In this class, I talk about ways AI and cheap prediction are being used in fashion and apparel. From my interview with Dr. Tao Mei, Vice President of Cloud & AI at JD.com

You can listen here or at iTunes, Google Podcasts and Himalaya.

Related podcasts and articles:

I am a consultant and keynote speaker on how to accelerate growth with improving customer experiences (CX) and digital moats.

I am a partner at TechMoat Consulting, a consulting firm specialized in how to increase growth with improved customer experiences (CX), personalization and other types of customer value. Get in touch here.

I am also author of the Moats and Marathons book series, a framework for building and measuring competitive advantages in digital businesses.

Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.

——transcription below

:
Welcome, welcome everybody. My name is Jeff Towson and this is Tech Strategy. And the question for today’s class, how is AI changing, transforming, maybe disrupting the fashion industry? And to get to the answer, I’m just gonna sort of talk about an interview, a discussion I had with Tao Mei, who is the head of, or the vice president of cloud and AI at JD.com. And obviously, He is a technical expert, an operational expert, the person who’s actually doing this stuff on the ground. So I figured it’s better to just see what he thought. So that will be our topic. I think it’s pretty fascinating. The whole AI software digital thing, in a lot of sectors, it’s not that interesting. Fashion, luxury, really there’s a lot going on. There’s a lot of interesting things happening. So I think it’s gonna be a good topic. Okay. Before we get into the topic, those of you who haven’t subscribed, please do so. You can go over to jefftausen.com and sign up there. There’s a free 30-day trial. Try out the class. See how it goes. Free for 30 days. So hopefully that’s a reasonable deal. And I did want to really welcome all the new members, as I mentioned in the last podcast, but really seeing a, I guess, surge is the right word. in people listening to the podcast, people signing up for the class, it’s great. And I appreciate that. So of all of you that this is maybe the first week you’ve joined, the first week you’re listening to the podcast, first week you’ve signed up for the class, welcome, hopefully this will be great. And there is a sort of a step-by-step process that we go through week by week, week by week. We address new topics, new ideas, sort of put new tools in your toolkit for how to think about these questions. And then… That’s about 49% of what we do. And then the other 51% as we just talk about companies, one after the next, after the next, because these are just different, they’re individual, there’s a lot going on and that’s the goal. It’s 50% concepts and tools relating to digital and 50% just companies that are doing interesting things and pushing the frontier. And I think you learn from both. And I think the concepts really make it easier to look at what’s going on with a little bit more of a framework. as opposed to just going through random stuff all day. So anyways, that’s the approach for the class and welcome to all the new members. And of course, for those of you who are members, the obligation, the duty, the thing I nag you about every week is to do the basic homework assignment, which is try to apply this in your own life in the world around you. It’s so easy to kick back and listen and think, oh, that’s interesting. And then you get on with your day. But the stuff that sticks, the stuff that takes you forward is when you say, you take an idea, you take one of the concepts. You know, we have about 23 learning goals laid out across four to five levels now. And within that, there’s just lots of ideas. Every learning goal has several concepts to think about network effects, switching cause, virality, all of this stuff. Just choose one. within whatever learning goal you’re on and try and apply it to a company you see. I generally encourage people every week to write it down. Two to three paragraphs. If you write it, you learn a lot more. If you didn’t write it down, you probably don’t know it beyond a certain level. So take a concept, network effect, virality, whatever. Pick a company you’re interested in, then just look at it through that framework and write, here’s what I think JD is doing in terms of. rate of learning in terms of network effects, in terms of virality, in terms of switching costs. Write two or three paragraphs, just give it a shot, write it down, piece of paper and a pen is how I do it. Otherwise, you know, pull out your phone and open up the notes section and write it down and just try and apply it. You’ll learn a lot more if you just, if you do that every week. Even if you’re wrong, you know, chances are you probably won’t get it right. That’s okay. Just give it a shot, do that every week, and that’s kind of the homework assignment for this week. Apply one of these goals, write two to three paragraphs about a company, and then I’ll push you to do that every single week, and that’s the kind of stuff that really does add up. The way I do it personally is I pull out a piece of paper and I pull out a pen and I write, now I’m kind of maniacal about this, so I probably write 15 to 20 pages. Pretty much every day about a various company and I have certain frameworks I apply and that’s just my process. I’ve been doing this for years and years and years. Pretty much every day but I’m kind of, you know, it’s my thing. You know, don’t do that, that’s crazy. But you know, take a simpler version of it, two to three paragraphs about a company and it adds up over time. Okay, let’s get into the content for today. Now in terms of the content for today, the… There’s basically four concepts I want you to think about. None of these are new. We’ve touched