-
art
Robbie Barrat's AI-generated Fashion Designs are a Glimpse of a Strange and Beautiful Future

Written by

No items found.
During his high school years in West Virginia, while you and your buddies were discovering new ways to get high (don’t get me wrong—your apple / snorkel piece was quite impressive), 19-year-old Robbie Barrat was tinkering 
with neural networks and AI, designing a program that trained itself to rap like Kanye West. The company NVIDIA noticed his talent and brought him on for a prestigious internship. Now, a year out of high school, he’s not attending a university—he’s working for one. Stanford, to be exact, at Purvesh Khatri’s esteemed lab, where he’s using neural networks “for tasks relating to molecule generation for medicine.” Feel bad yet? But Barrat is an artist, he wants you to know, and his passion for AI and neural networks is in service to
his creative ambitions. His latest project taps another interest of his—fashion—and uses novel neural network processing methods to extract commonalities from large data sets (in this case, thousands of images from Balenciaga lookbooks and fashion shows). He then feeds this data to another neural network, which draws from it to come up with its own interpretation of a Balenciaga outfit. The results, which he regularly posts to his Twitter page (@DrBeef\_[@DrBeef\_](http://twitter.com/DrBeef_)) along with other experiments, are strange and compelling, interesting aesthetically as well as for what they teach us about machine learning, AI, and fashion tropes.  “A question I get a lot about this is whether or not the AI is ‘creatively’ making these outfits, or if it’s just mimicking Balenciaga. I don’t think that the AI is being ‘creative’ when it comes up with this stuff, but I don’t think that it’s just mimicking Balenciaga either,” Barrat explains. “I think it comes up with interesting and strange results partly due to the fact that it only has visual information to work with. When humans are designing clothing, we know all about the nonvisual context our clothes have (like what bags are used for and why people carry them, why people wear coats, etc.) The network really doesn’t understand or care about this stuff, so instead of a bag it might instead just generate a piece of cloth for the person to hold—or just generate a pair of pants with a big compartment built
in because it doesn’t understand that bags are separate from pants, since in all the images it sees they’re always right next to each other. It also doesn’t understand symmetry at all, but I really love the asymmetrical outfits. It’s just, like, a totally alien perspective.”  Barrat is often disappointed by sensationalist _the robots are coming for our jobs_ responses that some have to his work. The way he sees it, AI is not a replacement for creativity but a powerful resource to augment it. He’s a huge fan of Balenciaga—“I own a few pieces. Not a lot—I’m not, like, super rich,” he laughs. Barrat has a great deal of respect for the designers. “I don’t think that clothing designers need to worry about this sort of thing, but I do see a future where artists and designers use AI tools like this to augment human creativity,” he tells me. “So maybe a designer could use the fashion neural network by generating outfits and flicking through them, hoping to find something interesting, and incorporate features of what they see from the AI in a human-made design. It’ll be a completely radical artist’s tool—not something that is going to replace artists.”  His work challenges our double standards around authorship, creativity, and, appropriation. Fashion is dominated by narratives of singular visionaries—Demna Gvasalia, Alessandro Michele, Tom Ford—and it is in part this association that commands respect (and justifies the dollar signs) in the eyes of the public. Would people still pay for pieces designed by a machine? Looking at these images, which Barrat made exclusively for Flaunt, how can we separate the contributions from the Barrat, the clothing designers, and the neural networks? Who is the artist here? It’s just a taste of the complicated and fascinating questions to come.