Our regular analysis selects one or more news stories from fashion technology, and presents The Interline‘s take on why they matter to our global brand and retail audience – as well as what they might mean for the longer-term future of fashion. As always, this analysis is also delivered to Interline Insiders by email and signing up continues to be the best way to get a fresh look at the fashion technology news, completely free, in your inbox.

Customisable and geared for growth: fashion’s new e-commerce model

One of the biggest stories of the past week is that musician-turned-designer Pharrell Williams is taking the helm as the Louis Vuitton menswear creative director. The appointment has been met with mixed feelings; some saying that Williams’s cross-functional skill set is perfect for the job, while others commenting that Williams will act more as a celebrity curator and that an actual designer should have been given the post. One thing is for certain, Williams is a strong proponent of technology and innovation. This could represent a greater shift towards technology being an integral part of luxury houses’ strategies, rather than it being siloed as a standalone concept.

One area in particular where Williams has considerable experience is in creating digital spaces that are designed with long-term objectives in mind; set up to be agile and geared for growth. This is the vision with Joopiter, Williams’s digital-first auction house and content platform. Joopiter uses commerce application Chord: an API-first, headless system — allowing for separation of the front-end and back-end systems.

Headless architectures might not sound especially exciting, but they are indicative of just how important technology remains behind the scenes of even the splashiest luxury applications. And they are a reminder that being a proponent of technology – as Williams is – can mean much more than digital marketing and engagement – especially as the world’s leading brands continue to work to get their digital houses in order.

Alongside this, big and small brands are recognising the importance of first-party data to assist in optimising customer experiences across digital and physical touchpoints. This is in the wake of Apple and Google shifting away from third-party cookies, making data-driven cultivation of consumer relationships essential.

Does fashion need a GPTZero equivalent?

Over the weekend, Google senior VP Prabhakar Raghavan shared a warning about ChatGPT, highlighting that those using the tool should be aware of its potential to give out inaccurate information that may seem convincingly correct.

Since its release in November last year, ChatGPT has garnered a vast amount of attention with people wondering what its impact on our cultural landscape would be, and marvelling at its ability to generate code, co-author content, and generally create in unexpected ways. But while people have been rightly amazed at its capabilities, that undercurrent of fear has been quickly developing – not just the fear that AI could replace humans in certain roles, but the concern that it could lead to a new way to propagate disinformation.

Enter GPTZero: a free and publicly available tool that can detect whether a piece of text is written by a human or machine. The software was developed by Edward Tian, an undergraduate student of computer science and journalism at Princeton University; whose intention was not to discourage the use of the ChatGPTs of the world, but rather to ensure that the technologies is adopted and applied responsibly. While fashion and art is judged by its creativity rather than its factuality; it does raise an important question around the products of generative AI tools like DALL-E 2, DreamUp, Midjourney, and Stable Diffusion, to name a few.

Earlier this year, The Interline’s Editor-in-Chief, Ben Hanson, analysed what effect generative AI is having on the fashion industry — where creativity and originality are paramount, and where large models often stray into copyright infringement with alarming speed.

Alongside this, as with ChatGPT, using generative AI tools in a fashion context means considering legal issues regarding ownership of what is produced. But to go one step back, if we are not told that something was generated by AI, it’s possible that we would not even know. So, does fashion need a GPTZero of its own, and to what end? To unpick the use of copyright and brand-sensitive images in the training data of generative models? To identify where inspiration crosses the line into appropriation? These are difficult, subjective questions for humans to decode – let alone AI.

It’s important to note that we’re not there yet. GPTZero uses language patterns and clues therein to identify AI-generated text and it is possible that a similar model for AI-generated images could be developed, but at the moment the approach to AI enforcement is more direct: identifying instances where generative models appear to have used a protected source as their base for creation, and then recreated something much too close to the original.

