondhand fashion, ethical style and sustainability.
