(function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start': new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0], j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src= 'https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f); })(window,document,'script','dataLayer','GTM-KX9RFV5L');

Mister Beacon Episode #186

Practical AI Applications with Computer Vision

February 27, 2024

This week on Mr. Beacon, we chat with Marty Beard, the Co-Founder and CEO of alwaysAI, a computer vision company bringing practical AI solutions to business owners. alwaysAI harnesses the power of existing camera infrastructure and AI models to provide real-time, actionable data.

Marty delves into their focus on key verticals such as mining, addressing the imperative need for efficiency in harvesting rare earth minerals essential for electric vehicles and smartphones. We also explore alwaysAI's endeavors in enhancing retail experiences, diving into the intricacies of privacy and union concerns surrounding employee and customer tracking. How does computer vision reconcile these concerns?

Hear about alwaysAI's approach to these concerns as well as the significance of multimodal AI in their operations. How does alwaysAI differ from a large language model like ChatGPT? We go into detail on many of the practical applications for AI in this conversation.

Drawing from his rich background at Oracle and Blackberry, along with a foundation in philosophy, Marty offers insights into strategic decision-making in the ever-evolving tech landscape.

Tune in as we navigate the intersection of technology, strategy, and philosophy with Marty Beard, paving the way for a future shaped by AI and Computer Vision.

Marty’s 3 Favorite songs

Tangled up in Blue, Bob Dylan: https://www.youtube.com/watch?v=QKcNyMBw818

So What, Miles Davis: https://www.youtube.com/watch?v=ylXk1LBvIqU

Grateful Dead (various works)


  • Steve Statler 00:00

    Welcome to the Mr. Beacon podcast. This week, we're delving into the world of Artificial Intelligence and machine vision with the CEO and Co-Founder of alwaysAI. So Marty Beard is an incredible executive, he has seen some amazing things in high tech history. He was vice president of Oracle online, president of Sybase 365, Chairman and CEO of his own Live Ops company, Chief Executive Officer of Blackberry, and is now a Co-Founder and CEO of AWS AI. You don't have to go far to read and hear about how Artificial Intelligence is changing everything. And I've always wondered how it works and what it can be used for outside of Chat GPT. And Marty as some amazing examples of how this has been used in stores and restaurants in minds. And I think it'll really pique your interest, hopefully, stimulate some ideas, and give you a bit more confidence in knowing this particular part of what's happening in the world of digital to physical convergence. The Mr. Beacon Ambient IoT podcast is sponsored by Wiliot, bringing intelligence to every single thing. So Marty, welcome to the podcast. Hey, thank you. So I'd love it. If you can start off just by introducing us to your company. You're the Co-Founder and CEO of always a I want to see a company do.

    Marty Beard 01:50

    Yeah, we're a vision AI company. So what does that mean, we have a software platform that allows companies to build computer vision applications, and there's a lot involved in that process, but build computer vision applications, and they'd actually get them out on the edge running, right? So they can collect data from cameras that are out in the real world that will help them run their business. So we're the we're the software platform that gives you an end to end capability to build computer vision apps, run those apps 24/7 out in the real world, and then get all that data that people are really after to help them make better better business decisions.

    Steve Statler 02:37

    are you building a platform that other people will develop on? Or would you go to General Motors and kind of give them a turnkey solution?

    Marty Beard 02:45

    Yeah, we do we do both? So it's a great question. So in some cases, the customer, they have IT departments, they have developers, and they're going to basically use our platform to create those applications themselves and get running out on their cameras and so forth. Everything the platform enables. That's great, right? It's built for them to just go crazy prototype, build as many apps as you want to get them, get them out onto as many cameras as you want. Other cases, a lot of cases, actually, the customer will say, Hey, I don't really have a whole bunch of folks. But I definitely need to do this, can you guys actually provide the solution. And then in that case, my engineering team will work with the business people, and we'll build the AI model will train the model will develop the application to their specs, will actually get it out running on the cameras and so forth. So it's both.

    Steve Statler 03:42

    And it seems like you probably have one of the classic entrepreneurial problems of you, you probably have some great technology, but it can be used for a million different things. Right. What have you decided to focus on as the Yeah, key use cases?