A consideration, though, is that in the long run, software like GPTZero might have the opposite effect of what it is striving for, and actually be used to improve the ‘realness’ of generative models. Machine learning models such as ChatGPT learn by minimising errors: their pre-training (the PT in the name) is based on generation and discrimination (in the computer science rather than social sense) and that networks leads, over time, to results that “pass” more often than they fail.

In principle, such ‘errors’ could include outputs that a GPTZero equivalent is able to identify as AI-generated. In the case of artwork, these are sometimes visible to people (odd numbers of fingers on models, impossible geometry and lighting on landscapes) but in the case of text they’re sometimes imperceptible to all but the closest readers. If GPTZero is applied, at least in principle, to visual use cases then it’s entirely possible that, over time, the generative model simply gets better as evading detection.

What may follow is an endless cat and mouse between generative AI getting more realistic and AI detectors becoming more sophisticated. And while that’s a battle that will surely play out across publishing and academia, it’s not yet certain if we’ll see the same pattern adopted in fashion, with brands chasing down generative models that take their styles as “inspiration” and generate results that, over time, become harder for the average user to pick out.

Investment: subjectivity and sustainability in the spotlight

Two major developments have taken place at the intersection of finance and fashion this week. The first is the formation of Looper Textile Co, a joint venture between retail giant H&M and German recycling company Remondis. The company will “collect, sort and sell used and unwanted garments and textiles and thereby extend the highest use of these valuable resources.” Secondly, PTTRNS.ai, a Netherlands-based fashion technology firm has secured €2 million in their seed round. The company has developed an AI-powered, SaaS platform for hyper-personalised fashion and eyewear shopping experiences, with the aim of reducing returns and curbing mindless consumerism. 

If you have been following the investment trends for 2023 so far, there has been a significant amount of capital invested into companies that have a focus on the following trifecta: making the fashion landscape more eco-friendly, driving efficiency, and prioritising the personal. The last few years have seen a shift of responsibility onto the consumer when it comes to recycling and more pressure to make conscious choices when buying clothing, accessories and footwear. Now, it seems that companies are willing to take back some of this responsibility and aid in the prevention of waste and the promotion of sustainable practices — and the money has followed.

The Interline believes that this is undoubtedly where investment of money and of energy should be going, and it does appear that the good intentions of those working at this junction will translate into action. This is because the pressure is on due to high visibility of where cash is going for even the smallest companies. Additionally, given H&M’s history, all eyes will be on the brand to curb its carbon footprint and take the steps necessary to meet its sustainability goals. 

The best from The Interline:

This week we published two further instalments in our series of digital product creation executive interviews, as well as an exclusive feature written by group DPC lead for manufacturing giant MAS Holdings – all extracted from our industry-defining DPC Report. We also published the results of our overproduction survey with Printful, and shared a look into what a next-generation fashion factory might look like with NEFFA | New Fashion Factory.

In the first of this week’s DPC conversations, we talk to Mark Harrop, Founder & CEO of WhichPLM, about how long-established ideas are now translating into wide-scale adoption.

In her feature, Anupama Fernando, Digital Evangelist & Group Lead for MAS Holdings, shares how she believes the time has come to tap into the existing reservoir of digital product creation expertise and craftmanship.

In the second of this week’s DPC conversations, we talk to William Wilcox, CEO of Clothing Tech, on redefining the ‘digital twin’.

In late 2022 and early this year, we worked with the team at Printful to survey the market to measure the scale, impact, and severity of overproduction in fashion. This week, we shared those results, which make a strong case for localised, digital, on-demand production – not just to minimise environmental impact, but to achieve greater agility, as well.

In advance of their showcasing a completely new approach to material and method at leading footwear fair MICAM, in Milan, from 19th to 22nd February, we worked with NEFFA | New Fashion Factory to explore what a next-generation fashion factory might look like, and why the argument could be so compelling for fashion to move beyond slow, unsustainable sourcing, cutting, and sewing and to explore entirely new possibilities.