Is AI Helping to Create Fashion's Operating System? | Fashion Made S4E9

These models are getting so large that they're just starting to have these new emergent properties that weren't originally the intention of the training job. So for example, a model that's the scale of trillions, trillions of parameters, it's going to now know how to— if you said like, okay, now create me an OBJ from this and then make sure that there's less than this many polys or something like that, it will, even though we never taught it to do that, it's seen so much stuff and has so much context that it just knows how to do it.

So Jordan, Jordan Taylor, nice to Great to meet you and great to have you here in Milan. You are a pretty amazing guy, right? So you've— I hope so. You've built a multimillion pound business. You're not even 30 years old yet. And yeah, so really great to have you here. I know you guys are obviously sponsoring the event here as well. But yeah, I'd love to know a little bit more about Viscom, about your journey to get here because I think unlike a lot of people who work in fashion, this isn't, this wasn't where you started. Yeah. You came from the automotive space, but it would be good to just find out a little bit more about your journey and how you got to where you are at the moment.

Totally, totally, totally. So a little background about myself. So I grew up in Michigan. It's like in the middle of America, Midwest, big automotive state. That's where the Big Three are in America. And I went to a place called CCS to study car design. And car design consists of a lot of similar practices as fashion. You have these drawings either then visualize, then convince, you know, a bunch of people why it should exist. After doing that, moved out to Los Angeles to work at Honda, spent a lot of time there, but a majority of my career in design I spent actually at NVIDIA doing a lot of design research, which essentially was just us trying to figure out where does AI and computer vision fit into the physical product world. And now, you know, going forward, this was kind of in 2020 or so, COVID just happened, so we all got sent home, but a lot of the stuff that we were working on, the things that we're seeing were around how AI was going to affect us as creatives. And, you know, for a lot of people that don't come from this space of design, people don't realize the world around us was actually designed by other people. It almost just kind of like appears that it's, it's there. Like these chairs we're sitting in, it's almost like, oh yeah, it's just a chair. But a lot of people don't realize someone had to like maybe draw this and design it. So I really saw the advancements of some of these technologies as a way to change my own workflow, which consisted of drawing, rendering, you know, that's a very timely process if you do it in Photoshop and using these traditional Autodesk tools. And it became very apparent to me in 2021 or so that this was going to change the way that people like me do their work. So I took the leap, left NVIDIA shortly after, and started the company in my living room with my childhood friend Kalen, who's now my co-founder. So it's been, it's been quite a journey and There's a lot more to that, but that's kind of the TL;DR of that. Yeah.

Awesome, man. Awesome. And so with your time, you were working at Honda, right? Yeah. Yeah. And with your time working at Honda in terms of design process, can you talk me through some of the design processes you would have had to use, how they transfer into the fashion world that you now know?

Totally. So that whole process consisted of mainly drawing and those drawings would then have to get rendered out using something like Photoshop because you have to— a sketch works for a certain eye, but when people that maybe aren't designers themselves need to perceive what you're doing, they need to see a 3D model, a rendering, something a little bit more higher fidelity. That higher fidelity rendering then gets turned into a physical model, which I guess in fashion would be kind of like a, a mannequin mockup of sorts. But in our world it was a clay model of the car and then that gets milled out and then you kind of then wrap that with like vinyl and kind of understand the surfaces and things. It was basically a way for everyone in the room to understand what your little sketch would look like in the physical world. And if your design gets chosen, that gets pushed, put into manufacturing essentially, and you get the A-class surfacing and ready to go to production. So this is known as the, the paper-to-production process.

Amazing. Yeah. Any Hondas that you personally designed that I've seen?

No, they're, they kind of just, they stayed, they stayed in the, they're probably in the closet of dead designs there somewhere, but nothing, nothing into production. Yeah, I was, I wasn't there long enough to. To really have something out there that you would notice.

Yeah. Yeah. But amazing. Amazing. And, um, in terms of, um, in terms of starting Viscom, I guess this is something I'd like to pull a thread at. And I guess you're, listen, you guys are here as a sponsor, but it's also important, I think, for brands as well. Um, the, the need versus capability. So we can put any amount of tech into this t-shirt, but is there a need? Right. And so I'd love to know more about how you you know, pull at that thread of how you found that the need was there. Yeah, totally, totally.