    Marty Beard 04:00

    Yeah, no, you're right. You're right. Well, it turned out that there's a couple of industries or verticals that are our early adopters, right? So one is the industrial space, and specifically mining and manufacturing, turned out to be a early adopters of computer vision. And you probably wouldn't have picked that if you were building a business from scratch, but it was like, Oh, I get it. You know, if you think about mining, it's, it's under a lot of pressure. Because the you know, we need to get more and more of these metals out of the earth. Why? Because everybody's buying electric vehicles, and we're all using iPhones and, you know, laptops and so forth. And all of that relies on on metals, and but they need to run those mines more efficiently so cameras can show them where they're being inefficient, or where they're having issues or problems and maybe labor constraints or whatever. Also health and safety. Same thing with manufacturing, we work through example with a lumber manufacturer and, and they want to understand how efficient are they cutting lumber? And how safely? are they cutting lumber? And are they getting the kind of output that they need to get? Are they catching anomalies or mistakes before they become a real problem? So industrial is one space that we really focus on a lot. The other one, almost 180 from that is retail. So retail is obsessed with what with customers? How many customers do I have? Where are they going in my store? What brands are they? Are they most focused on? Right? So there's all the customer analytics for a retailer, but also behind the kit, what's called behind the counter. So for example, if you're a restaurant, how efficiently am I making stuff like it? Am I quote manufacturing a sandwich or a pizza as quickly and efficiently as I can. So these are two areas that are really hot. It's all about operational improvement. It's all about getting real time data, like real time insight, so you can act on what's actually actually happening. You can catch problems before they become big issues. And this is where computer vision really thrives.

    Steve Statler 06:12

    How would your product help subway? Monitor the sandwich making process in a way that can be actioned?

    Marty Beard 06:22

    And actually, yeah, yeah, so we do things like, we can actually assess what's called the average traversal time from the very beginning of that process of making the sandwich. To the very end, how long did that take on average. And if you know that number, then you can benchmark different subway stores against each other. You can say, Okay, this store right here, does that in two and a half minutes, and they have really good results, and they sell a lot of sandwiches. This one over here seems they're in the five and a half minute mark, that's twice the other guy why? Like, what's what's going on? Right? There's something going on there. And then they can diagnose it, which they literally do by zone. So one zone might be bread, and the other zone is condiments and the other zone is meats. So they can start to drill down into oh, we're really starting to have a problem right here in this one section. Let's fix it. So that's, that's, that's a simple example. But that's that's a real example. So before that it was human beings trying to figure it out. Or you might look at weekly or monthly data and try to understand what's going on now you're just literally getting in real time on a dashboard. And you can see the dashboard. And you can you can make decisions based on off their real time data.

    Steve Statler 07:43

    I think as an employee, I would have mixed feelings about this, I want to and I feel like, Oh, this is good, because this AI is impartial and dispassionate. And just because, you know, the, you know, I've got my hair partying, while my app binding is a complete mess, but it's gonna judge me objectively. And then the flip side is the obvious Big Brother thing. I'm sure you will have had to deal with that. And I can't imagine that unions love this. But I guess probably a lot of fast food shops. Very few are unionized. So maybe that's not so much of an issue.

    Marty Beard 08:21

    But it's it. Yeah, but but construction sites and mines and others are now it's it's a real? It's a great question. And it's, it's an issue that needs to be addressed. So on the privacy side, let's start with that. You can't identify somebody by name, right? You can't, you can't do that. In fact, in every situations that I've mentioned, and others that I can tell you about, you have to blur faces, right? So you literally as part of the computer vision application, you're not, you're blurring the actual identity of, of a person working there, and you need to protect their, their personal identity, but the camera is identifying, okay, that's a specific individual. And they're here in this zone, and, you know, and so forth. So there is that level of detail. But that's it's an important issue. And the company needs to be upfront about how the data is being used and so forth. Now, on the flip side of that, which is what is interesting and not expected, is that what companies have started to do is there's upside in the sense of, if you use this tool accurately, there's bonuses. There's, you're almost like gamifying, the process, you're kind of like, hey, store A you guys can make a little bit more money than store B because you're performing better. Right? So then there's this awareness that like, okay, I can use that tool to actually potentially improve the performance of the store and maybe that filters down to everybody that's involved in that. But these are real issues that the that the industry is working out. We worked with, did some work with Burger King and in Europe, which has very strong rules and regulations around personal identity and so forth, and at Burger King wanted to know exactly how long it took from the time somebody entered the door, they ordered their meal, they picked up the bag and they left. So again, that amount of that amount of time. But we had to blur like, literally everybody's face and the person is identified just by a number, there's no, you have no idea. That's Marty beard or Steve, or you don't know that it's just, it's just a number. But these are real issues. And these, these are issues that I think AI in general is dealing with, and then computer vision, specifically, it's going to have to do.

    Steve Statler 10:33

    Can you give feedback to an individual employee? You know, there's a there's an AI company that that kind of monitors sales calls and tells people Oh, you're talking too much and try using you know, the best performing people use this word. Can you do the same for sandwich production? Or wood production? Can you be a coach?