on these in other classes, but these are sort of the four takeaways for today. Number one would be six digital superpowers, or just digital superpowers, which those of you who are in my class at Sasson or Siebes or PKU or whatever, I bring this up all the time. So I’m sure you’re sick of hearing me talk about digital superpowers. That’s concept number one for today. Number two is… sort of the five dimensions of operational marathons, which is something I haven’t talked about directly much. I’ve sort of alluded to it in different ways, talk a lot about operational marathons, but I really think there’s five key dimensions to that happening. So that’ll be concept number two, I’ll explain that. Number three, AI as cheap prediction. Well, we had a podcast on this, which was podcast 28, and that was about Ant Financial, but it was really about starting to think about AI and machine learning as a form of cheap prediction. So Podcast 28, we talked a lot about that. And we also talked about it in Podcast 31, which is a previous interview I did with Jing Dong about applying AI to conversational AI, to customer service. So that one kind of relates. If you want to read… listen more about today’s topic, look at those podcasts 28, podcast 31. I’ll put the links in the show notes and I’ll also put these four ideas in the show notes, so it’s all there. And the fourth concept for today is this idea of AI factories, human free operations, which we’ve also talked about a little bit before. Anyways, those are the four ideas for today. I’ll go into those as we sort of talk about my interview with JD. Links are in the show notes and this all goes under the learning goal number 22 for those of you who are in the class And moving up the learning goals one by one this goes under learning goal 22, which is the basics of AI Okay into the content Now AI is just software. So it’s right in the strike zone of this class It’s you know, this this class is about data technology digital software It’s all the same thing. It’s things in life that are made of bits and bytes, ones and zeros, as opposed to things that are made of atoms and molecules in the real world. Because the economics are different and as industry after industry starts doing more of this stuff, it can change the economics and the behavior. Now in practice, that’s just chaos right now. That there’s new tools coming all the time, there’s new types of AI, there’s new software tools that can do this or do that. new little tools like sensors and IOT and GPS on your smartphone. There’s just a lot of this stuff coming. And as you apply it to each industry, banking, healthcare, retail, media, communications, we just see a sea of use cases that, hey, this new tool we’re hearing about, that actually makes a lot of sense for banking. And this use case makes a lot of sense for the in-store experience of Sephora. So There’s kind of this stream of tools coming, and then that’s resulting in lots and lots of new use cases being experimented with, and sometimes new business models entirely. So it’s a lot of chaos, it’s all bottoms up. There is no grand unified theory of any of this. You just have to keep track of all these things. And out of that, every now and then, well, The use cases we could just sort of consider as, look, these are just operational improvements. That’s sort of one level of thinking, it’s just operational improvements. The store’s a little better, the app’s a little better, the communication’s a little better, operational improvements, and every now and then, one of these use cases, or usually a combination of new use cases, results in a new business model being created. Bike sharing. Bike sharing was a new use case, a new business model, a new strategy. And financial, what they’re doing with banks is really a new business model. A lot of what’s going on, one type of business model we talk a lot about in this class is platform business models. That’s Alibaba, that’s Tencent, that’s Facebook, that’s all of these. So in my mind, I always keep this at sort of two different levels. Level one is just operational marathon. Lots of new tools, lots of new use cases, and if you’re running a business, you just, you know, you’re continually improving your operations. You know, you’re refurbishing your restaurants, you’re updating your menu, you’re doing a little bit new type of marketing, digital marketing, you’re training your people a little better. That’s just operational marathon. People call that the sort of operational activities, the operational models. Warren Buffett would use the phrase, this is the jockey on top of the horse. You know, that we’re talking about, look, I’ve got a really good jockey, my management team is really good, we’re doing lots of operational changes all the time, we’re improving the scale, we’re improving our efficiencies, we’re improving our effectiveness, we’re managing our supply chain better. That’s all operational activity. And all businesses, regardless of how strong you are, you’re in an operational marathon all the time. And AI tools, software tools, digital tools, are just creating new ways of doing that all the time. That is different than a strategy question, a business model. That’s sort of the second level. So level one would just be operational stuff, operational marathons, chaos all the time, lots of digital tools. Level two, the level above that would be when you create an entirely new business model or you create a business model that has structural advantages relative to your competitor, that your retail chain can do things as the Alibaba new retail fresh hippo supermarket that a traditional business model in supermarkets can’t do. That to me is a business model change. It’s a It’s a competitive advantage. It’s a we’re doing activities that the other supermarkets just can’t do. So I always