The Role of AI at the Edge with Andi Huels

Elevate the Edge hosts Jo and Maribel discuss the Role of AI at the Edge with Lenovo's Andi Huels.

Transcript

Maribel Lopez  0:00  
Hello and welcome back to the podcast. I'm Maribel Lopez and I'm joined here today with my fabulous co host Jo Peterson. Hey, Jo.

Jo Peterson  0:08  
Hey Maribel.

Maribel Lopez  0:10  
And as always, we're excited to be joined here today by another great technology guest we're joined with Andi Huels and she leads Lenovo is an enterprise AI business in North America. She is the go to market executive with experience in artificial intelligence, machine learning data analytics. Andi advises global and fortune 500 companies on how to drive agility and reinvent their operations with AI, which I think we can all appreciate is something that's really, really important these days. She's been working in AI solutions with senior executives and retail QSR CPG, supply chain manufacturing, we could go on and I'm really excited to hear that she is also recently been awarded and Gattaca is top 50 women in tech. So wonderful to have you on the program and be excited to talk to you about the edge and AI. 

Andi Huels  1:11  
Well, thank you for inviting me to be part of elevate the edge.

Maribel Lopez  1:15  
So Andi, there are a lot of definitions kind of rolling around in the industry right now. And I was wondering if he wanted to spend a minute helping us level set? And then what is edge AI? And how does it work?

Andi Huels  1:27  
Sure, absolutely. That's a question that I frequently get. What do we mean by the edge is at the edge of my desk, the edge of the universe? What are we talking about? Ai server is actually a very small device, it's not much bigger than a laptop. And inside of an edge AI server you have several components that make it edge AI, the first of which is typically a GPU, you also always have a CPU. And the secret sauce is your independent software vendor. So whatever that AI software is, it gives us a unique use case. And Lenovo is portion of this is we actually manufacture those edge AI servers. Now what do I mean by the edge simply means that you're on prem. So rather than processing all of the data in the cloud, you're keeping it on premises near the source of that data.

Maribel Lopez  2:23  
Yeah, we've we've shown I've been talking to a lot of people about the edge. And we're always amazed at how many different definitions of the edge we get, and how many different things are included in the edge. So it's great that you already spoke a little bit about the what is an edge AI server? And that was going to be a question that either Jo or I were going to ask you, but Jo, I'm gonna pass it back over to you. And I know you had a question that you want to talk about in the analytics field.

Jo Peterson  2:53  
Yeah, I do. Well, but before we go there, I wanted to see from Andi, if we could talk about, you know, we know what an edge AI server is, but how is that different than some of the other devices that a customer is going to have in their environment? What do you think, Andi?

Andi Huels  3:10  
Yeah, so what is doing is actually processing that data and deriving insights in microseconds. Rather than sending all the data collected by IoT sensors directly to the cloud edge computing processes, this data within the network, and only relevant data is sent. And that has a lot of benefits. Would you like me to go through those?

Jo Peterson  3:32  
Well, yo, I actually I was going to ask you about, you know, the benefits of that instant data, and how it can help companies differentiate them in the marketplace? Because I'm sure as you in your in your travels, you get to see some of that.

Andi Huels  3:48  
Yeah, absolutely. So I'll give you a couple of examples. I'm sure a lot of you have been to say a Kroger grocery store, right? Well, self checkout is actually edge AI. And if you think about if you're a red customer versus a green customer red customer is, is out to steal or to, you know, malign the grocer, whereas a green customer is just simply, you know, not being able to scan their bag of frozen peas, right. And in the case of a red customer, if we're sending that to the cloud, and then waiting for it to come back down that information, you'll already be in the parking lot. So latency is that concept. It's we can't have a delay in the processing of the data. Another example would be McDonald's drive thru. We do natural language processing voice to text in the drive thru at McDonald's. Well think about it if you're placing your order, but it needs to go to the cloud and then come back down. That's going to actually make the drive thru speed longer. And instead because we're using edge AI, we've taken 30 seconds off of the time it typically takes to go through a McDonald's drive thru. So latency is one area that's very important that that's a reason you need to be on prem. Another is simply security, right? Think of all of that data that we're processing, whether it's intelligent store analytics, and where you're walking in the store, and what you're picking up and looking at, and so forth. That's private. And we don't want that going up to the cloud, or even leaving the building. So that type of information, that's another reason why we want to use an edge AI device on prem. Another would be communications, I mean, think about your own home, when your internet goes down, that simply can't happen. Right? So we use edge AI servers in use cases where it's absolutely imperative that our communication is 24/7. Always on, like self checkout, again, we can't hope that the internet is always on. So those are some of the reasons that you want to use an edge AI server.

