In a previous post we looked at Amazon's The Fix, an in-house affordable accessories label available exclusively to the retailer's Prime customers. Yet, as reported by the MIT Technology Review, the Internet-based retailer may be aiming not just at reselling fashion products, but at designing them thanks to a tailor-cut (pun intended) algorithm.
Developed by Amazon's San Francisco-based Lab126 - the company's research and development hub - the algorithm uses a tool called generative adversarial network (GAN). The GAN employs two deep neural networks to remember the characteristics of specific styles and trends.
Learning from raw data, the tool can transform an existing piece of clothing to fit in a particular style. In a nutshell, the algorithm may spot a trend on Instagram, Pinterest, Facebook, or in its own collection images generated by Amazon's Echo Look camera, and come up with new styles.
The results of this project were presented during a recent workshop co-chaired by Amazon and featuring academic researchers working on how machines can study and understand fashion styles and trends. There were further lectures at the conference focused on other projects that employed algorithms to identify fashion-focused social-network accounts, a person's correct size, or generate new items of clothing from scratch (a research by Tim Oates, a professor at the University of Maryland in Baltimore County).
In the last few months Amazon has been heavily focusing on fashion, and online apparel shopping is currently thriving, so this can be the next logical step. Yet some issues may arise from the use of this "fashion designer" algorithm. As seen in a previous post, Amazon's The Fix is already a compendium of the most glamorous and desirable fashion tropes or - to put it more simply - it offers consumers products that are copied by designs made by more famous brands. Now, will the results be incredibly different if an algorithm starts selecting the most popular fashion trends to create a product derived from them? Who knows. For the time being the algorithm may help Amazon spotting a trend first.
At the moment, though there are other services, such as Stitch Fix, using algorithms to provide personalised recommendations that help consumers selecting items in accordance with their tastes and needs (Amazon launched its own version of Stitch Fix - Prime Wardrobe - in June; like StitchFix, this service lets customers try on clothes before buying them). Stitch Fix also started producing a small collection of items via the Hybrid Design project, based again on the results that algorithms are providing them with, so Amazon wouldn't be doing anything radically new here.
It is a bit too early for fashion designers to worry as the algorithms developed so far haven't invented innovative trends, but they are working on existing styles and can only produce new designs after selecting a wide range of images, so we still need the creative power of a human being to provide that input. What we know, though, is that, while Stitch Fix seems to be doing well, previous projects involving fashion and algorithms have generated embarrassing results (remember Google And Zalando's risible Project Muze?), it is therefore easy to wonder if one day Amazon's dream of the fashion algorithm may turn into a dystopian nightmare of AI runway shows full of ill-fitting copied clothes. At the same time, you perversely hope the "fashion designer algorithms" will succeed, as they may help giving a rest to stressed designers (or maybe they will prompt designers to do their job without behaving like gods and stars...).
The most interesting thing, though, would be to see what may happen if historical or powerful houses - think Dior, Louis Vuitton or Prada - started using this technology allowing algorithms to remix their archives. What kind of creations would we get? Guess time will tell.
Comments