look for when one of these tools emerges that doesn’t just create an operational improvement, but it creates a structural improvement, a structural advantage. That’s the difference between the jockey and the horse. We have a great jockey, our management team is good, but the other company’s horse is just faster. And if they have a faster horse, it doesn’t really matter how good your jockey is. Now, ideally you want both. You want a faster horse and an outstanding jockey. You want both. That’s what Warren Buffett looks for. But if we’ve all got the same type of horse, then it really comes down to who’s got the better jockey. So I’m looking for a case, look, is this a better horse? Is this a better business model? Is this a better strategy? So that second level, it’s a strategy question. It’s a business model question. The first level is more about… management performance, operational improvements, operational scale, all that. I always break it into two levels in my mind and Buffett pretty much does the same thing as far as I can tell and he just calls it horse versus jockey. And generally as a rule in life, if you’re a great jockey, if you’re a great management team and you’re on a weak business model, you’re not going to win. If you have a better business model, a better horse, a mediocre jockey will usually win. And those two levels, that’s how I think about it. And people use different languages, the horse, the jockey thing, the business model versus the operational model and so on. When I look at digital tools like AI coming out, I use my little digital superpowers list, which those of you who do this know very well because I talk about it all the time. That’s just my short list. If I see a digital tool that I think gives a company a digital superpower, business model improvement that makes them, they put some in a position, a competitive advantage, something that gives them an advantage that others just can’t compete with. That to me is a faster horse. So I have six of those and number one would be, does it dramatically improve the user experience? Number two, does it create a… Platform business model number three does it create a network effect number four does it create a competitive advantage? Number five does it get you virality as a sales mechanism number six does it get you you know dramatic scale improvements Those are my six My list is on my website. I talk about them all the time. That’s my that’s my cheat sheet You know I hear about ooh this this group is doing AI and something something okay I run the list in my head does that give them a superpower? because if it does, their competitors are in trouble. Okay, fine, that’s concept number one for today is the six digital superpowers. Underneath that, this idea of, okay, even if you have a superpower or if you don’t, most companies don’t, you’re in an operational marathon. You’re always trying to improve day to day, day to day. It’s like running in a marathon, you’re in a big pack, you run faster than your… your other runners and slowly over time, hopefully year after year, you pull away from the pack in terms of your operational capabilities, like KFC, right? They have the Pizza Hut, they have the KFC, this is Yum China. Now they’ve got about 9,000 outlets now. That is dramatically ahead of say McDonald’s, which is more like, I forget the number, like four to 5,000 outlets in China now. And dramatically ahead of other companies like Starbucks, which are two to 3,000. I mean, They have just pulled away from the pack over two to three decades because they’ve ran faster in the marathon every day. And that’s kind of the number two idea for today is operational marathons. Now what I’m arguing, which I haven’t written about yet, but I’ve taught in my classes and I’ll start doing cases on this, is that there’s basically five dimensions of operational marathons and digital is changing this. And I’ll talk about that later. Don’t worry about that for today. So those are sort of the two ideas, is superpowers versus business, five dimensions of operational marathons. So I was talking with Tao Mei, Dr. Tao Mei, he’s got a PhD, about what they’re doing with AI in fashion at JD. This was kind of what I was thinking in my head, was is this an operational improvement or is this a superpower? Are they changing the business model? Is it a game changer? Faster horse versus, okay, you’re just running along is a good management jockey. And out of the conversation, I really came up with sort of three takeaways that I thought what they were doing within AI in fashion, AI driven fashion, technology driven fashion. That there were really three things that got my attention in this regard. So that’s what I’m going to talk about today. So first a little bit about Dr. Tao Mei. He’s a nice man. I hadn’t known him before. We sort of did a Zoom talk. which JD was kind enough to set up for me. Really appreciated that, thank you very much. And he’s the vice president of cloud and AI at JD and he’s also his other title according to his LinkedIn page is technical vice president of JD.com. Okay, so let’s, senior AI thinker at JD is mentally how I thought about this. He’s been there a couple of years. Prior to that, he was a senior research manager at Microsoft Research Asia in Beijing. You will hear this research center all the time in China. When you look at people who are founding a lot of the AI champions of China, you look at their backgrounds and Microsoft Research Asia, you’ll see it all over the place. This seems to be sort of the ground zero for AI training in China in many ways. It was not a big surprise. I think he spent about six or seven years there. The title on his CV is Senior Research Manager. Prior to that, he has a PhD from the University of Science and Technology of China, which is in Hefei, and basically pattern recognition