Jo Peterson  5:56  
Oh, those are good reasons. I you know, I'm going to date myself a little bit here. But you're saying that the Wayne's World donut example for goodness there's you know

Andi Huels  6:17  
the combination of computer vision

Maribel Lopez  6:24  
OK, Wayne's world. Wow, that is the Wayback Machine and kind of funny to have that in Edge AI in the same conversation. But he right. So we've talked a little bit about the edge AI server, we've talked a little bit about where you know why we want to have things on the Prem, what are some of the characteristics around that around latency security? Sometimes it's caused, a lot of times it is also around data residency and other things. And you gave us a couple of examples as well. But I thought maybe we could dig into that a little more, because I know that you're doing so many interesting things. And can you share maybe a few more really common edge AI use cases? And maybe some that we have not heard about yet?

Andi Huels  7:13  
Sure, absolutely. Another example that's becoming more common, you probably see it when you're in an airport is autonomous shopping. So this is similar to checkout, except there's absolutely no, you know, human involved, you're going to scan your phone and take your items, leave the store, and then within a few minutes, you're going to receive a receipt. Now that is absolutely critical that you know you have an edge AI server because otherwise literally people just leave the store, and and take whatever they'd like. So that's, you're gonna see that growing a lot more. We talked about, we talked about self checkout. So that's a very prevalent one that you see a lot. Another is one of my partners radius AI is doing intelligent store analytics. So we think about where do people go in the store? What do they look at? And this type of information is absolutely priceless, right? Think about how many people are in line at the store, that helps us solve the issue of of labor, because instead of guessing when we need to have employees, we know especially after you collect a year's worth of data, you know exactly your busiest times and when you don't need as many staff think about, you know, lot or pump analytics at a convenience store. Right? How many people that pump gas actually come in and buy anything? How many use the restroom? How many buy one product versus another product, you don't even have to know who the person is or see their face, you can use their gaze to know what what their analytics are, right? So that's very, very popular now because stores are realizing, we want to know when we when we have customers and what they're doing and what they're looking at. It's valuable information. Another area that's becoming very hot topic for retailers and restaurants is safety. It used to be theft. We used to have the security guards that were always watching you to make sure you didn't steal, which is it's still an issue. But now the bigger issue for retailers and restaurants is the employee and customer safety. So what are some things that we can do with AI at the edge that can increase safety? Well, one of those is facial matching technology. So Maribel has just robbed a convenience store down the street. Jo can then if she has the facial matching technology can then match Maribel so that she's not going to be able to come into another store and Rob as well. So known felons and so forth can be detected and so you can really increase safety that way. We also have abrash of organized crime, where groups of say 10 will come in at the same time I'll scoop up product and, and rush out. But with computer vision and edge AI, you can actually detect when 10 People are coming together in a pack moving quickly throughout the store. And some of the the gestures think about, you know, hands up in the air is not something you typically do when you're in a store lying on the floor is not, you know, those type of gestures can immediately be we can alert the authorities. So that's another one. And also think about your highest value products like say, Tide Pods. Now if Maribel is in Kroger, and she's filling her entire cart with Tide Pods that might give Jo the store manager something to, you know, wonder why in the world does she need that many, maybe she's going to steal them. So what we can do with computer vision and ngi is watch those high value products. And when someone's filling their entire cart with those, we can lock the wheels of the cart, right so that she doesn't pose a danger of flying through the store or trying to steal that entire cartload. So those are just some of the the areas that are kind of interesting for the for right now. And in the future of NTi.

Maribel Lopez  11:06  
Yeah, what I what I love about the intersection of these two technologies in one sense, you've got basically rapid analytics happening for for like a more late layman's term on this. And on the second hand, you have just the concept of taking vision not even tied to somebody if we take the Public Safety and Security out of it. But really being able to understand various store environments and having a real understanding of the experience that's happening. What are the things that are impacting the experience? Even understanding if people like how how people in general behave? You know, they they pick up items and put them down? How often do they pick them up and actually buy them versus just putting them back down. And that can actually tell things to the buyer without even needing to know if it's Maribel or Jo, that you know, these products are resonating, these products aren't resonating. Or people just don't like the store because the layout differs on the other stores. So they're not actually navigating through the store in a right way? Or you're creating a new fan experience in a stadium. Right, which is another thing? How do we create more personalized experiences to different people? So I think that there's a lot of opportunity at the intersection of looking at computer vision with you know, high performance computing, right?

Andi Huels  12:33  
Yeah, realize that they collect all that data online. You know, when I'm shopping online, they can see everything. And then market to me, Well, how do they do that when they don't have that data from the you know, the actual retail experience. So now they're realizing they can use edge AI and computer vision to collect that same type of data. And, and market more effectively, a lot of retailers also sell their shelf space. So if you know that the Maybelline has, you know, so much time, you know, people look at it, and then they look at the cover girl, and then they buy the Maybelline, you can sell that data back to Maybelline or you can charge them more for the shelf space. based on actual data. This is how many people are coming into our store to look at your products.

Jo Peterson  13:21  
This feels, you know, like a great marriage between IT and marketing. As you go into these meetings, Andi, who do you see in these meetings? Or Is that who you see at the table? Where does this sort of idea start in retail organization?