    Marty Beard 10:53

    Well, in the, in the case of like, for example, mining if, if you if somebody is is being unsafe, right, which is a real I mean, that's, that's, that's a real issue. This is not like somebody that's going to stub their toe, right? This is like something bad may happen, then, yeah, you absolutely want to take action on that particular person. And you would know that particular person, because the cameras saying there's a person that's not roped in appropriately in that area right there and you go take care of it. Or in the restaurant example I gave earlier, let's say the dashboard is showing a manager that we're really having backup in the bread section. Like for some reason we're having backup in the bread section, you're gonna go to the bread section and talk to the person there and say, what's going on? It may be Oh, I don't have enough bread. Okay. You know, let me let me help you out. So yeah, you can definitely take take action in real time based on yours.

    Steve Statler 11:51

    And you talk a bit more about the retail applications that tends to be a vertical, we cover IoT digital physical convergence, but for some reason, I think it's basically just because I'm super interested in retail. But yeah, what are the other retail applications,

    Marty Beard 12:06

    We have a we do work with Lazy Boy, furniture, outlets, you may have heard of those. And, you know, that's called specialty retail. So we were just talking about restaurants. But now we're into so called specialty retail furniture, stores, hardware stores, you know, things like that. They have a very simple, neat, which is, who's coming in? How many people are coming in? Where are they going in the store? And am I am I assigning the right salesperson to that person that came in? So you would say, oh, couldn't a human being do that some of these stores are enormous, right? And so it's, they want to know, okay, that person came in, they're very interested in in couches, the right salesperson is, is Dave, let's get Dave over there to go meet with that person, etc. And so it's just a little bit more real time kind of ability to mix and match appropriately human beings, you know, the customers of the right salesperson. We're in about 80 of those stores now. And that, and also, obviously, they're understanding who's coming in how many people are coming in. demographics of the people coming in, they're starting to understand a lot more about their customer base and their prospect base. Fascinating. So it's a simple, simple, simple application, but it's, it's sophisticated computer vision, these are deep learning AI models that are deployed out in the facility. And, and, you know, there's a lot of a lot of technology that goes behind that.

    Steve Statler 13:43

    And how are you delivering the prompts the nudges, to the staff to, to look after someone who's been roaming around in the mattress area? Yeah, they have a

    Marty Beard 13:52

    Yeah, they have a dashboard that they're that they're using, the manager can literally see color coded, you know, all that information about how many and where and zones and so forth is kind of captured in a nice in a nice dashboard that is put together by our partner, which is called a company called trek well. So we partner with Traco, which is really focused on that analytics presentation. We're all AI magic behind that, man. It's kind of feeding feeding that and yeah, so it's a dashboard. Just think of an iPad where somebody has it and they can see what's going on.

    Steve Statler 14:32

    And with retail, there's obviously a lot of technology there. But also the buyers are famously unimpressed by technology for technology's sake. Now do you sell to a company like that?

    Marty Beard 14:48

    It's going to be you're absolutely right. In fact, the buyers that would apply to all industries. Yeah, it's you know, retail. Yep, very tight margins and and I don't want anything that gets in the way of efficiency and meeting those margins. You need to demonstrate demo here. This is what it does. And this is how you will get ROI from this application. I guess it's ironic, if you're running your computer vision company, you better visually show people how it's going to help them. You can't hold up a PowerPoint and say, I would like to talk to you about vision. Right? So you need to show people that humans are very visual. And I think they get it. You know, when they see it, they kind of get it will do very simple proofs of concept for people where if you just put up one camera, by the way, I should even mention, in most cases, there you have the cameras. Yeah, yeah, you just you can argue they're already using it, you know, those black kids, you know, you see the camera up in the ceiling, it's probably just doing simple security footage. But you can get to that camera through always AI and then you could do much more sophisticated computer vision off that camera. So you just quickly set it up and showcase counting people zone analysis, you know, it's easy to set up a few zones. And, and go from there. And then I think, you know, when people kind of go, Oh, I get it. Okay, this all right, can you do this? Can you do that? And now you're getting into the more sophisticated needs.

    Steve Statler 16:27

    And so what was your approach to getting that deal? Did you say, Look, we think there's an opportunity in specialty retail, and we're going to find this partner, and we're going to create a demo, and then we're going to go and call on all the specialty retailers.