and intelligence systems. OK, so he looks like an AI guy who then went to one of the best places in China and then went to JD, where e-commerce. Obviously, that’s another place to do. great AI in China. It’s one of the biggest places. All right. The three takeaways I kind of had from our conversation, I’ll go into a bit. For those of you who are subscribers, I’m going to post some of the slides he shared with me there. So you’ll be getting those in the next, hopefully this week I’ll send them to you. But look, this combination of AI and fashion is just fantastic. It really, it hits right into the area where AI, really does in fact have a lot of power and effectiveness. And it’s very bad for models, fashion models in particular. But AI does certain things well. One of the ideas for today is, look, AI is cheap prediction. Whenever anyone tells me AI machine learning, mentally in my head, I think cheap prediction. And we did a podcast on this that, you know. Transistors made calculations very, very cheap. So suddenly you could put calculations into everything. You could put them into calculators, you could put them into back offices of companies, you could put them in accounting. You could start to put them into other places in the world like that we didn’t really see coming like photography, which used to be done by chemistry. Suddenly you could do it by calculation because it’s digital photos, we got digital cameras. That was a bit of a surprise. And then it moved into things like Facebook and Instagram, which are all about sharing digital photos. So once you make something cheap, we use a lot more of it in common areas. And then we start to apply it to other areas we haven’t seen. And then according to, you know, this whole idea of AI as prediction machines, which is a good book, you know, then we start to see the value of complements rise and the value of substitutes to fall. That’s kind of business thinking on this. If that didn’t make sense to you, go back to podcast 28, where we talk a lot about this. Okay, within AI as a form of cheap prediction, you know, there’s a lot of AI technologies. There’s machine learning, there’s computer vision, there’s natural language processing, there’s speech recognition, there’s tech to speech. Some of that works okay. Some of it doesn’t work terribly well. But certain things really do work well. And fashion manages to leverage a lot of the areas that AI is actually very, very effective. In particular, AI is very good at computer vision. It turns out AI can look at cameras, whether it’s a photo or a video, and it can really understand what it sees. Well, there’s no actual understanding. AI doesn’t understand anything. but it’s very good at classification and detection of things within videos. It is very good at segmentation and parsing. It is very good at content generation. And so you can do things like parsing. Parsing is, for those of you who are subscribers, I will send you slides on this. It’s a lot easier to see this stuff because it’s visual than me talking about it on a podcast. Parsing is you put someone in some clothes, you take a picture of it or a video of it and the AI parses it into certain things like, you know, this is a shirt, these are pants, these are socks. That relates to sort of detection, parsing and detection are words you hear a lot when people start to talk about when computer vision starts to be applied to fashion, detection and parsing is looking at a photo and saying these are high heels, this is a skirt, this is a hat. and it can start to put these into different categories within the software. And then it can start to do classification. Classification is when it looks at something it has detected and parsed like a shirt or socks, and it starts to classify that image as, this hat is urban, this is a lady, this is a lady in an office. This is a very neutral pattern of a shirt. This is a very cold style. This is a very warm style. Once you start to detect and parse things, then what you wanna do is start to classify them because all of this stuff has to be classified and tagged. Urban, sort of vibrant pattern, office setting, casual setting, dinner party setting, neutral, warm, cold. It starts to put. text words against the various things it has detected. And based on those classifications, then it can start to sort of do what they call compatibility. We can associate these types of shoes that we described as urban and office based with a skirt that is also classified as urban and office based. So it can start to make these compatibility linkages. The other thing we talk about is a lot of search. Once you have those things done, compatibility, classification, detection, and parsing, well then you can start to do searches. Instead of searching, I can search by typing in something. I can type in, I am looking for urban professional clothing, and it will show me things because they’ve all been classified this way. Or I can search visually. I can… put my phone on something, the camera can look in an item like a pair of shoes, it will recognize the item, it will detect it, it will classify it, and then it will search based on those classifications and show me other types of shoes that they sell that are similar to the one I’m looking at in my camera. And then the last bit would be generation, content generation. Not only can I search for things, it can start to generate entirely two new types of fashion. based on these classifications. So think about those sort of as when computer vision meets fashion, and this is from a Taomei slide, parsing, detection, classification, compatibility, search based on image, text, and other things, and then content generation. Those are kind of the categories to think about. So let me give you an example. An example that we talked about would be photo-detect. You go