Artificial Intelligence Creates Best-Selling T-Shirt Design, Replacing Even Creative Work

Two computer algorithms produced one of the best-selling T-shirts for Indian e-commerce site Myntra.

One algorithm generated random images that it attempted to pass off as clothing, while the second algorithm distinguished between images and clothes in the Myntra inventory. The two algorithms produced a feedback loop to become better at producing images that looked like clothes and better at determining whether they were similar and not identical to products.

Myntra Chief Executive Officer, Ananth Narayanan, said the computer algorithms created designs whose sales are currently “growing at 100%” and that “it’s working.” Companies are recognizing the use for artificial intelligence (AI) and are using it to determine which clothes to stock and to recommend to consumers in the realm of fashion.

An economist at Massachusetts Institute of Technology (M.I.T.), Erik Brynjolfsson, and a Carnegie Mellon computer scientist, Tom Mitchell, argued last year in the journal Science that most jobs would be affected and become partly automated. “A much broader set of tasks will be automated or augmented by machines over the coming years.”

The entry of AI into the fashion industry shows that technology could even replace jobs typically known for their creativity. Online styling service Stitch Fix, which sends customers boxes of clothing, maintains profiles of customers to personalize shipments primarily relies on algorithms that project how clients are expected to buy in a given situation months into the future. For instance, an individual is expected to expand their wardrobe if they start a new job, and the volume of clothes people tend to buy. The AI predicts the styles people with different profiles tend to favor as well, like a nurse with a child in New York.

The Bureau of Labor Statistics expects wholesale and retail buyer employment to contract by 2% over a decade relative to a 7% increase for all occupations. Some of this may be attributed to automating tasks like inventory or purchases from less stylistic demanding consumers.

An economist that studies automation at M.I.T., Daron Acemoglu, said the good news is that these jobs may still pay substantially more than many positions available to low- and middle-skilled workers in recent decades and these jobs may be hard to automate in the end.

Leave a Comment