Andi Huels  13:38  
That is a great question. Because working at Lenovo, initially folks at Lenovo were like, we're going to sell to the IT department like we always do. But that is not the case with with Edge AI. Really these these ideas, these, this digital transformation has to come from the top. If you take say target CEO, Brian Cornell, he said the most are gonna spend $4 billion a year on one thing, and that is speed. How is he going to get speed? Right? How is he going to get that type of information quickly, because now we all have this insatiable need for speed, right? And AI is the way that you're going to do it. If you think about Doug McMillon, he said at the CES keynote, and he's the CEO of Walmart. I'm going to focus on three things next year, big data analytics, robotics and artificial intelligence. Right. So when it starts from the top to answer your question, it's usually at the sea level. So usually I'm talking to somebody that's a digital transformation executive, the CTO now we're seeing a lot more chief innovation officers and even some, you know, far fangled titles like the next gen supply chain leader, you know, that type of title but almost never as the IT manager. It's it starts from that then that trickles down.

Jo Peterson  15:01  
And so you bring me to my next question. It seems obvious to me, but I'd love to get your thoughts on it. These folks that are forward thinking like the people that you just mentioned, that realize the need, how much further ahead is it going to put them from their peers, from their competitors, this sort of AI differentiation that you're discussing,

Andi Huels  15:30  
it's really going to give them a quantum leap ahead of their competitors. And right now, you know, when I present to companies, they they're still some of them are not ready, they think that AI is going to be on there, it should be on their, you know, technology roadmap for the future. And that could not be further from the truth, you can, it's very easy to deploy. And as long as you know, all of the ecosystem players, which was what Lenovo brings to, we bring you the ISV, we connect you with Nvidia, Intel, AMD all of the partners, so that, you know, you can implement these use cases. But some of the advantages that those retailers are going to have, we touched on self checkout, for example, decreases customer friction, right, instead of having to stand in line, you're easily able to do it yourself and get out of there quickly. Safety is another huge one. I'd rather shop at a store where I knew there was, you know, known felon detection, because I feel safer in that environment. Another area will be will be just efficiencies, right there, it's going to be the layout of the store is going to be set up to be more efficient for us to get in and out quickly. Because they know how to do that. And you think about a quick service restaurant, we can use edge AI to actually, you can use your voice to act to speak the order. And that's going to speed up the process, you just walk up to the counter your orders already ready. So and also labor, you know, there's a huge labor shortage right now. And so there'll be able to predict when they need more employees so that we're not waiting in line quite as long. So there's all kinds of advantages, and it's just going to absolutely speed up their operations. So it's going to be table stakes in the future. Not a nice to have.

Maribel Lopez  17:19  
I absolutely agree with that. Andi, you know, you're out talking to a lot of organizations, is there something that you think the people that are listening to this podcast should know about edge AI, or something that's counterintuitive, or that you'd like to set the record straight on for those that are listening and wondering what to do next with Edge AI?

Andi Huels  17:42  
Sure, here's a couple of tips. First of all, there's there's a rack server, which is very large, but as large as my wingspan here, and and then the edge AI server, what we don't want to do is stack these devices on top of each other. And when you think especially retail and restaurant environments, they don't have a lot of space, right? And even one edge AI server takes up, you know, it's about the size of a laptop. So what you want to do is you want to think about what use cases do I want to deploy? Because what happens is you tend to deploy one, and then you're like, Wow, that's amazing. And you get this like appetite for AI and you want more. So if you actually if you purchase a server that can do multiple use cases, you can that's always the space you need. And our sc 450 can do up to 14 ai use cases in the same box. Right. So that's one tip is to make sure your your future proofing yourself thinking ahead about what do I want to do with AI. And we can do an AI innovation showcase for you so that you know all the different use cases. And you can pick from those. Another is think about your technology, even our iPhones after three years, you're like I want a new phone, I want the latest technology. The same goes with NS AI servers. So what I highly recommend is that you consider doing an OP X strategy versus capex meaning lease it basically rent the equipment because in three years, the technology will be significantly better. And then you can refresh that equipment at no cost to you. Right. So you'll have the latest and greatest. So those are a couple of tips that I would highly recommend for especially retailers and restaurants.

Jo Peterson  19:30  
They'll end up going great today, we always ask One fun fact to round out and close out the podcast. And it doesn't have to be about tech. So have you got anything what fun fact do you want to share?

Andi Huels  19:43  
Well, you know, you're asking a tech nerd or something fun so I'm gonna have to probably give you a my idea of a fun fact is that by 2025 Edge compute is expected to be four times larger than the cloud and generate 75% of the world's data. So not only is that fun in Andi's mind, but it's also great job security.

Jo Peterson  20:06  
It is it is job security. Well, you've been so fun and so gracious today. Thank you so much for taking time with us. We really appreciate it. And we're excited to see what you do next, Miss Andi.

Andi Huels  20:18  
Well, thank you so much for having me on elevate the edge. I enjoyed meeting you both.

Maribel Lopez  20:24  
Take care. Bye bye

Transcribed by https://otter.ai