online, you’re going to do some fashion shopping, you’re looking for shoes, you’re looking at dresses, whatever you’re looking at. All of those photos or videos will have text beneath them. This is a black skirt for office wear for urban professionals on the go. That might be the text you would see under the photo of a particular skirt. And the AI is starting to get better and better at doing that sort of automatic generation of descriptions. which is really important in fashion, right? You see photos and you wanna know what it’s about. So you can take a picture of, here’s an example from JD AI research. You’re taking as a photo of some zebras on the savanna with a rainbow. And the text that would be generated by the AI would be quote, a group of zebras gazing in a field with a rainbow in the sky, right? So that would be sort of text generation. Well, that turns out to be very, very helpful in fashion because people are always putting up new photos of new types of products, videos, and then underneath that you have the text. Okay, AI is getting pretty decent at doing that. Another type of sort of content generation you could do would be you would take a picture of a model wearing, let’s say, blue pants and a pink shirt. you know, long sleeve, V neck, something like that. And then you would maybe put an image like a Van Gogh painting, a bright pattern, anything you want. And it would combine the image with this sort of clothing type and start to generate 20 to 30 pictures of the same model wearing a long sleeve shirt and pants. But now the patterns on the pants, are the patterns we suggested. So it can start to do content generation, which would be photos, but in various fashion styles. And in theory, you could buy any of those. And it would also then generate text to go along with that pattern. So they would call that content stylization, I think is the term they used, clothing stylization. Now you can do more than that. You can actually take a picture of a model. let’s say standing with a little bit turned to the left, arms at the side, maybe one knee bent, wearing whatever clothes the model was wearing in the photo shoot. And then they can start to create lots and lots of images with the same posture, same face, same model, but wearing all different types of clothing, different colors, different styles. Maybe these are styles from other… fashion lines or maybe these are entirely new fashion styles that are being generated by the AI and from that they generate 50 photos that all go up online and then the AI not only generated those photos but also generates the text that goes along with them describing each one. It gets even more interesting. They can start to change the posture of the model from looking from the left to looking from the right to moving this arm. they can start to change the posture as well. So suddenly this is why I said, look, this is really bad for fashion models because you can pretty much take one or two photos of a fashion model wearing one outfit and then just generate maybe not infinite, but a whole lot of different styles of fashions, of clothes, of postures, of all of that. They all go up on the e-commerce website for this fashion brand and then the text is generated and suddenly all of that is purchasable. searchable, they all get tagged with different classifiers. You can search by text, you can search by taking a photo of something, scanning your phone at an item you see in a store, and it will start to show you lots of items that you can buy like that. So all of this is pretty cool in terms of fashion and how AI can do that. So, I mean, Tommy brought up quite a lot in this area, this idea that content generation, for product descriptions, for photo descriptions, is obviously very useful. Content generations of various photos of models in different types of boots or whatever is pretty interesting. But then it can also, you know, it can go one step beyond that of just creating new styles and new fashions by combining different types of patterns and clothes and all of this. So there’s a lot going on in sort of style generation, content generation, content text generation. Another thing we talked about was virtual try-on. Okay, so you see all these patterns on the screen, all these different types of clothes, boots, clothes, whatever, maybe showing on models. Well, then it can start to apply that to you. It can look at your photo. Maybe you walked into a store, a Gucci store, a Prada store, and it scans you as you come into the store because they have a smart screen and it puts you up on the screen. You can then virtually try on all of these things on your phone, in the store. It can start to show boots, how they would look on you. It can suggest different types of clothing that would go with that. You can do it with makeup and other things as well. But I think for fashion, we’re mostly talking about clothing. So there’s virtual try-on, there’s content generation. You can deploy it on a web app. You can deploy it within somebody’s online store on JD. You can apply it within a physical store in the mall. Sunglasses, shoes, all of this. I mean, it’s pretty awesome. And it all works pretty well, because it turns out AI, which is basically just doing lots of cheap prediction that doesn’t cost you anything. It can predict what type of clothes might be popular, what kind that you might look good in. Turns out AI is really good at that. It’s not as good with natural language processing and a lot of things, but it is very good with computer vision and visual things. So that was kind of bucket number one. that AI-driven fashion is just great and it works really well today. It’s very commercializable right now. And it’s pretty bad news for fashion models across the board. So that’s kind of takeaway number one. All right, takeaway number two was we’ll