The Edit is our weekly show, where Social Editor Grace Robinson quizzes editor-in-chief Ben Hanson on some of the most significant fashion and technology stories from the past seven days.
This week Grace and Ben examine AI-powered motion analysis in garment factories, the promise and risk of AI shoplifting detection, the US push to rebuild textile manufacturing, the scrutiny facing social media and AI around teen safety, and three revealing technology investment stories.
This show is available in video format, embedded below, as well as audio, on Apple Podcasts and Spotify.
For deeper interviews with brand and technology leaders, our prestige interview show releases on Thursdays everywhere you find podcasts – and will be coming to YouTube soon.
Subscribers to our mailing list will receive notifications for both shows.
Grace Robinson: Welcome to The Edit from The Interline, the show where we run a quickfire analysis on our pick of the most important fashion and beauty technology stories from the last seven days. I’m Grace, the Social Editor, and I’m joined by Ben, the Editor-in-Chief. Together we have around 25 minutes to give you our analysis on the stories we think really matter.
Ben Hanson: Hey, Grace. How are you doing? You’ve got a different background this week.
Grace Robinson: Yeah. I’m just at my friend’s chalet, so a bit different, but luckily still a place to record.
Ben Hanson: Hey. Well, you know what? I was away the previous couple of weeks, so I think it’s fair that we trade. Let’s do it.
GSD RealMotion: AI-powered motion analysis on the shopfloor
Grace Robinson: So the first thing that we’re talking about is AI because it’s been a very AI-heavy week, but I suppose we have to go where the news is. And I think it’s fair to say that it’s been a bit of a broad spectrum with the AI stories: it’s neither AI good nor AI bad, but a bit of a balance.
The first story is one that you actually sent my way, and it’s something that The Interline published just a few days ago. So Coats Digital, which owns GSD Labour Quantification, has launched something it calls GSD RealMotion, which is essentially an AI-powered motion analysis for garment factory operations. So I wanted to know what prompted you to pick this story. I’m going to hazard a guess this is a bit of a testament to just how far AI is being embedded into the supply chain, but let me know what you think.
Ben Hanson: You’re correct: it is a bit of a testament to how AI is being embedded in the supply chain. Labour quantification is not the official term for this; it’s the one I’ve ended up using over multiple years of looking into it. GSD is the accepted — and I mean that in the most literal sense; it’s accredited by the International Labour Organisation — method for what you refer to as a motion study or time study: quantifying the amount of time that it takes to do a specific factory operation. That can be a sew operation, a pick operation or a put operation — all of the things that happen on a sewing line for somebody to attach one pattern piece to another. It is scientifically derived and data-backed. It’s not the only one of these, but it is the most prominent one for apparel, at least. There are comparable things in footwear. SATRA has one called TimeLine, I believe.
But the long story short is that GSD, in general, as a methodology, is a way of mathematically and scientifically determining how long it should take to do something in a factory. You stack that up and it’s how long it would take you to make a product.
It also includes what you refer to as in-cycle time and out-of-cycle time: things they refer to as value-added, and things they refer to as not value-added. From that, you can start to work out standards and allowances for people going to the bathroom, taking breaks between work and things like that. So it’s the foundation upon which a lot of costing, negotiation and relationships between brands and their factory partners should be built.
Now, it’s not universally used for that standard. A lot of people will talk about SAMs — standard allowable minutes or standard minute values — and SMVs, depending on where you are. But what they’ll actually do to get to those values is historical averaging, based on how factories have performed before, and costing based on what ought to be. Costing in low-wage areas, in particular, is done through historical averaging and guesswork rather than this scientific side of things. The reason is that these kinds of method, time and motion analyses can be tricky to implement. They can be tricky to centralise around. Although they are a standard, they’re not a standard that everybody has instinctively embraced.
The way those standards are built is literal stop-watching. They use that in a glib sense, but each of these operations is assessed by somebody there with a racing-style start-stop, and those people are called practitioners or assessors, I think. Elevating AI to the status of being a foundation for this kind of thing — being a foundation for pairing operations that you can see in video to pre-existing defined operations with standard minute values, or even creating new ones — is a massive vote of confidence, from a company that is really tightly embedded in the supply chain, in the accuracy, robustness and accountability of AI in these processes.
