I read a recent article in Forbes that presented a new fashion startup whose differentiation was designing clothes without designers or merchants. Instead, this company uses machine learning algorithms to analyze chat data and then creates and manufactures designs based on the trends the company discovers in the data. I do not mention the brand because I am not here to critique its business model specifically, instead, I became intrigued by the proposal if AI has the creative capability to replace designers? And should it?
To answer this, let us first define creativity. Margaret Boden defines creativity as “the generation of ideas that are both novel and valuable” further defining value as “determined by social groups” (Boden 2007). In his paper on Artificial Creativity, sociology professor Anton Oleinik concludes that “creativity is hardly possible without one’s capacity to think metaphorically, to coordinate proactively and to make predictions that go beyond simple extrapolation.” From my experience and training as an artist, I have come to understand creativity as the ability to connect disparate images/thoughts/subjects into something that is new and exciting. This ties back to Oleinik’s need for metaphorical thinking and Boden’s need for a product that has determined value.
Many explorations with AI and creative development have been conducted where AI was applied to compose music, paint like Rembrandt, and write poetry, but the big question at the end remained if the outputs were new and valuable or derivative? Machine learning uses neural networks and large data sets to learn from and generate new outputs. The limitation is that the outputs may be new in the essence that they are not a replica of any of the inputs, but the outputs are derivative in the sense that they are similar to the data set of inputs. There is no capability to consider options outside the dataset. Creativity requires stepping outside the obvious and connecting uncorrelated dots into something new and exciting.
Are we still talking about fashion here? Yes! Let me know bring this into my world now. How does this apply to the fashion industry? Let us look at the example of the AI design system DeepVogue and its 2019 win as runner up in a major fashion competition in China. After the event, headlines referred to DeepVogue as an “AI fashion designer”. In reality, the company used a group of designers to decide what the inputs were for the AI design system: key words, themes, and images. This team of designers then reviewed all the outputs and filtered which pieces would get produced and enter the competition. The designers were the ones that led the creative direction and determined the value of the outputs. The design system analyzed the data and through GAN technology created derivative variations of the inputs. This event did not mark the rise of an AI designer, but rather exemplified quite well how an AI powered design system can stimulate and drive creativity in humans. AI did not beat the other human designers. A team of human designers assisted by AI won.
Let us now consider the revival of Gucci. From 1994-2004, Tom Ford made Gucci sexy. From 2006-2013, Frida Giannini made it sensual. Meanwhile, sales growth had begun to slow down considerably and Gucci was ready for a shakeup. Then in 2014 came Alessandro Michelle and he completely reimagined the Gucci customer. His vision was irreverent, retro, geeky, and chic. He combined textures and silhouettes from multiple fashion eras and created a new way to dress that thrilled the fashion world. Could DeepVogue have come up with Alessandro’s vision on its own? No. Alessandro’s creative success did not rely on the data set of Giannini’s or Ford’s designs. Instead, he aimed to look forward and realized the shift he intuitively felt from his social surroundings. Alessandro used his social intelligence to predict where fashion should go, and he was right. From a hypothetical perspective, if DeepVogue had been brought in and fed the last two decades of Gucci fashion, new, yet derivative designs would be the output, but not a forward-looking vision on par to what Michelle created. Gucci did not need more of the same, it needed something new.
As new and existing brands consider AI and its use, the following question must first be answered: Are we a fashion forward or derivative brand? Do not get me wrong, fast fashion is highly derivative and the major brands who play in this market segment have made billions, but creative and forward-thinking brands EXPLODE in sales. As a trained and experienced designer, I genuinely believe the path to success is the following: NEVER GIVE CUSTOMERS WHAT THEY WANT! GIVE THEM WHAT THEY DON’T KNOW THEY WANT! My learnings from a year at MIT and from researching this article is that the best use of AI is not to replace human capabilities but to enhance them. AI is a powerful tool to help analyze and infer insights from data, but it is currently not capable of independent creative thinking. When we rephrase AI to mean augmented intelligence, it is easier to understand its use as a tool. During my time as a fashion designer working for a major label, I spent 90% of my time on admin and 10% designing. Imagine if the percentages had been reversed? To succeed as a new fashion brand, define your path as a follower or a leader and invest in the technology to get you there. Give your designers the tools to make them more efficient and the capabilities to be even more creative. Do not replace your creative teams with technology. To succeed, embrace the collaboration of creativity with technology!