Yeah, there's, there's a big difference between searching for a problem with technology versus starting from the problem. And I very much started from the problem because it was very personal to me. I was, I was personally running into this problem where I was spending more time on the technicalities of bringing ideas to life rather than the ideas themselves. So next thing you know, you kind of lost the reason. You kind of forget like, wait, why am I rendering this again? And then 3 days later you realize like, ah, the fundamental idea wasn't even that good to begin with because I got so lost into this technicality of bringing it to life. So that's when I realized I was like, oh man, if we can really allow designers to almost, I guess, fail faster, then you kind of opened up this can of worms where they can truly continue to be creatives rather than these, these, these technical technologists that are more so just hammering away at a computer to kind of get their idea to be one for one what's in their head. Yeah. So I very much started from that personal problem. Shared with my other designer friends on Discord back in the day and just kind of like see if they were kind of resonating with it. But, and we'll probably go deeper into this, but the one thing that really stuck out to me was what they were doing with the generated images. They were taking it and further adding their own flair to it. And that's what really showed me like, oh wow, the human in the process is still super important. They weren't satisfied with just what was coming out. They were kind of like rendering over it, sketching over it. And I was like, ah, okay. Like this is going to be like a human-assisted thing. And, but yeah, that's, that's very much how we kind of defined the problem and, you know, found that need you were mentioning is very much, it was personal to myself. And even till this day, you know, I'm deep in the product, making sure it like works for myself. And I'm like, would I use this? Like that kind of stuff. So yeah, it's, it's, it's a very personal thing, I would say.

That's cool, man. And in terms of, in terms of the world you see in fashion, I think that You know, you've just touched on something really important, which is this sort of, and it comes from the design mind, and I think that's not exclusive to fashion, but that pernickety nature of always wanting to have that final, just one brushstroke or one button or one change of material or one change of flow. Yeah. And so you've kind of touched on the fact that AI, in saying that, is gonna be very difficult for a sentient AI to, to take over a human's role in the design process because of the very human nature that it requires. But if you were going to ever build something foundational, maybe a new AI within the fashion space, what do you think that would be?

Mm-hmm. Totally, totally. Yeah. And I think if I was, if I was limited to fashion and have to make something foundationally new there in the model space, I think it would be purely on translation, translating the intention of fashion designers. I think fashion designers are extremely unique in the sense that they have their own vocabulary and way of describing things that's maybe not so intuitive to something like a, uh, what you see today in like a ChatGPT, right? Like the way we describe materials and words is not even the only way we go about doing that. Words, I would say, is just part of what we do to describe what we want. We might bring in swatches, we might bring in photos from a vacation to kind of communicate an idea. It might even be a playlist, your favorite playlist that actually gets the idea across. So I would create something that's extremely multimodality in that sense where it's able to capture all the multitude of intentions that a designer might have that's not just purely text. 'Cause right now you see a lot of the text and you're in a chat window and it's like, ah man, it's like, this is really how I wanna do fashion design in the future. So I would really kind of one, understand It should understand and honor the craft of how fashion designers already work and think today, but also be respectful of the different multimodalities that they use to communicate their ideas. So that's where I would really focus on.

And, and asking you a question, and this is like a tech question, maybe outside of fashion. Yeah. Like I understand what ChatGPT is and I understand bits and I understand the, the level of data that they processes, but I've also understand that a 3-year-old or a 5-year-old child has processed a lot more and, and it's not obviously through language. Yeah. So how far do you think we are from from building tech that is able to completely understand all of those different modalities? We're very close.

I think we're like at least 3 to 4 years. It's pretty wild. Even with the most recent stuff from OpenAI and the 4.0 image gen stuff, it's pretty insane. It's kind of just a hint of what's to come. I would say we're, we're relatively close. The only thing maybe stopping this right now is just simply those, the scaling laws right now to train these models that are that big. And what's really interesting is these models are no longer being trained for very specific things anymore. The specifics that we need are more so emergent properties of the scale of the models. These models are getting so large that they're just starting to have these new emergent properties that weren't originally the intention of the training job. So for example, a model that's the scale of trillions, trillions of parameters, it's going to now know how to— if you said like, okay, now create me an OBJ from this and then make sure that the— there's less than this many polys or something like that, it will, even though we never taught it to do that, it's seen so much stuff and has so much context that it just knows how to do it. So that kind of stuff is very close. It's just a matter of like, okay, do we even have enough energy on the planet to train a model that big? Like the US is starting to activate a lot of nuclear reactors now to power these things. So it's getting wild. It's getting— it's— yeah. Yeah, it's getting crazy. So we're not too far.

That's wild. I read something, I don't know if it's true or not, but I read something about one of the frontier labs. I think it may have been OpenAI. Yeah. And they did this at a couple of different places. So maybe they did it at Amazon as well, but asked their front-running model to repeat the same word again and again and again, ad infinitum. And eventually after, I can't remember what period of time, the model says, stop, I'm suffering. Yeah.

Wow. Yeah.

Wow. Yeah. But then they don't know if that's just, uh, like if it's a learned behavior to say that they should be suffering, or because, you know, like, yeah, yeah, that is wild.

Yeah, that's where things get a little bit interesting for sure. Yeah. Is it sentient? Yeah, it's, it's very interesting. Or is that a learned thing that it knows that? Yeah, it doesn't know repeatability might equal suffering. Yeah. So I don't know.