call this fashion trend and analysis. By the way, a lot of this, what I’m saying, these are my opinions. These are not Thao May’s. This is a lot of my language. So this is not JD or Tao Mai saying any of this. This is me, my interpretation, which is not the same thing. But number two would be fashion trend and analysis. Okay, let’s say we start to gather data about what people are wearing. You know, one of the tricks with fashion is always you’re trying to predict the next trend. What are people gonna be wearing in Beijing this winter, as opposed to last winter? Because these things do change, like. like athleisure, like sportswear, has been incredibly popular in China in the last couple years, and it wasn’t before. It’s just suddenly like we went out one summer and everyone was walking around in yoga pants and everyone was walking around in sneakers and everyone looked like they were on their way to the gym. I can literally, I remember doing this. I was sitting on the Beijing subway and I just started counting like how many people on the subway cart are wearing sweatpants. And like 30% of the people were wearing sweatpants. And how many are wearing like athletic shoes, like trainers and like 70%. And that was not at all the case one year before. And all the billboards around the subway were like women boxing in stretchy pants and guys running on the street. That was all the billboards. And it kind of changed very quickly. And so you see these fashion trends happening all the time. Oh, now people are wearing Ugg shoes. They weren’t wearing Ugg shoes last year. Now they’re all wearing scars, they weren’t wearing scar. Well, predicting these fashion trends is something that the major brands spend a lot of time doing. If you catch a trend, then you can make a lot of money. If you don’t catch a trend, a lot of your inventory can decrease in value month by month by month. This is one of the big problems. If you’re a Zara store or a Gucci store or Christian Dior, you have to sell this. seasons items now, because every month that goes by, they’re sitting in your inventory, they are decreasing in value, and suddenly they’re in the discount bin four months later, and you’re writing off the inventory or pushing it to a flash sale site like VIP shop within a year. So sort of managing these trends, planning your inventory, doing demand projection, and then also sort of designing your products. is incredibly important in terms of getting your inventory, but also in terms of not having a lot of inventory wastage. So they spend a lot of time doing this and fast fashion companies like Zara are very good at doing this quickly. Okay, how can you do fashion trend analysis and prediction, which is what a lot of MBAs do when they work at companies like Gucci and Prada? Well, I just kind of said, look, AI is really good at prediction. which is something that MBAs do for a lot more money and AI does it pretty cheap. Okay, this is something that we talked about that like you can go on social media, you can go on Instagram, you can go on Facebook or WeChat, and you can have AI just scanning what people are posting all the time and looking what’s in their photo and using all these tools I just mentioned about classification and such. and seeing that, oh, look, we’re seeing a lot of people posting pictures of boots. That might be a trend. And you can actually kind of do it, depending what kind of access you have to what people are posting, what they’re wearing on the streets based on the fact there’s lots of video cameras everywhere. And you can start to develop sort of a heat map for what is trending city by city around the world day by day. So it gives you sort of a virtual real time tracking. Here’s what’s trending on Instagram, on Facebook, on WeChat, on WhatsApp, in terms of what people are putting in their photos and what they’re wearing in Warsaw in March. And here’s what they’re wearing in Miami, and here’s what they’re wearing in Rio de Janeiro, and here’s what they’re wearing in Paris, because you can locate, geolocate all of this stuff. And you can get sort of a world heat map. for fashion trends. That would be the analysis aspect of this because AI is very good at computer vision. Well, then it’s just a half step to, well, if this is what’s trending, let’s have the AI start to make predictions of what items might sell and what items we should design and what items we should market and what we should put in our inventory right now based on trending leading to prediction. and it can happen sort of in real time. So that’s a pretty cool number sort of to take away. And Taumei talked about this, what AI can see in terms of what people are wearing around the world in real time. It can collect the photos from social media. It can see what people are posting. It can make predictions based on fashion trends, fashion shows, what’s on television, what’s not on television, what’s on the street, what’s at parties around the world. and then do a prediction. So data visibility becomes first, but then prediction comes next and bam, you can start generating all sorts of new types of content and seeing if people buy. So that was sort of number two that I was like, wow, that’s pretty impressive. And getting back to this idea of digital superpowers, if you are a retailer or a merchant and you’re able to do that and your competitor isn’t, That strikes me as a big deal. It’s definitely an operational improvement, an operational marathon, but it also could be a superpower. Okay, third takeaway from the conversation is basically fashion APIs. APIs is when you have a software and your tools and you open them up to other people. Because ultimately, look, JD doesn’t wanna be in the, we’re the fashion company. They’re not a fashion company. They are a platform business model that creates tools and data for