    Marty Beard 16:41

    If that was you got it? That's what yeah, that's I mean, literally, our partner Traco already had some experience with Lazy Boy, and saw an opportunity to provide them more sophisticated AI, they know us. And so they said, Hey, we want to let's go ahead and we demoed, and exact, you just laid it out exactly. It was kind of like, okay, that's what we need. And then we worked, and we built it. And then we started in a couple of early stores, optimize the application, you know, AI is very dynamic, right? It's, it's garbage in, garbage out. It's all based on the model, right? So you want a model that's really optimized for the specific use case. And that optimization process takes time. Right, you need to get out in the real world, and you need to collect that data, and you need to look at it and try it and refine it and so forth. So we did that in the early stores. And now we're in a very touching, touching, touching, touching, just roll it out, store by store.

    Steve Statler 17:44

    And can this technology be used to combat shrink? That seems to be one of the big issues in retail, can you spot people out to steal things?

    Marty Beard 17:54

    You could, you could also shrinkage also could be for example, like in the case of, of, you know, a fast casual restaurant like Mendocino Farms or, you know, or a quick service restaurant like McDonald's or so forth. They're just maniacal about not wasting food or not wasting things. So wastage is kind of similar. It's like, let's not waste food. Right? So cameras are fantastic. Right? You can really analyze Yeah, we're wasting here, not here, so forth. And then obviously, shrinkage, that's more of a security. You could potentially have the cameras on the checkout self checkout, which is where a lot of that seems to happen. You could use cameras there. But we're not a security company per se. We can help with security, but that's not really our forte.

    Steve Statler 18:49

    And what does the deployment look like in terms of the infrastructure you introduced? Your company is? Edge I use the word edge a couple of times. Yeah. So I have local servers. They like with a lot of NVIDIA GPUs or.

    Marty Beard 19:06

    So maybe in this case, let's back up from from what happens. So I'll just take the case of listing some real cases. So like, Lazy Boy, okay, there are cameras, right? Let's say there's four to five cameras. These are just simple what are called IP cameras, they're there, they're able to hook up to the internet, right? They can accept Wi Fi, they can, etc. Those are the cameras. They're capturing the images. Those camera feeds go to what's called an edge device, right? And a brand that is everybody knows called nividia. Sells edge devices. So to a lot of other people, I'm just using the videos as an example. And they have a line of products called Jetsons. They're these boxes. They're about this big and the camera feeds into the box. What's the box doing the box actually has the AI model, the actual models there. And it's leveraging the GPUs, the graphical processing unit sitting in the in the video box. And it's taking the camera feeds in. And it's applying the model that was built, how many people are aware of those people demographic, so the people with zones or the whatever the model was that was built, if the magic is happening literally on that device. And and that's so again, if I'm backing up, and then always AI is the platform that built the app, deployed the app, put it all on that device to actually give you that, that that real time fee, okay? Now, in some cases, you got the cameras, and instead of going to an edge device, they're going to the cloud, right? So a company might say I don't want to put out edge hardware, I just want to go directly to the cloud. That's fine. And we do we do a lot of that. The only problem with that is it's expensive, right? Because the cloud is going to charge you for the for the inference, right? Touching, touching, touching every time, you know, the inference is just the logic, it's just saying is that a person is that a person is that a person and that's touching, touching, touching? Right in the cloud on an edge device, there's no charge for that. It just runs 24/7, there's no, there's no charge. But that's the Djinn generic architecture, you got cameras connected to either an edge device or the cloud. And then you've got the magic happening. The AI really algorithm, the mat, I call it the magic is either literally happening in that edge device or in the cloud. And the magic is all created and built by always in our in our situation.

    Steve Statler 21:44

    And How expensive are those video devices?

    Marty Beard 21:48

    Getting going backwards to the cameras? You know, I mean, literally cameras now like, what 70 bucks, 100 bucks, right?

    Steve Statler 21:54

    I assuming that's free. That's already there.

    Marty Beard 21:57

    Yeah, about the the edge devices are, you know, these are 700 800 900 bucks, something like that. But they can feed they can see many, many cameras. If you want to get to a beefier box that can do more and run faster and have more GPUs that now you're into maybe a couple of $1,000. But you know, in the case of of lazy boy, they're not even using the video, they're using a different product. And in their case, it's an edge device. And I think it runs somewhere around 130 bucks. Something like well, so it's the edge cost. This is Moore's law, right? Would you and I've seen, it's like the power goes up, the cost goes down. Ai magic can happen now on the edge. It couldn't you know, five years ago, it was really difficult. Now you can do.

    Steve Statler 22:50

    Interesting. So I remember. I'm sort of interested in your opinion on maybe an adjacent application for same technology, Amazon Fresh. The all of the cameras looking at the yeah, basically streamlining payment. Yeah. Which just seems amazing. Yeah, yeah. I guess I should ask you if that's something that you have worked on. But one year, one of the criticisms there is, hardware is really expensive, because they have so many cameras, and you can't just use the security cameras, and you got to have like, a lot of compute to do that level of processing. Is that a valid criticism?