Now, don’t get me wrong: there’s human review in all of this. But philosophically and practically, this is probably one of the biggest votes of confidence for the fact that you can take AI, in a nebulous sense — it’s a vision model applied on top of video footage — and use it to do things fundamental to relationships between brands and factory partners, to costing, to everything else, and to a standard that is fully independently accredited by the International Labour Organisation. I have not spoken to GSD about this since I saw the news; it only came out a couple of days ago, and it’s niche in the sense that some people have glazed over in that description. I think it’s a big deal, and I think we’re going to see some things come out of this in the near future.
AI shoplifting detection: Iceland’s savings meet Sainsbury’s false positive
Grace Robinson: For the next story, we’re actually going to pull on the same thread of AI detecting things from video, but in a very different setting: supermarkets. So this week, Iceland — not the country, but the value supermarket, which is known for its cheap frozen food here in the UK — announced that it had cut its theft-related losses in stores by 80% after implementing what the stores have dubbed AI crime-detection technology. Shrinkage is actually a very real issue for retailers in food and beverage, but it’s also a massive issue in fashion, where theft often happens in fitting rooms, which grocery stores obviously don’t have, but it is a big issue in fashion.
Based on the success of this, Iceland is now apparently rolling out AI-assisted face detection as an extra lever to prevent shoplifting from known prior offenders. But, really interestingly, another story this week actually comes from Sainsbury’s — another UK supermarket that does actually sell clothes in its stores — suggesting that this kind of face-detection technology might actually create some issues. Sainsbury’s has paused its use of Facewatch, which is software that’s also used in other UK retailers, after an incident in which a person was escorted out of their store in London after Facewatch incorrectly identified him as a prior shoplifter.
So it’s a very mixed bag with this technology, and it’s also the flip side of the same class of AI that we’ve just been talking about with GSD RealMotion. I wanted to know what your take here is.
Ben Hanson: You’re correct that this is a very UK-centric story, by the way. We are going across the pond for the next one. You’re correct that it is the same class of technology: GSD RealMotion works based on phone video or other video sources of people performing sewing operations and factory operations, and it then uses an AI vision model to derive the operations from that. Facewatch and other detection technologies for in-store use seem to work the same way. They are not architected on top of new cameras or new sensors or anything like that.
As far as I understand it, Facewatch works based on CCTV feeds that already exist and are very prevalent in retail here in the UK, and the other one, which is not named, but that Iceland has already used to reduce theft-related shrinkage by 80%, is also layered on top of existing CCTV video feeds, but that one is done at the product and shelf level. So the Iceland one that they have is not looking for people; it’s looking to track products and where they have left shelves and then subsequently not been purchased or not been returned.
It sounds like one of those is more reliable than the other. It sounds like object detection is probably slightly more accountable than face detection. They’re both fundamentally the same thing. Now, what we’ve just said about GSD being a real vote of confidence in how far you can embed AI in very sensitive operations upstream demonstrates there’s a lot of variability here, because these are very sensitive deployments downstream. Very sensitive deployments in the sense that they are culturally part of the same issue we’re seeing in the US with Flock cameras. Those are not used in retail; they’re used as licence-plate detectors. But there is a huge groundswell of anti-Flock sentiment in the US, which stems from people not wanting to be surveilled.
The reasons that people don’t want to be surveilled, generally speaking, are that they want to protect their own personal privacy. But nobody likes being accused of a thing that they didn’t do, and nobody likes knowing that the machinery exists to pin something on them that they didn’t do. In the early days of generative AI, in particular, there were a lot of stories about lower courts and things in the US using it for traffic violations and things along those lines, to basically automate a bunch of that stuff, and a bunch of folks who were not rightly accused ended up falling foul of it. It seems like that’s becoming the case here, with the face-detection side of things as well.
Now, the part that worries me a little bit about this is not that sewing-operation detection is seemingly better than basic face detection. For all we know, that could just be down to the quality of the CCTV feeds or something else. It’s that all of these things can get better. Generally speaking, the trajectory of AI is towards progress, maturity and greater success and use in verifiable domains like this, and right now we get to have a conversation about what happens when face detection in shoplifting scenarios goes wrong. I’m equally concerned about what happens when it goes right in the longer term. This is a case where I understand the retail industry’s prerogative to reduce shrinkage, but it walks us straight into a cultural quagmire that I think people need to be keeping watch on — what’s happening with Flock in the US, to see how that’s going to go.
FutureTEX: US textile reshoring meets defence spending
Grace Robinson: The next one is actually a quicker story and there’s no AI in it, at least not outwardly. And we are also crossing the Atlantic for this one. So the Pentagon and the Department of War have partnered with Georgia Tech and also other bodies across education, NGOs and industry to modernise textile manufacturing in the United States to support the defence industry. So the initiative is called FutureTEX and apparently it’s promising up to $480 million of domestic investment over the next decade, starting with $36 million in the first year. So I know you’ve done a bunch of interviews with people in manufacturing, materials and ingredient innovation over the past few months. What’s your perspective on this story?