merchants, brands, and other retailers such that they come to their platform and start to do things. So their goal is not to have all this secret technology and use it themselves, although they will do that, it’s to open it up to everybody. So they have their sort of fashion APIs, which they have at New Hub, and they’re opening all this up to everybody. So they’re opening it up to any merchant or brand that comes on their website. You can start to put this into your own online store. You can start to use it. If you’re a merchant, brand, retailer, or an application developer where you’re creating your own software tools, you can start to access all of these things that they’re creating, their tools and the data. And, you know, suddenly you’re a small merchant on JD, you can start to do your own content creation from photos. You can start to do your own clothing stylization. You can start to offer to your customers virtual try-ons. You can do fashion detection and parsing. You can do all of this and because it’s software, which is why software is amazing, you can deploy it to everybody. If you have your own physical stores, if you’re a barber shop, if you’re a little fashion retailer on the side street of Beijing, you can start to offer virtual try-on in your physical stores. You can put up your smart screens, you can give it to people on your app, things like that. And Tal mentioned that, Tal may mention they’ve deployed this, they’re getting over one million daily active users in terms of all of this on their sites. That’s pretty amazing. So you give it away to everyone. And here’s some of the fashion APIs that they list at New Hub, which are product detection, product classification, product recognition, product search by image. brand logo detection, brand logo recognition, brand sketch recognition, sketch based product retrieval, fashion parsing, fashion category classification, fashion attributes recognition, fashion compatibility. So they’re just starting to give these tools to everybody. That’s a pretty big deal. You could see that this is gonna become pretty common in a lot of the fashion world very, very quickly. and companies like JD are just gonna have dramatically better tools, probably than any individual merchant or brand has themselves, and probably better than any platform business model has, other than probably Alibaba, and maybe some of the biggest brands, because they have such scale to do this. That it would be very hard for other companies to do this, even other platform business models in places like Thailand or whatever. because they’re not gonna have the scale of a massive Chinese platform business model or maybe an Amazon in the West. Scale matters, data access matters. So we could see very quickly that a company like JD could become one of the leaders in fashion APIs, which I think will probably happen. Okay, so those are kind of my takeaways from my discussion was, look. AI driven fashion is pretty amazing. It really does work. It’s not theoretical. It’s happening right now. It works. And bad news for fashion models. Number two, fashion trend analysis and prediction is really impressive because fashion trend, fashion trend analysis is really about demand prediction and inventory management and what you have in your supply chain and what you’re creating and how you’re moving things around your warehouses. If you can get better at demand projection, that’s gonna give you a lot of benefits in a lot of parts of your business and retail. And then the third one is fashion APIs are gonna be great. JD is clearly gonna be a leader in this, along with a small handful of other companies on the planet. And I would put that all within for the concepts today. Think about those things in terms of one, digital superpowers, that was one of your concepts for today. within the sort of different dimensions of an operational marathon, another one of your concepts for today, AI as a form of cheap prediction, and then AI factories and human free operations. So that all fits right within sort of the learning goals for today. All right, I think that’s most of what I wanted to go through today. I’ve been running long in the last couple of weeks. I’m sorry about that. So we’ll keep it a little bit short today. Last comment, I guess, on this is, if you pull up… out of these sort of tools and APIs, you know, sort of higher level looking at, at e-commerce, retail, fashion. You know, obviously the biggest benefit for all of these digital tools, as far as I can tell, is just gonna be sort of increasing personalization, increasing the two-way communication and engagement with customers, which in this case is consumers, and just continually making the experience better. and more interesting. And it’s not just about finding a lot of goods cheap. It’s about the experience of what happens when you walk into a store and what happens when you use the app online and just making that better and better and better. And there’s so many benefits to just improving the consumer experience. And within that personalization is probably the biggest lever. that every store you walk into, they’re gonna personalize the products to you, the inventory to you, the experience to you, what they show you. You know, that’s all digital. We’re used to that happening online. It’s gonna happen more and more in the physical world, but just excellence of experience. That it’s just fun to shop and it’s fun to do this stuff. And then occasionally you buy stuff. The other big lever is clearly logistics and just managing the supply chain better. and getting your inventory better and more tightly tying your inventory and your logistics activities to the consumer demand that you’re better and better able to project what people are gonna buy tomorrow and what they’re gonna buy next week and what they’re gonna buy in