    Marty Beard 23:34

    Is that just? No, it is? It is? Yeah, I think I think in that specific case, the way that Amazon went about it, it's amazing. But it's really, really expensive. The way that they they did admit it. And if you've noticed, they've backed off that a bit and kind of shut down some of those those stores and kind of kind of backed off because I think the way they did it is you literally have probably hundreds of cameras in there. Now. We did it a very different way. We have some experience with what's called contactless checkout. All of us have seen this now we all get used to like going to a target or something and then we we self checkout, right? It's all barcode scanning. And what's happened is now people are really looking at different modalities of contactless checkout, where you don't interact with the human being you take your items, most people don't have 100 items, we have 10 or less, and you put them down on something. Cameras look at it, immediately identify it, you tap your phone, you walk out, right RFID tags are another way that people are doing that that's not computer vision. That's a different technology. So I think there's a lot of effort and money going into making that process faster. The way that Amazon went about it, I think was amazing and like, but too expensive to really roll out big time. Right. And I think people are still working on this this challenge. We did do a lot of work like I said in in contactless checkout. Interesting, difficult computer vision challenge for sure.

    Steve Statler 25:13

    Yeah, there's a there's an Amazon Fresh still near where we both flip. There's just inland from you and just south of me in the camo mountain area, and it's just been sitting there and they haven't turned it on. It's like very frustrating I kind of experiences.

    Marty Beard 25:30

    Yeah, yeah.

    Steve Statler 25:32

    Where would you say this part of AI is in terms of Crossing the Chasm. Are you?

    Marty Beard 25:43

    Yeah, I think I think it's in the in the middle of that process. It's, it's, it's, it's got a couple of, you know, maybe the way to answer it is, it's very different from, for example, generative AI and kind of open AI, open AI, right. So that's all of us have browsers, we we boot up our browser, we go to, you know, chat GPT. And we interface with it. And we have that kind of magic. But that's, that's looking backwards. It's historical. So it's grabbing data that already exists out in the world on the internet, and pulling it together for you in this amazing, human like manner. Right? It's so it's almost like Wikipedia on steroids. It's kind of like, I'm just going to pull together information from you, or I'm going to generate information for you. But the way I'm doing that is I'm looking, it's a mirror on the past, right? I'm just grabbing information that already exists, and I'm going to present it to you. Okay, when you get into computer vision, this is literally real time, new data that does not exist. Right. So this is what's happening now. This is not Chad GBT, which goes up to November of 2022, or whatever the cutoff date on the on the model is, so now you're dealing with kind of real time information. And you also have hardware, right? So you've got physics, you have atoms are on, you know, you've got metal devices, cameras, you've got drones, you've got cars, you have things that are literally out there that are also you know, have to be enabled to to take this AI and make it work for you to get that real time data. So my point being it's, it's more complex. And the industry is working its way through the best ways to do that. The most optimal ways to do that, the cheapest ways to do that, right? And so we're all versus, you know, open AI, which I can jealously look at as it didn't have to worry about that. Because everybody has a browser, on their laptop, their iPad, their phone, they already and it's literally just accessing magic through that. This is you got to get it running on a on a physical device. And then from there, you need to collect that information and then present it right. So it's a little bit, it's a little bit tougher, So long answer your question, but I think we're, we're in the middle of that process.

    Steve Statler 28:05

    So this is an area that Elon Musk is famously very keen on. So kind of jettisoning all these extra sensors from the cars and just focusing on the on the camera input. And one of the things that he uses to kind of bolster the perceived value of Tesla, is the the corpus of training data they have they're gathering more training data than anyone else. That's something Well, first of all, do you buy that argument? And does that apply to always AI? Do you see you building a kind of a barrier to entry? Because you just have more hours of video streaming into your system? No.

    Marty Beard 28:48

    Well, I so first question I do I do buy his argument, because the more data you have, the more training you can do, the better the model will react, and therefore be able to predict, and that's just an ongoing thing. In his case, in in the case of like open AI, they're dealing with massive datasets, like, you know, I think I read something like more molecules in the more than molecules in the ocean. It's, it's kind of like just massive, massive, massive datasets. And yeah, the more data, the better the predictions will, will be right. In our case, in in computer vision, the models are much smaller. Right? So just just physically smaller, they don't take as much room as needed on those situations. That's, that's so that's a really important distinction. And why is that because you can you can train a computer vision model off a relatively small set to do exactly what you need it to do, right? That's one of the reasons it can actually get on a physical device. That's not a massive servers sitting in a data center somewhere, it's literally on a, on a smaller device. What's unique about us is our end to end capability, we are totally unique and that you can go all the way from collecting data to build the model, train the model, build your specific application, literally get it out running on a camera and get the data that whole end and process. You can log it always out and do all of it. By the way, you can do it all remotely. So if you're, if you're on 100, cameras, out in 10, different locations, you can see everything going on there just from your, your desktop or your laptop. So that's our uniqueness is the end end. It's not really a data dip, you know, like a data model battle, as much as it is actually a practical AI implementation.