Ben Hanson: The people who come to mind are Stephen Bates from RHEON Labs and Arne Arons from Unspun. And the most relevant interview for this is one I did towards the tail end of last year with Nick Reed, who’s the CEO and founder of Neem London, which is a menswear brand. Pretty universally, these people agree that being able to innovate in the way we create materials and produce fashion is a necessity if you’re going to reshore these things. I think nobody believes that taking the offshore sourcing and manufacturing model currently in effect in China, Bangladesh, Vietnam and so on, and relocating it to America, France, the UK, Italy or anywhere else, is the way this should work, because it inevitably leads to a massive increase in prices.
We talked about New Balance previously, and we commended them for showing that made-in-UK or made-in-USA shoes are significantly more expensive than the ones made overseas.
The other thing all of those people agree on is that, in order for brands to be able to place a lot of domestic orders, either for materials or for manufacturing, the infrastructure for that has to exist. It’s like a chicken-and-egg scenario where domestic factories and domestic mills are not going to spend a fortune of their own money on building out on-demand, low-MOQ, innovative capacity if brands are not going to then place orders for it. Brands are not going to build collections on the assumption that that capacity exists until that capacity actually exists. So you end up in a weird cycle, which is what Nick from Neem talked about. Now, I think everybody also agrees that state-level sponsorship and subsidies and other programmes are the way to bridge that gap. They are the way to shore up the domestic supply chain for materials and production.
Now, $480 million is not a lot of money to do this in a country the size of the US. $36 million in the first year is absolutely not a lot of money, but it is a step in the right direction. And I think the interesting part for a lot of this is that you’re not going to get that kind of investment and subsidy from governments who just want to reshore textile production or manufacturing. I don’t think the political will is there. I don’t think the industry ambition and kind of achievability is there.
Every country, but the US in particular, is spending a lot more on defence at the moment. If you can successfully hook those subsidies, investments and sponsorships from the government into defence contractors and ambitions, the upshot of that is that you end up with a domestic industry for textile production, or a larger one, or a domestic industry for garment manufacturing, or a larger one. That’s what it’s going to take, I suspect. And that won’t be wildly different from what it took around the Second World War or other initiatives when the kind of domestic industrial build-out happened as a function of defence, and then afterwards those same logics around assembly lines and everything else were repurposed for the automotive industry, heavy steel or all those other kinds of things.
I think it’s sad in a way that that’s what it takes, but history tells us that that is what it takes. And I suspect that this will be where a lot of the success for upstream innovation ends up coming from, from a pure ‘who’s paying for it?’ point of view.
Teen safety: Meta’s lawsuit and OpenAI’s ChatGPT for Teens
Grace Robinson: I was hoping that, for the next story, we could have a fun conversation about Crocs’ new mascot, Niles, which the press release we got sent calls “a bold new era of character-driven storytelling”, but which is literally a person in a crocodile suit. I don’t know about you, but this kind of whimsical storytelling for the whole family is something that I really enjoy. But instead, unfortunately, we’re going to talk about two things that, in a much less wholesome way, have also been marketed across generations, but now they’re facing a bit of a reckoning for it.
The first topic is social media. Meta is the subject of an ongoing case brought by 29 US states, accusing the company of knowingly allowing Facebook and Instagram to contribute to negative health outcomes for minors. You and I have talked about social media bans before, one of which is coming into force here in the UK next spring, but this is something very different: active, large-scale litigation based on the foundational principle that social media is addictive and harmful.
Then we’ve also seen this week that OpenAI, probably in response to this trial, has announced that they’ll be launching a new ChatGPT for Teens, which includes built-in safety protections, controls and structures designed to support learning. All of this is aimed at people between 13 and 17. There’s obviously so much about all of this that we don’t really know, but I wanted to get your take on what you think this means for fashion.
Ben Hanson: This is a long question, but I think my answer is quite short to this, actually. We have previously talked about social media bans, which are not exclusive to the UK, and similar things are being debated in France, although it’s struggling there. Australia is obviously at the forefront of a lot of this. All of this is based on teens and young people being the population most at risk when it comes to the deleterious effects of social media and potentially ongoing parasocial relationships with AI.
I think fashion brands are aware of this; we’re not telling them anything new. I think a lot of fashion brands don’t really advertise to 13-to-17-year-olds, although I also wish we were having a conversation about Niles because I like the dude in the crocodile suit. Give me more mascots, and generally speaking we should probably have one at The Interline at some point.
