AUSTIN, United States — The “robots” are coming — but rather than nudge out the need for humans, artificial intelligence stands to enhance the creative process — experts say. The implication of AI on design was a major theme at this week’s SXSW festival in Austin, Texas, where experts from a range of fields convened to discuss hot topics in music, film, media, art and beyond.
Entrepreneur Camilla Olson was in town to promote her fashion-tech software solution Savitude, which uses AI to recommend clothing based on a shopper’s shape and proportions. Before Savitude, Olson founded two predictive modeling companies and designed an eponymous fashion label, both of which informed her insights into solving fashion’s fit issues.
Savitude was selected to be a part of Target’s retail accelerator, Techstars, which included two pilots of the software on Target’s website. Olson said that part of the appeal of Savitude to Target was that the approach was neither too quantitative (meaning too reliant on science and numbers) nor too simplistic. “Someone who has the mathematical appreciation — engineering — will look for perfection and overkill” in solving the fit problem, Olson said. “If you have expertise [in fashion], you know where to draw the lines in product design. You have a gut feeling of what the market needs.”
Olson’s perspective reflects the growing tension between human and machine. As science gets smarter and is able to make recommendations on what is most likely to sell, traditional approaches are facing irrelevance.
Fashion designer Gretchen Jones, who is the former fashion director of womenswear at Pendelton Woolen Mills, found that her role as a designer had become more “defensive” than proactive. “I was fighting against big data that would often negate the creative design directions,” Jones said. “I was speaking through my gut and they had paperwork that could prove another black mock turtleneck was the thing that sold. But rarely can a customer tell you what they want that hasn’t been created yet, and that was stifling my ideation.”
Jones’s solution was to pursue a master’s degree in fashion at the University of Arts London, where she researched the role of data in the fashion business. What she found was surprising: she learned that data analytics can be valuable in empowering the creative process — if the business side invites the creative side to participate.
Human creativity isn’t algorithmic, it is illogical and abstract, but we can use AI to overcome the limitations of our mind.
“It’s not just guys in suits or Mark Zuckerberg dudes,” Jones says. “We need to disrupt data; it’s a tool, but not the only thing.” Designers, she said, are wise to acknowledge that customers feel that aesthetic choices are an extension of their identities, and that a designer is designing for them, rather than creating a vision that is delivered to the customer. In this way, Jones found that data could help designers understand the emotional connections that customers have with a brand.
Jones added that leaning too heavily on either the creative or the business side — whether that’s expecting a miracle by appointing Raf Simons to chief creative officer of Calvin Klein or former Starbucks executive Adam Brotman as president and chief experience officer at J. Crew — will not save fashion. “Dictatorial creativity is a failure,” she says.
Actor and entrepreneur Brooklyn Decker, who co-founded digital wardrobe app Finery with Whitney Casey, thinks that artificial intelligence will take over the role of the fashion influencer, using the computer generated “influencer” @lilmiquela (who has 740,000 Instagram followers) as an example.
“This person can be anywhere and fit any size and appeal to any audience, based on the data [the brand] layers on top,” Casey says. Decker adds, “and if the content is interesting enough, I don’t think she becomes [advertising cartoon] ‘Tony the Tiger.'”
Some experts suggest that in certain cases, it’s even possible for an algorithm to mimic human intuition. Jenna Niven, who is creative director at advertising agency R/GA, explained that “the gut” is the brain’s organic algorithm, and because a person’s knowledge base is limited to one worldview, humans can lean on AI to enhance creative capabilities by creating associations between huge amounts of data.
The increase in number of possible designs leads to more creativity, as designers see more possibilities and inspiration.
“Over time, it has been ingrained that creativity is an elusive thought process that happens deep in the sub-conscious,” Niven said. “I don’t think human creativity is algorithmic. The rest is illogical and abstract, but we can use AI to overcome the limitations of our mind.”
At a conceptual level, Niven said, fashion designers could look to AI to generate designs to come up with stimulus, in a way similar to what Google did with DeepDream, which used computer vision to alter images. “You look at that and you think, I never would have been able to imagine that before AI produced that. It’s taking all of these possible combinations and producing them really fast so you can comprehend and use that as inspiration. Funnily enough, fashion is one of the few industries that is taking advantage of AI before the curve,” she said.
Although algorithms aren’t generally creating new garments, they are being used to educate designers about what is needed in the market at companies such as True & Co., RocksBox, Rent the Runway and Amazon. Stitch Fix uses data to both inform designs for its in-house labels and to scale the capability of its 3,400 stylists, who lean on AI to curate an assortment of product recommendations.
Eric Colson, who is the chief algorithms officer at Stitch Fix, firmly believes that human designers are still very much the curators of fashion, but that machines can expand the number of possibilities that a human designer can consider. “The increase in number of possible designs leads to more creativity, as designers see more possibilities and inspiration,” Colson says. “Because apparel is both personal and emotional, a design has to strike a chord with a fashion designer before it goes into production.”
He also thinks that machines can estimate the probability of a design’s success, although it’s still difficult to predict which totally new concepts will succeed. In other words, predicating the popularity of “the cold shoulder” is “revolutionary,” but tweaking that concept with elements such as back and side cutouts is “evolutionary.”
“Machines can capture elements of style and allow us to manipulate them further. Imagine saying, ‘Take that skirt by Theory, but add a Kate Spade touch.’ Deep learning algorithms can, in theory, do such things,” Colson said. “It’s able to learn what makes Kate Spade, Kate Spade. Once they learn it, they can apply it to anything.”
Finally, Niven, of R/GA, had some encouraging news for (human) designers worried about proving their worth: “If you look at something that is mass-produced, it ends up losing value,” she says. “So if we are constantly producing garments out of an AI machine, the garments produced by AI are going to be devalued, and hence the value of a garment produced by a human is actually going to increase.”
This article originally appeared in BoF.