    Steve Statler 30:51

    And so what's next for you guys from an r&d perspective? Where are you taking the product.

    Marty Beard 30:57

    So we're at the biggest thing for us is what's called multimodal tech loves to come up with expressions that are like sound cool. And just so multimodal, all that means is another mode. In our case, for example, sound gets combined with vision to make the application even better. So for example, in in a mining scenario, I'm not only seeing what's happening, I can actually hear what's happening, which could be really important in a mine things are exploding. And, you know, there's noises that are important, I can combine that into into a synthesized AI model. And it's even better, even smarter and even more human like, because all of us are seeing and hearing and so forth. So that's, that's so we're we just announced this, that we're able to handle multimodal capabilities. So we've gone from vision, to sound, to now we're starting to deal with text and large language models and so forth, which are also starting to find their way into our world. So that's super exciting, because now it becomes really a AI platform, not just a vision platform, but again, still very focused on practically getting it working. Right, we're not the guys that are in the back office just doing data science work. This is like, we want to take that work. And then we want to get it out and get it working in the real world. So super exciting.

    Steve Statler 32:26

    Multimodal, so multimodal, does that include text, the commissary the conversations that people are having in their work, okay.

    Marty Beard 32:38

    Yeah. Yep, you could start, you could start building in voice into the model, you could build in text into the model, you could build in sound coming from the sensor. So all of this is getting fed somehow, by a sense, right. But, you know, it's, it's obviously heading our way. And we can see that in some of our early adopters and so forth.

    Steve Statler 33:06

    And what about text? So I, that seems like a separate thing. So can you unpack that?

    Marty Beard 33:14

    So. So text is being used for so the very beginning with this whole process of any AI is the model, the model that you develop? Right, you know, garbage in, garbage out, and you want that to be as good as you can get. And building a model is an art form you because you have to, you know, you collect images, you annotate images, you're trying to train something, look for this, right. There's a whole category called synthetic data, which is that which is, you know, data that's not quote, real video images from the real world, it's just coming off the internet. And what's happening is, you're now starting to be able to use text. To describe what kind of synthetic data set you want, please give me 50 images of people walking in a shoe store and looking at Nikes. Right, and then you get a synthetic representation of that. And then that helps you make your model better. Now you also have real images of people walking in stores and looking at Nikes. But now you can also augment it with the synthetic data, it just makes everything faster. And and and kind of gets to the magic quicker. So that's super interesting, because it started to marry the unreal world with the real world, this synthetic kind of the real world.

    Steve Statler 34:35

    Amazing. Well, Marty, it's been a real pleasure. We've got a whole other chat to this conversation where we delve into your amazing history you been part of some amazing companies shoot some amazing things and thanks very much. I'm talking about alwaysAI.

    Marty Beard 34:52

    Thank you. I was great.

    Steve Statler 34:53

    So Marty, you've had an amazing career spanning Oracle side Bass CEO, Blackberry co founder CEO of always AI, and I always look at children's work to Oracle. That's like, that's quite a thing in my mind, because Larry Ellison runs a tight ship. And they're like, yes, legendary. How did you get to be vice president? at Oracle, what were you doing?

    Marty Beard 35:25

    I, you know, maybe maybe it was just I got lucky on the on the timing, who knows? But yeah, this is really early in my career. And the web is really having a big impact on E commerce and all kinds of business applications. And Larry, Larry decided that he wanted to see how much of his sales he could do online. Right. So he, he was tired of having to rely on this huge sales force, you know, of human beings, right, that that would go out and try to sell Oracle database and a legendary Salesforce legendary.

    Steve Statler 36:05

    Yeah, listen, very, very effective. Yeah.

    Marty Beard 36:07

    And at the time that I got there, they were doing a lot of consulting under a gentleman named Ray Lane, famous executive at Oracle. And Larry was sort of wanted to get back to software sales and try to make it a more direct sale. And I was in a group that was focused on doing that for small to medium sized businesses, it was called Oracle online. And it was all about, you know, selling database licenses online, and really leveraging the web to do that. So it was a fascinating project, Larry was very involved in it. I just happened to be at the right place at the right time. And they had a great opportunity to, to not only work with him, but just work with a wide variety of amazing talent at at Oracle. And we did pretty well, we sold, we sold a lot of database online. So it was kind of the start of that more online sales. And that became just normal within the enterprise software space.