But all of that is a lightning rod for a wider conversation about social media’s negative impact on people of all ages and backgrounds, and potentially the same with AI as well. I’ve recently done an interview with somebody who works on the fintech side of things — the deferred-payment stuff. I had a conversation with him about something that you and I have talked about relatively recently, which is the idea of AI being persuasive and encouraging people towards transactions, and those transactions then becoming more seamless as a result of deferred payments, buy now, pay later and so on.
You can extrapolate a lot of that to Instagram, and you can extrapolate a lot of that to TikTok. I think this is the beginning of a wider anti-social-media reckoning that’s going to spill over into AI. And for fashion’s purposes, the concern is very simple: where do you put yourself? Where do you put your communications? Where do you put your advertising dollars?
Tech investment roundup: Higgsfield, Anthropic–Decart and Dimension
Grace Robinson: Our last story is a bit more of a segment, and I think this may be one that we end up making a weekly fixture. We get a lot of investment and acquisition news sent our way, and for this week you and I have picked the three announcements that we think are most relevant for fashion, even if they’re not necessarily fashion stories. So this week, Reuters reported that Higgsfield, the AI creative platform that’s been at the centre of a lot of conversation around generative video, has raised another $400 million, which values it close to $5.5 billion.
Then Bloomberg reported that Anthropic is in talks to buy Decart for $6 billion, and I know you’ve spoken with Decart before, off the record, about their Lucy models, so I’m keen to hear your thoughts on this.
And finally, the LA Times reported that Dimension, which bills itself as an operating system for TikTok Shop, has raised a much more modest $1.6 million. So let’s talk about what all of these stories mean for fashion.
Ben Hanson: I think we do end up making this a weekly or biweekly segment. Let’s see. And I think I can give you some quick answers to each of them. I find Higgsfield fascinating. It is by far the biggest and most public AI creative platform. It’s primarily expressed as video, and the part that I find the most interesting about them is that they do have fashion-specific pushes in what they’re doing. They have a model — I can’t remember the name. I want to say it’s called Soul, but don’t quote me on that. They have a model and a workflow that is fashion-specific, but the most interesting thing that they’ve done — I don’t know if you saw this — is that they’ve pushed out a couple of purely AI-created films starring real actors.
There was one called Cully Boys, I think, which is bizarre to watch. I would encourage everybody to spend a little bit of time looking through them just to benchmark the state of the art in video generation, but also to see just how big the community has become around a lot of this.
Higgsfield also does its own content production. It has in-house and partner studios that create films and shows, and it will do short films and open-source them, which is again a fascinating concept: open-sourcing media. It will open-source all the prompts, workflows and everything else that went into them to allow people to experiment and remix and everything else. I think this is one of those valuations that probably took a bunch of people by surprise: well, it’s just a front end on a lot of different AI video-generation models, with Seedance 2.5 being the most popular one. How does it earn that valuation? I think it earns that valuation by basically being the de facto platform for a lot of the community-building and experimentation that comes with all of this.
Now, Decart: you’re right, I did speak to them off the record a little while ago. Their Lucy model is the first — or the first I have seen — real-time video model. So the benefit of it is that you and I could be having this conversation and, with a slightly perceptible but not massive delay, the model would ingest the video stream and then generate a new video stream out of it that shows you wearing different clothing. And it’s pretty good. It has the same limitations of not being a real simulation that every AI model has, but I think this would be Anthropic making a fashion play, essentially, because yes, you can use that real-time video model to do a bunch of other stuff. It’s primarily been expressed in Decart’s marketing and everything else as a fashion application, as a new frontier for virtual try-on. So I think $6 billion is a lot of money, but maybe that’s a valuation that’s earned.
And then the final one, Dimension. I’ll use this to plug an upcoming interview I’ve done in our main Thursday interview show, coming in a couple of weeks. I’ve interviewed somebody who works in social strategy and influencer relations and creator relations and things like that, and the main takeaway from that conversation is just how professionalised the creator economy and the seller economy around these kinds of things is becoming.
I think people tend to think of live shopping, influencers and creators as being a bit scrappy, and there being a lot of unstructured money sloshing around. I think all of this is way more operationalised, professionalised and industrialised than people give it credit for. Now, I’ve not used Dimension. I have no opinion about the product itself, but this category of financial tools, logistical tools, operational tools, product tools and things for creators and sellers is a category that I see growing pretty quickly.
