“Look cool, wear pizza, support the AI fashion scene” - David
Yesterday, a new t-shirt company launched on Product Hunt. The twist? All of the shirts are designed by AI, having learned from a training set of over one million open source images.
In fact, the entire company — from the logo to the marketing copy to the brand name (“Cross & Freckle“) — is generated by AI. Don’t worry — the makers behind the project, Paul Blankley, Tyler Becker and Sarah McBride, are real humans.
“The original idea was simply to use AI to generate the doodles. Then we started seeing other generative websites and tools pop-up on PH, like Hipster Business Name and Talk to Transformer and so we decided to see just how much of a startup you can generate with AI.“ - Sarah
How they did it: The team trained a machine learning model on Google Creative Lab's “Quick, Draw!” dataset. Then they trained the model on NYC-themed drawings, with the output being minimalist doodles of rats, dogs, pizza and pigeons.
While the concept of generating an entire retail company — from clothing design to brand name — is a novel one, the intersection of retail and AI is a growing sector.
Retail spend on AI is forecasted to grow to $7.3 billion by 2022, a significant bump from the $2B spent in 2018. As retailers are learning to leverage artificial intelligence, the customer experience gets more personalized, brick-and-mortar inventory can be optimized and marketing investments can be hyper-targeted.
And some big retailers are already doing this; Levi’s uses AI to recommend ideal sizes and placements for inventory within its stores, North Face uses IBM Watson’s cognitive computing to make personalized recs on what coat customers should buy, and online consignment store ThredUp recently released “Goody Boxes,” which use an AI algorithm to remember customer preferences.
“The applications for this technology are two-fold: as a creative tool and as an efficiency tool. For example, a designer could use a training set of trending Instagram photos and generate ideas for new clothing items based on what people are already wearing and responding to. From an efficiency stand point, such a generative algorithm could feasibly take a core design created by a brand and apply it to many different clothing item designs and dimensions, cutting out the manual and labor intensive work that is currently required to do so.” - Sarah