How AI Can (and Can’t) Make Better Fashion
Thanks to breakthroughs in AI research, computers are now unearthing efficiencies and solving problems that even the most intelligent human beings can have a hard time with. Computers can tell you the fastest route to work; pretty soon, they’ll be driving your car too. If you have enough data about a problem, it’s reasonable to assume that you can use a computer to model it and create an AI that finds a measurably good solution.
But many human creations transcend merely “good” solutions. If we gave an AI the task of designing clothing, for example, we’d get something that covers our bodies, protects them from the elements, and does so at a low cost. It might come up with a one-size-fits-all jumpsuit made from biodegradable materials that you simply toss when it’s too worn to wear any longer. Good!
However, we human beings would be less than satisfied with the solution, even if it would be a measurably better way to clothe ourselves — because the choices we make about how we dress ourselves are not entirely logical.
Consider buttons. Elastic was invented in the 1820s and zippers have been around since the 1890s. Still, you’re probably using buttons to keep your clothes fastened, even though we’ve solved the keep-my-clothes-from-falling-off problem with at least two more functional and convenient methods. There’s no practical reason we still use buttons. We use buttons because we’ve got history with buttons.
What if we put our history with buttons aside? If we had all the time we’ve spent fiddling with buttons back, what would we accomplish? Why stop with buttons? Given the computing technology we have today, we could optimize everything about how we dress ourselves. After all, at fashion’s core purpose, we’re just trying to protect our fragile bodies from the elements and other people’s eyeballs.
Unfortunately for AI, fashion carries meaning. Even super-savvy humans don’t always understand the whole story behind their style. Back to our example of buttons: men’s shirts button on the right, and women’s on the left. Nobody is completely sure why; one theory is that men’s jackets were once designed this way to allow for better access to swords. Men don’t carry swords regularly today, but their shirts still have buttons on the right.
Unwritten rules like this one, and myriad cultural and socioeconomic inputs, shape the design and meaning of fashion. It would be nearly impossible to distill fashion into a set of rules that a machine can effectively use to solve problems. In short, AI might never design a good garment on its own, because when we consider the culture around fashion, what makes it “good” is difficult to define. Despite this, AI can help us design in other ways.
Here at W+K Lodge, we applied AI to help explore a similarly culturally confusing domain — food. The choices we make about food are possibly even more inscrutable than the choices we make about clothing. Food, at its core, is about fueling your body. Nevertheless, food is also culture; people are constantly inventing new foods that don’t necessarily provide optimal nutrition. We eat them anyhow because we like them. Sometimes the trendiest foods are the least sensible things we could put in our bodies.
Our hypothesis is that people don’t really understand why they’re eating a trendy food. Fun, trendy food, like fashion, makes a splash in culture and creates trends that we can’t necessarily comprehend or predict with great accuracy.
In partnership with our client Soylent, we built a machine learning tool named Food Is Great to help us identify and react to trends in food as they emerged. We found that by listening to and categorizing social chatter among influential foodies in key US cities, we could identify food trends just as the first stories about them started to break. After a few weeks of training — first on how to identify a topical post about food, and then on how to identify the most influential posts — our algorithm performed remarkably well. We identified some trends around ketchup, energy bars, and national food days that we didn’t know existed.
There were several ways that we considered using this data. We briefed a team of writers to create videos about the silliness of human food trends, as told from Soylent’s ultra-logical point of view. We thought about introducing our office to the latest rainbow pastry craze. But, for all our ideas, we couldn’t train our AI to create a social commentary or dream up the next viral treat — we trusted Food Is Great to track our cultural quirks, not to interpret and act upon them.
The limitations of Food Is Great remind us of the importance of a human-centric process in industries where AI is being applied to creative and design challenges. Just as our writers played a crucial role in interpreting data to create content for Food Is Great, human designers have a crucial role to play in fashion design. While AI can help us to identify fashion trends and optimize for quality and speed, human designers apply a necessary understanding of difficult-to-distill cultural norms. Despite early initiatives by Amazon or IBM, even the world’s most capable technologists have yet to design an algorithm that understands fashion well enough to create culture on its own.
Food Is Great served its greatest value by inspiring our team to react to trends in food as they emerged. Today’s AI shines as a creative partner for human designers — quickly exploring culturally relevant designs in combination with available materials and manufacturing technologies. Supported by AI, designers can do their best, most human work — pushing their ideas into new territories by seeing, interpreting, reacting to, or even rejecting fashion currents before a majority of consumers even know they’re trends.
Illustration by Daniel Savage.
Frontiers is a forum for the engineers, designers, strategists, and producers at W+K Lodge to consider how the world is changing through the lens of emerging tech — and how brands can evolve to keep up. In our first frontier, Retail, we explore what it means to combine brand creativity and technology in brick-and-mortar and e-commerce.