    Steve Statler 36:08

    How would you describe his management approach? He's like this Zelie character that pops up whenever you watch a documentary about Silicon Valley, whether it's about Steve Jobs, or Elon Musk, or whatever, yeah, kind of appears as a feature character. But what's the actually like when he's working on a new project like this.

    Marty Beard 37:24

    My very first meeting that involved him, there was a bunch of people in a room in the foreground. And he found out that there was a group of people in Sacramento that were taking orders off fax machines, and then rekeying them and he fired them all. So as always, sitting in the meeting, thinking, Okay, this is how things roll. So he found this massive inefficiency, and, and he got upset about that. And that wasn't even on the agenda. It's just, that just happened. Right? And then, and I later found out I said, is, you know what happened? And it was like, Yeah, some people, some people got impacted, not everybody, but he did find something that shouldn't, shouldn't have been there. The other, the other thing I noticed about him is he so he would make decisions like that very, very quickly, people would come in and spend a month on PowerPoints and put them in front of them, and they'd be perfect, and he would not look at them. He would just talk and just immediately get to the key issue and immediately resolve it. So very quick, very focused, very smart.

    Steve Statler 38:38

    How do you prepare to deal with someone like that?

    Marty Beard 38:40

    I think you go into knowing your subject matter from every angle that you can think of, right? And so if it was pricing, you know, online, what are we selling the database for? How are we pricing it, you know, what are the different permutations etc, you just need to make sure that you add all that cupboard. And I found him like, you know, obviously, he's legendary for his temper and all kinds of things. And outside of that situation I just mentioned and really see it that much. It was more very, very business focused very, very get to the essence of the problem and try to figure out how to how to fix it. So I found it good training, you know, to see that.

    Steve Statler 39:19

    So it was to be a little scary dealing with someone who just fired a bunch of people because they were just doing the wrong thing. But did you feel like did you have space to actually perform? Yeah, be creative?

    Marty Beard 39:33

    Yeah, I think so. Yeah. And again, this is me as part of a team. This is not you know, Marty, they're sitting there having this one on one with Larry Ellison. That was not that was not happening. But yeah, I think the team, look, Oracle, a lot of really smart people. Larry hired really, really smart people from a lot of different backgrounds. Definitely a technology oriented place, high high value of engineering backgrounds and so forth. Eat. And just a lot of people around the table that are just smart Larry moves really fast. Honestly, he's probably the smartest guy sitting in the room. I think everybody actually knows that not just because he owns the company, and is that multi billionaire, but actually, really, really smart and fast. So you're on your toes, you know, your subject and, and then, you know, he liked another give me another simple example. It was a lot of people were building a website to try to make the interface better for the customer, and so forth. And Larry walks into his office and boots up and asked somebody to go to amazon.com, which was relatively early on. And he said, Why are we redoing things? Do you think they know how to make things work where they're selling directly to customers? And we're like, Oh, yeah. Books? Yes. Beyond books? Yes. Why are we doing this? That's probably a paradigm that we should follow. Just follow that. That's the way it worked. That makes sense. Practical, you've probably figured this out. Maybe there's some things there that are best practice that we can learn from, etc. It's just an example. Very practical guy.

    Steve Statler 41:03

    So all that things from working in that environment, I realized this kind of one relatively small part of your career, just a few years. But as you're the CEO, now, you're kind of in his, his position. Are there? Is there a set of things that you learned from him that you're putting into practice and other set of things that you said? I'm not going to see that? You know?

    Marty Beard 41:32

    Yeah, I think, yeah, definitely a set of things from him. And then after him, I worked for a gentleman named John Chen for many years. Also incredibly smart and very operationally focused and so forth. Yeah, I think I think learning that in a meeting, it's just completely focused on the business problem. Right, that that came across really clearly from Larry and John and, and really trying to understand the business problem, focus on that, that's the most important thing, talk, get the right people around the table, talk about the problem and try to resolve it and move on to the next. The next problem. So I guess, you know, focus, and really focusing on the right issue, not a lot of just smoke around the issue or things that are kind of tangential to what you're trying to try to accomplish. I think on the other hand, however, and I would say especially my experience at Sybase, and so forth, maybe I try to also focus a little bit more on on innovation, and kind of also allowing time for people to really, especially when you're talking about roadmaps and the future, so it's not so much, here's a business problem we need to solve. It's more about where's the market going? And what kind of innovation do we need to focus on? That's a very different process, right. And that's, that's something that's really, really critical to add that juice into the system. So you, you know, you've got the new innovative technology that you need, because the market just constantly moves. Right? If you're only focused on the box of existing business problems, you're gonna miss the next box. Right. So. So I think, you know, trying to do those two things, were those different hats at the right time.

