Is Artificial Intelligence Good? Clothing Companies Say Yes
16 März 2020 - 08:55PM
Dow Jones News
By Jacob Gallagher
For years now, Emily Bode has had a brain-power problem. The New
York designer runs Bode, a fast-rising brand that produces
whimsical, fresh garments from aged textiles. Its signature piece
is a quilt coat, each made from one of the hundreds of one-off
throws the brand sources from a network of pickers. The issue: Her
staff regularly called upon Ms. Bode, their resident quilt expert,
to identify the quilt before they shipped it off to production, or
to describe a coat's provenance to a potential customer. She could
confirm that, yes, the quilt in question featured a 1950s "Log
Cabin" motif, but the process was tedious.
The solution came from an unlikely source: Microsoft. After a
series of conversations with the brand, the Redmond, Wash. tech
giant developed the Bode Vault, a proprietary piece of artificial
intelligence that can identify a quilt faster than Ms. Bode can
type out a text.
One of many AI applications the fashion industry has begun to
employ, the Bode Vault system -- when fed a photo of a quilt --
analyzes it for "hot points" or key design details of specific
quilt types. The technology works less like a barcode scanner,
which would analyze the picture in full, and more like a reader
interacting with the book "Where's Waldo?" It searches for the
equivalent of a "Waldo" in a given quilt, and if finding, say, six
of them, uses that data to identify the quilt variety. Before this
can happen, the computer must be fed hundreds of images of around
30 different examples of a given quilt type to create a baseline
from which the Vault can work.
While the machine can dissect and name over ten types of
patterns so far, some quilts stump it: Most "suit quilts," for
example, are irregularly hodgepodged together from old suits, and
the patterns are generally too varied for the system to learn what
a "typical" suit quilt is. Thus far, the technology also lacks the
ability to date a quilt or identify where the materials are sourced
from. According to a representative for Microsoft, Ms. Bode was not
paid to participate in the development of the technology, nor did
she pay Microsoft to undertake the project.
Exploiting AI is becoming one of the largest behind-the-scenes
trends in the fashion business. Technology companies like Edited, a
10-year-old outfit based in London, uses machine learning (a form
of artificial intelligence) to analyze the effects of pop culture
and current trends in the retail industry. "Back in the day, there
was a lot of guesswork when it came to determining what kind of
assortments that [brands] wanted to sell, what kind of trends to
put their money behind," said Daphne Duong, Edited's content &
communications manager. The reports that Edited whips up from its
data track the evolution of color palettes at menswear shows, or
dissect the driving factors behind the prominence of Scandinavian
style. Edited tracks more than 480 million products across 90,000
retailers and brands including Marni, Chloé, Zara, Mango, Tommy
Hilfiger and Puma.
In 2018, to identify key trends that were engaging its target
market, online fashion retailer Yoox tapped machine learning to
sift through social media and data on its existing customers'
shopping preferences. It then used the results to inform its first
private label collection for men and women, including items like a
plaid topcoat and a blue corduroy suit.
Artificial intelligence has also been used to influence
consumers' purchasing decisions. Labels like Levi's have used
automated chatbots that steer customers toward a new pair of jeans.
For some companies AI technology is the cornerstone of their
business model: Stitch Fix, a nine-year-old online personal styling
company, uses algorithms to generate specific clothing sets for its
customers.
Beyond the fashion industry, global spending on AI systems
overall is expected to reach $49.2 billion in 2020, a 31% increase
over last year, according to market-research company International
Data Corp. With AI creeping into more facets of our lives, a
category of fashion has also emerged to battle these potentially
invasive technologies. It includes items designed to confuse
cameras that use machine learning to identify a given individual's
face: say, a pair of goggles with LED lights running down the
lenses or a headscarf covered with images of other peoples' faces.
"There is a big distrust, maybe not necessarily in the fashion
industry, but in general around AI," said Maruschka Loubser, the
director of global brand marketing partnerships at Microsoft. Ms.
Loubser, who worked on the Bode Vault project, felt it was
important to "demystify"AI technology.
The Vault is not the first time that Microsoft has created an
AI. platform for a brand -- other retailers and labels have
deployed similar technologies made by Microsoft, particularly on
the production side -- but it is certainly its most ballyhooed
partnership to date. The Vault application is not available to the
public (Ms. Bode expressed hope that customers will be able to
download it eventually) but Microsoft is still running slick
Instagram ads touting the technology. The fact that Microsoft is
able to tout the partnership in marketing materials could explain
why the company developed the vault at no cost to Bode. As used by
Bode's salespeople in its New York City retail story, the
information the app generates can help convince a customer to spend
the hefty $1,554 for one of Bode's quilt coats.
That kind of storytelling "aids in selling," said Ms. Bode. "If
someone tries something on, it's a really great way for the
customer to understand the value." The app also has applications
inside Ms. Bode's studio where textiles are stored. When the brand
receives an order from a retailer like Browns's in London for, say,
15 "Log Cabin" raw quilts to be turned into coats, the team uses
the Vault to pinpoint where in Ms. Bode's extensive stash those
quilts are located.
After witnessing the Vault in action, one can see why Microsoft
is shouting about it -- as applications of AI go, few would seem
more benevolent, less sinister than identifying historic quilts.
For Ms. Bode, the greatest potential of the app is in preserving
those quilt stories. She was getting inundated with so much
information both from her suppliers and through online search --
but those anecdotal tales were getting lost. Each quilt listing in
the Vault has spaces where members of her team can input the "quilt
story" which could include where it was from, who made it and any
other intriguing backstories. Of course, here as elsewhere around
the industry, AI's capabilities are still relatively limited. Those
stories must still be keyed in by a human hand.
Write to Jacob Gallagher at Jacob.Gallagher@wsj.com
(END) Dow Jones Newswires
March 16, 2020 15:40 ET (19:40 GMT)
Copyright (c) 2020 Dow Jones & Company, Inc.
Microsoft (NASDAQ:MSFT)
Historical Stock Chart
Von Feb 2024 bis Mär 2024
Microsoft (NASDAQ:MSFT)
Historical Stock Chart
Von Mär 2023 bis Mär 2024