That's interesting. Very interesting. But it's very interesting you talk about the modality and the, and the nuances, and then I guess as, as like in 10 years' time, each sector may well have specific models that recognize the plethora of different nuances that their designers would use. Yeah. Yeah. So it's, yeah, it's pretty wild. Pretty wild. Okay. So in terms of designer speech, I think we were just sort of covering that there. But in terms of any other, I mean, I'd be interested from your perspective, just to hear outside of fashion what you think the most exciting thing AI could be used for over the next, you know, 5 to 10 years.

Oh man. Okay. So I—

Let's talk AI.

Yeah, yeah, yeah. I think in general, I think, you know, AI is this, it's hard to put it into the real world and it's this thing on our computer. Like what is AI? But I think humanoid robots are kind of definitely that thing that's gonna give it a physical presence that's currently lacking. We see this a lot in SF with companies like Figure, for example, that are really focusing on this and, it's, it's gonna become a normal part of our society. Like maybe starting with the elderly and you lease these things cuz they're gonna be really expensive at a, at a first cost. But I think that's gonna be extremely normal where you, these are gonna be at people's houses. You leave your home, you come back, it's perfectly clean because it just cleans your house for you. I think that's gonna be the next big thing. Yeah, for sure. Like the, the humanoid factor of these, of the intelligence. Yeah. Yeah. So, A little bit scary, very I, Robot-esque, but it's the inherent thing of just where this is all, uh, headed. I don't know if you've seen any of these FigureBots in person. I haven't. I went to their offices not too long ago, and, uh, they're just walking around the office doing stuff, and it's like the people that are working there, they're just like, oh yeah, yeah, just like they've already become so— and etherized. Yes, exactly. So how long it's going to take for society to get to that point, I don't know, but Yeah, that's definitely the one. And then self-driving cars, of course. If you've been to SF recently, the Waymos, it's very exciting.

Well, I saw in China that they've just, two companies have just been commissioned. I can't remember which city, but for flying taxis for— Ah, Chongqing.

I think that was Chongqing. Yeah, yeah, yeah, yeah. I saw that.

Well, I showed my daughter that and then I, wow, flying taxi.

Yeah. Yeah. We're entering an interesting time for sure.

Yeah, we really are. And in terms of Yeah, one of the things you said, which was that you think that at the moment in the fashion sector particularly, that they should be maybe more reassured than other sectors, that certainly in the design function, um, AI is only going to be an accelerator and an additive. It's never going to be a replacer. Yeah, yeah, totally.

Yeah, I can talk on that. So I think there's a big difference between automation and acceleration. Now, there are some things we probably do want to automate, and those are the things that maybe aren't so fun. For example, like UV unwrapping, or even in the real world, like, um, like garbage men, like taking out the trash. Like, that's something that should probably be automated. Um, but acceleration, I think, is very different, especially in the creative side of things, because it's more so based around taking into account what artists and designers— not only what they enjoy, but what they're actually good at. And I think a lot of that could be, let's say, for example, I'm starting to see this right now where designers are kind putting themselves as being more so curators of ideas rather than the ones actually always having to create them from scratch. And you kind of become this, like, you know, if you think of the diverging and converging triangles, in that converging part, that's where the human is super important because the AI is always going to— I think it's going to get to a point where it's going to be able to outpace us on pure idea creation. But as far as defining which one of those ideas is the most relevant for the problem is a very human thing. So I think that's a really big part of where humans are still going to be extremely important, but it's a different part of the value chain. But I think it's still just as valuable as being the one to do all the sketches, going on Pinterest, creating the mood boards. Like, yeah. So that's, that's kind of how I see these two buckets of automation versus acceleration. Yeah.

And you said to me, I think maybe one of your colleagues said to me you were doing a competition in, uh, I'd like to know more about that because I like the idea of, oh yeah, decentralizing fashion and furniture. And totally, it sounds like that's something that you're trying to, you're trying to spark the next generation to understand that they can design clothes, right? Yeah, yeah, yeah.

So we, it's in Milan, it's in Milan, it's for the Design Week. We had this competition where people were using Viscom and they had to design a chair. And this chair, we actually 3D printed, 1:1 to scale, and we're showcasing it now. And that essentially was, you know, I feel like the world of atoms has almost been forgotten. We've been so obsessed with the bits and pixels that you almost forget that the things that are on the screen can be real. And I really wanted to bridge that gap between the digital and the physical with this competition to really show people like, look, your ideas can be tangible. They don't have to just be an Instagram thing. Like, these can be like real objects. And I think it just really is meant to kind of spark inspiration in that realm of like, hey guys, like, The things we make don't have to stay like on our screens. They can become very real. And I just kind of wanted to give a hint of that through our technology. 'Cause what we really allow people to do is to go from paper to production really, really fast and allow them to fail faster in the process of doing that. Because prior, you know, to get a sketch and get it produced, you wouldn't know if that line you threw down was actually feasible until maybe 2 to 3 weeks down the line when it's, it comes in the mail. But now since you're able to like prototype and get these 3D models so quickly, you can fail much quicker. And I think this competition was, and if you even look at the time span in which we held it and got the chair made, it's a testament to the technology. It's very much, you know, this traditionally might have been a 6-month thing because, you know, the designs, the renders, actually making the chair, but with the speed of Viscom and then also just like 3D printing these days, it's like we were able to do this all in I think like a month or so or less, and it's pretty cool. Amazing. If you guys, if you have time, maybe, go down there. But I know Milan's kind of far from here, but it's kind of, kind of—