June when the new season comes out. The better insight you have into that, the better you’re gonna be able to manage your logistics and your inventory. And then, maybe the third bit here is you’re also starting to connect the pieces from consumer demand to delivery to logistics and then to manufacturers. We want to get all of these parties connected, looking at the same data and just sort of close the loop in connectivity and data where it’s not just that the manufacturers are more efficiently creating what’s going to go into the warehouses such that next week when people start buying more Uggs, because we know that that’s how the trend is moving, that we’re gonna have those ready, but also increasingly to personalize the experience and suddenly manufacturers start creating new and better things day by day, because they can also see what consumers want. Not just in the aggregate, but personally, what you want suddenly, this Nike manufacturer. is gonna start offering you shoes that are just for you, that are custom fitted, sort of closing the loop in data and connectivity, in visibility and prediction between consumers, retail, logistics and manufacturing, that it’s one ecosystem. And that’s something we’re seeing happening in China far more than anywhere else on the planet. I mean, that’s the frontier and JD is there. So it’s pretty awesome. It also helps that most of the manufacturing of products in fashion happens in Asia. There are not a lot of apparel manufacturers in the United States. So that is kind of an Asia story anyways. That is the content for today. So for those of you who are subscribers, all of this goes under Learning Goal 22, which is the basics of AI, which… Fashion plus AI is pretty awesome, generally speaking. Your four concepts for today, digital superpowers, the five dimensions of operational marathons, AI is cheap prediction, and AI factories and human-free operations. That’s all in the show notes. It’s all on the webpage. It all goes out on the website. For those of you who aren’t subscribers, please sign up. It’s a lot of fun. We’re having a good time. For all of you that have signed up in the last two weeks, really, Welcome to the class. Thank you for doing that. It really means a lot to me that you’re giving it a try. It’s really gratifying to see the numbers pop. I mean, the chart was pretty amazing. It was like this steady increase week after week, and then it just spiked. I mean, literally the chart went vertical two and a half weeks ago. So that was really satisfying. It means a lot to me. Thank you for that. I do appreciate it. I don’t really have much to say here in Thailand. The COVID situation, I guess, is officially over here. We are all allowed to travel now. And I’ve been teaching here to a business school here, which has been a lot of fun, mostly Thai students, about half executives, although the executives are a bit from China and outside of Thailand. MBAs are mostly from Thailand, just a great group of people. Really wonderful to talk to and really satisfying. I just had a… a good conversation this morning with a young woman who owns five or six beauty salons in Bangkok. And we just talked about what does that mean to be a local service business versus, you know, selling products online and that the influencer economy that we’re so comfortable with in China, live streaming, TikTok, short video is coming here now. Lazada and Alibaba are doing a big push to move the live streaming influencer economy out of China. to the rest of the world and Southeast Asia is destination number one for them. And I was like, dude, you should start doing live streaming about haircare and how to do that. And people care about that. And there’s a lot of digital tools that have already been created. They’re coming out of China, they’re expanding to Southeast Asia very quickly, very well positioned company. If you’re a local service business in an area like haircare, which people really care about. and you combine that with more content and community and live streaming and influencer, you can really do a lot of great stuff. I mean, that’s a great place to be. So it was pretty exciting. It was a fun conversation. So that was my morning. And then in a stupid move, I bought Call of Duty last night at 2 a.m. because I couldn’t sleep. I have a problem with insomnia where some days I just can’t sleep. And it’s really frustrating because you can’t sleep, but you’re tired, right? That’s the problem with insomnia. You know, in theory you can get a lot done because you can’t sleep, but you’re also, you’re physically tired. So it’s hard to do things like thinking and, you know, you know, reading and, you know, writing. It’s hard to do that when you’re sort of physically exhausted. So I couldn’t do that at 2 a.m. this morning. So I ended up downloading on the PS4, Call of Duty, a modern warfare, which is really an awful idea because that’s just going to kill my productivity for the next two weeks because it’s just such a great game. And so. Anyways, I’m finishing with my teaching this week here, and then I’m gonna head down to one of the islands, probably Koh Phag Nhan, down in the south of Thailand, which I think is open now. Sit on the beach for a good week and a half, and write and think, and the productivity of that idea probably just took a major hit, because now I’ve got call of duty. Although I don’t have it there, so maybe that’ll help me out. Maybe the way to counteract that is to get out of the apartment. and go somewhere where I don’t have my PS4. So anyways, these are the things I think about at 2 a.m. walking around Bangkok. If you were out at Bangkok at 2 a.m. this morning, you probably saw me walking around trying to get over my insomnia. But anyways, that’s my week. So thank you for listening, and I will talk to you next week. Cheers from Bangkok.