Last December, M+ hosted a fashion show in which brands displayed collections created with the help of AiDA, a locally created artificial-intelligence software. We ask both the designers and a software creator whether AI marks a pivotal point in the fashion designing industry’s evolution.
Te furious cacophony of shrieks and self-righteous wrath that’s been simmering in the embattled cauldron we call Twitter might well boil down simply to this: AI art is bad and should be banned. Indeed, the AI-generated carbon copies of the works of digital artists on social media inspire circumspection, if not plain horror, raising a question that could make science-fiction novelist Isaac Asimov’s corpse grin: “Can machines create?”
Rather than wrestling with that, perhaps we should be asking something a little more straightforward, which is whether AI can be used to produce original works in fashion design, to which the answer is almost certainly in the affirmative. For example, in 2021, Refik Anadol created the mind-blowing installation at New York’s Museum of Modern Art titled Unsupervised, which aimed to depict AI’s dreams. The abstract piece comprises a series of ever-changing images warped on themselves, generated after the algorithm was trained on the metadata of MoMA’s collections that span more than 200 years of art. After all, if a piece has enough references – say thousands of them – it might as well be original.
The same goes for fashion. At a first-of-its-kind AI fashion show at M+ late last year, organised by the Laboratory of Artificial Intelligence in Design, CreateHK, Hong Kong Polytechnic University and the Royal College of Art, young couturiers used the AiDA AI-based interactive design assistant to help “interpret” their collections. And though the pieces shown – more than 80 outfits by 14 designers – could hardly rival those in Paris Fashion Week (or even that of Copenhagen), few of them screamed “derivative” (or at least not too glaringly).
Then there was the opening look, which was nothing short of spectacular: a deep-blue catsuit with a black moiré mini-dress, and elaborate harness made out of clear beads and a bulbous handbag conceived by emerging designer Irene Siu. For the rest, the presentation felt more like a symposium rather than a fashion event. The highlights (other than the prolonged panel talk at show’s end) were the hybrids of dresses and coats à la Chitose Abe paired with Balenciaga-style nude masks, courtesy of Korean label FBESFXXK, and ruffled dresses by Hong Kong-based designer Mountain Yam who, to quash accusations of mimicking Alexandre Vauthier, put transparent plastic shields on his models’ heads and necks.
So did the use of AI in fashion design lead to original looks? On this evidence, the answer is an equivocal “yes and no”. The CEO of Hong Kong-based AiDLab and creator of AiDA, Dr Calvin Wong, explains that his invention generates two-dimensional designs based on four variables: abstract images (mood), sketches, prints and a colour palette. He tells me the output is infinitely customisable and AiDA’s modus operandi is unique. “In recent years,” he says, “everybody’s been talking about using AI to create artworks. The main operating system for this is called GAN (Generative Adversarial Network), a common tool for creativity used in visual art, music composition, animation or visual communication. The key mechanism of GAN lies in feeding the core product to the model to train it to produce random images.”
No two AIs are made equal. According to Wong, AiDA centres on designers and their ideas, putting all the various sources of inspiration together, akin to an IRL assistant (though one with far less hunger for constant validation), with countless opportunities for augmentation. In other words, AiDA helps designers find inspiration and is not, as he repeatedly emphasises, meant to replace designers – it simply cannot function without human participation, unlike GAN-based systems.
Today’s oft-spouted mantra of the “democratisation of fashion” is Wong’s main motivation. “You can still use AiDA if you have no design ideas,” he says. “Maybe you love fashion but have no sense of trends and want to create your collection. If you like a particular colour or print, you can upload those, and the system will generate options. It’s suitable for all levels of designers – from beginners to savants.”
Which makes us wonder whether this might spark fashion’s transition from dictatorship to a form of parliamentary democracy, or become the catalyst that permits more designers to push the boundaries of creativity, making the industry yet more competitive. Whatever the outcome, one facet is laudable: AiDA’s gamified approach can make fashion design education more accessible – a Rubicon soon to be crossed, given that the system will form part of the PolyU’s school of fashion and textile design curriculum.
Ahead of its December launch, AiDA has attracted a small core of fans that includes Anteprima’s creative director Izumi Ogino. The brand’s Unbounded Energy is an extension of its spring/summer 2023 collection, for which Ogino used AiDA to study Anteprima’s signature silhouettes over the past 30 years. The result, both
for her and design director Dilys Yu, was unexpected. “The brand’s existed for 30 years already and, somehow, we have a fixed mind, especially when it comes to the technical part,” Yu says. “Knitwear, for example, is all about calculation. With AI, when it comes up with a design, I want to see if we can make it in knit, simply because we never thought of it before. So it lets us think outside the box.” Despite Yu’s fascination with the magnitude of AiDA’s ingenuity, Ogino cautions about its efficiency. “Soon I’m sure the processing time will be shorter,” she says. “But currently it takes twice as long compared to our normal workflow. We need a lot of time for input, and AiDA doesn’t support enough variables.”
Her concern isn’t shared by emerging designers Eugene Leung and Dan Tse of Sydney-babel Injury, who also sent their creations down the M+ runway. Leung, the founder, designer, artist and musician, grabbed the opportunity to experience AiDA as a way of innovating new design methods and formats of fashion presentation. “The user interface is easy and spontaneous,” he says. Tse concurs, saying that “AiDA makes things more efficient”. Injury’s oeuvre transcends materiality: since 2017, the creative duo has been exploring digital fashion, especially in the context of film, for which they’ve won accolades at Diane Pernet’s ASVOFF (A Shaded View of Fashion Film Festival), and similar festivals in Berlin and Tokyo.
While it may be reasonable to assume that any system learning from physical clothes can be limited by ideas of physicality, some young designers clearly disagree. “I think AI can help with digital fashion,” says Leung. “I don’t think it’s bound by anything.” Tse agrees: “Sometimes digital fashion can be even more surreal, so it’s a good match with AI.” He also notes that he would enhance the models produced by AI rather than using them as they are. Leung compares AiDA to modern music production. “All DJs are sampling music. AI is doing the same thing. I don’t think it’s something negative. “I think AI can do many things by itself, but I don’t think it can fully take over from humans. After all, humans have emotions,” he says. “Compact discs never took over vinyl, digital cameras never took over analogue.”
Blasé though the comparison may be, it’s also surprisingly profound. For as long as designers hold themselves accountable for what they put on the runway, the critics of AI can lay their slings and arrows aside. As to whether it heaps pressure on young designers’ integrity, we’ll have to wait and see.