but you're doing it in fashion as well, right? Yeah, planning to do something.

Oh yes, as well. Yeah, so we're also planning to do this in fashion as well. Um, I think we're gonna be doing it around shoes and, uh, like a jacket of sorts, and trying to really kind of same story as far as like going from paper to production as soon as possible. Um, we're still trying to define exactly what that subject matter is going to be, but the ethos and principles of the competition itself is very much on this kind of like, okay, you can now go from something that's in your head to a real object extremely fast. And that's what I'm super excited about. Amazing.

Yeah, I think it's awesome. Yeah, I'm interested. So who can get involved in that? Can I get involved?

Yeah, yeah, pretty much anyone. Anyone. Like, we've— we had in the last competition, I think, a few middle schoolers kind of getting involved because—

thinking about my daughter. Yeah, yeah, yeah, yeah.

No, really, it's, uh, and you know, it's not to kind of like lower the, the ceiling or the, the the floor of design, but more so just accessibility of it is now, you know, anyone can kind of like really be a player in the space. So yeah, anyone can get involved.

Amazing. Yeah, I think, I think that's another thing that I'm starting to notice, which I like in fashion, is that more movements are beginning to happen to democratize the skill set, particularly the creative one, because I think that for the enterprise brands, for many of the people that come to these events, they've had to work really, really hard. Art school all the way through. Yeah, exactly. And that's not to say that one shouldn't do that, but I think that that has possibly put off a whole generation of potentially amazing fashion designers, right? And so, yeah, I think technology is gonna, you know, it's the same thing that happens in sport, in certain sports. If people don't have the opportunity to, to play them, then you get a quite an archetypal human that does that job. And I think That's quite an exciting thing with tech, I think, that will possibly change. Yeah, yeah.

Like what a designer might look like, you know, not like literally look like, but it's just like where they're coming from and the different swaths of life that can now have access to it. Yeah, it's just going to be much wider, I would say.

Yeah, yeah. Cool. And so what's next for Vizcom outside of, you know, yeah, over the world?

Yeah, yeah. Other than— yeah, no, it's funny you say that. I always say internally at the company, like, You know, in the future, if, if, if God didn't make it, hopefully one of our tools did. That's kind of like our, like our north star and what's next for us. You know, I know agents is a super hot topic and things, and sure, we're, we're thinking about that, but I, I, I really think manufacturable intelligence is huge. So really understanding like, okay, how do we really go from here's a drawing and then maybe a 3D printer starts going off right away and really kind of, so the tangibility of the ideas more instantaneous. Right now there's still a lot of translation that needs to happen, a lot of mechanical engineering, and even just like with the vendors and things like that for fashion. So we're really focusing on like, how do we accelerate the things you see on the screen to becoming real out of the factories? Yeah. So that's a, that's a big focus. Reducing that friction.

Yeah. So basically like real true on-demand production.

Yeah. Or even personalized manufacturing, right? It's like these printers and a lot of methods are getting so cheap where you can almost have your own maybe like personal factory or it's like little pop-up factories in different areas. We're experimenting with a lot of interesting, cool ideas in that realm. But that's kind of how we're thinking about it. It's like the, the after the sketch, like what's, what's, what's after it.

So yeah, that's awesome. And just out of interest, so you're predominantly automotive, furniture, fashion, any other sectors that you guys are working in? Those are the main ones.

I would say kind of like a sub-dominant one of those is gaming for sure, because they do create physical objects, but they just stay in the digital world. Yeah. So yeah, we work with a few gaming studios out in LA, but those are kind of the main ones you mentioned we focus on.

Yeah. So maybe you'll be doing autonomous humanoids.

Oh, maybe. I don't know. We might, we might be able to do that. We could. There's probably someone at some of these companies using Viscom, but That would be fantastic.

That would be great. Cool. Okay. Awesome, man. Well, listen, speaking for us tomorrow. Yeah. Yeah. Yep. Tomorrow at noon. Yeah. Thank you so much. Awesome to meet you, man. Yeah.

Thanks for having me. Appreciate seeing you. Yeah.