Generative AI is proving tough for the next generation of fashion

Instructors at Maxim Too’s Black in Central Saint Martins in London initially expressed reservations about him using artificial intelligence to create the designs for the final collection. According to the recently graduated of her master’s degree in fashion design, they weren’t sure how the concept would translate into the finished garment.

This was a few months ago, before the release of the latest generation of Midjourney and other AI tools that can generate realistic images from text prompts. Black does most of his work mid-journey now, but at the time he was using AI models he created with the help of friends in the tech field wearing clothes he owned, borrowed, or found. It was trained using his dataset of 3,000 photos of himself styled. Vintage clothing store. The new images it produced looked slightly distorted, like a computer glitch, but that was part of what attracted Black. He then turned it into actual clothes.

“I think it was a good start, especially for a traditional school like this,” he said. “I wouldn’t be surprised if students in my grade below are using it.” [AI] now i think so [the school] I would be very open about it now. ”

However, it’s less than you might think. Several educators told the BoF that many students, despite belonging to a generation stereotyped as innate tech enthusiasts, still have a mistrust of AI. .

Fashion schools are at the forefront of preparing the next generation of talent to enter an industry that could be disrupted by generative AI. The recent proliferation of powerful new AI image generators has made it much easier for anyone with a computer to come up with new design concepts. The situation provokes both the excitement of the possibilities and the fear that workers will lose their jobs and that design will be robbed of human creativity.

Schools such as the Parsons School of Design in New York and the London College of Fashion have long offered elective courses that expose students to AI-powered design if they wish. Other courses are moving more slowly and have not formally developed related courses.

But it’s not just the school’s reluctance, says Peter Jun Ho Tsang, who helped develop the Fashion Tech MBA program at IFA Paris. In this program, students come from all over the world to learn the art of fashion and often dream of working in luxury homes. They want to learn skills such as traditional pattern cutting. Technology is usually not part of the plan.

“It might be talked about here and there, but it’s not necessarily the focus yet. It’s for business students.” [and] Marketing student. For design students, it’s completely forbidden,” Tsang said.

Technology is a means, not an end

Educators discussing AI argue that they are trying to use it to push the creativity of their students, rather than using it as an easy fix.

“I ask my students to talk more philosophically about the nature of these technologies, why they were developed, how they can improve humanity in a way, and what they can do from humans. We’ve encouraged them not to doubt it too much by talking about not taking it away,” said Margarita Luca, digital lead for CSM’s fashion program. “But I’m sure there are still students who are very upset that I am involved in something like this.”

Those in favor of introducing generative AI to fashion students also say it’s important to take a critical approach. Louca said many students are now experimenting with it and encourages them to use it to explore new possibilities rather than creating a final product right away. She cited Black as a good example, noting that she’s not a fan of students simply taking mid-journeys to bring out fully formed ideas. (Black said that when Luca first brought him up on AI, she was in favor of using it.)

At Nottingham Trent University, undergraduate Imogen Hawkes used Midjourney in combination with more fashion-specific tools to create a collection of knitwear designs. Hawks made the decision to use the technology himself, but Ian McKins, lead instructor at the Fashion Knitwear Design Program, said in an email that the school is helping students use AI to design. He said he was looking for a way to let them know. However, he added, it’s still up to the designer to decide if the output is novel and relevant.

“The challenge is to give them the insight to understand what is a derivative work and what is a transformative work,” said Parsons professor Jeongki Lim. He has been teaching interdisciplinary courses on AI-powered design for several years, and he’s fascinated his students. on a fashion show. “With a machine that can quickly produce a large number of derivative works, students and we need to develop, like a pressure cooker, the ability to discern what is interesting and what is new and novel.”

Lim’s approach to teaching AI design has changed with a series of new tools. This is because the process becomes very simple. Also, since the generative AI is trained on a large amount of data collected from the internet, it may include the works of designers and artists who have not given their explicit consent, so he is free from labor and copyright issues. Discuss the problem with the student as well.

prepare for the future

Educators don’t just exist to nurture creativity in their students. They are also responsible for preparing the fashion industry workforce of the future, which may one day require the use of generative AI.

“It will definitely affect my job,” said Matthew Drinkwater, head of the Fashion Innovation Agency at the London College of Fashion. “The fact that we are talking about prompt engineering as an area of ​​expertise is something we didn’t discuss even a year ago.” means to create

The agency, which aims to predict and explore which technologies will be important in the next three to five years, began experimenting with image-generating AI systems in 2019. At the time, you had to harvest the internet to compile your own datasets. Generate new images using a high-performance computer.

“For us, to make sure we have access to those tools, to understand how those tools work, what the datasets are, how they work, etc. I was thinking about it step by step. [students] They may start to understand what images they’re looking at and how they’re generated and how to implement them into their workflows,” said Drinkwater. “The tools available to them now will change the process of how they create.”

But even at other schools that have courses that give students the experience of LCF and generative AI, the classes are elective, so students can only take them if they’re already interested in the subject. Luca, who has been in the position for less than two years, said she is still exploring CSM’s approach to technology, but in her view the school should not have a separate technology course. , says that the curriculum should focus on general skills and skills. It has a mindset that is more important than any particular piece of software and can quickly become obsolete.

Whatever approach you take, educators thinking about generative AI want their students to engage with important new technologies rather than fear or ignore them. As Mr. Lim of Parsons says, “We are not ostriches with their heads in the sand.”