    Steve Statler 43:16

    So how did you end up transitioning to Sybase? Obviously, it was a promotion? So I imagined that featured in.

    Marty Beard 43:22

    Yeah, I think such a, I think for me, you know, after three, four years, at Oracle, I think Oracle had reached this point where it was, at that time, just one of the most highly valued companies in tech and, and was was really one of the picks and shovels for everybody going after.com, right, you had to you were using an Oracle database, probably on a Sun Microsystems machine. And you were that that they were just riding that wave, it kind of felt like not tapped out, but like, okay, that we're just gonna do this forever. I ran across site days, which was had competed with Oracle in the database space, but for a whole variety of reasons, had kind of gotten beat up the dead end was really looking to reinvent itself. And John Chen had come in as the CEO and it was really in a in a place of like, okay, I'm going to do some, I'm going to make some moves I'm going to do, I'm going to acquire some companies, I'm going to really how to do more. And he needed some strategic thinking something that I always been really strong at and, and just wanted somebody to help him kind of make those moves. And it just seemed like a such a unique opportunity that grabbed me, and it turned out to be a wonderful experience.

    Steve Statler 44:41

    You were there for a long time. Well, yeah. Do you look upon as some of your biggest achievements?

    Marty Beard 44:46

    I mean, we took well, we transformed a company that was a, you know, database company that was having a hard time growing into a fast growing mobile enterprise company. So I mean, it was literally pivoting. from kind of a stayed, you know, stable but having a hard time growing database business because again, Oracle was really winning at just but taking that base and the cash that you were able to generate from that and then move into mobile enterprise and really taking a bet. Now it's just obvious like our businesses are run our mobile phones, but at that time, it was it was a fairly big move. It was like, Okay, I'm gonna approve POS and approve, you know, run my HR systems and financial systems, I'm able to do that or not on a mobile device. So we made a huge move, you know, under John's leadership and, and others to move into mobile enterprise. We bought a lot of companies, we reposition the branding, reposition the technology capabilities, reposition the Salesforce, the partner ecosystem, the whole thing. And what happened at the end of that is SAP bought us. And for a very high premium, it was a huge payoff for, for savings shareholders and for the management team. So it was, it was a fun experience.

    Steve Statler 46:04

    Amazing, and that success gave you the platform to move up to Chairman and CEO of live ops.

    Marty Beard 46:13

    Yeah, yeah, yeah, I think coming out of that experience, the Sybase experience it was like, All right, I've worn a lot of different hats done a lot of exciting things had a big success. As part of a great team and all that stuff. Now I want to go do it myself. Right, I want to see if I can be CEO of GE, and you know, I had run of large division of of Sybase, it was like a $250 million business but not as CEO, right, I really wanted to do that. And LiveOps was a super interesting company that was run by a gentleman named Maynard Webb. And Maynard had come he was pretty famous guy that had come out of eBay. And he got fascinated by live Ops is kind of a, it was a lot of people that worked out of their houses. And it was sort of labor on demand. In some ways, it was a precursor of like, thinking of like, Uber was kind of transportation on demand. This was like labor on demand. And not outsourcing, but actually literally on demand, like I need somebody for four hours, I need somebody for six hours. That was super interesting. And then the platform that that that live ops, that bill was used by a lot of call centers and customer service. super interesting. And it was trying to leverage social and all kinds of different channels. So it was yeah, it was it was a fun opportunity for me to come in and, and, and run the company and, and become chairman of the board and go through that process. Great VCs were destined to. And yeah, ran that for quite a while.

    Steve Statler 47:48

    And so what persuaded you to go to BlackBerry and where were they in their incredibly storied history? I mean, this is a company that was just a colossus. At one point, I remember, when I first joined my first startup, I left IBM bought the company I worked at, and I'm like, okay, now's the time when I get to go from being a director during a VP and running sales and everything. And first thing I got from our CEO who's just an amazingly gifted guy, it was was a BlackBerry device. And that was the iconic kind of accessory you had when you were going to Sand Hill Road and doing all the things that you do.

    Marty Beard 48:30

    That's right. Yeah, no, I mean, come on. It's, it's iconic brand, right. And really had established the smartphone space. But yeah, when I when I got involved in that. So what happened was John Chen, who I mentioned earlier, and had run Sybase, he was asked to turn which will, you know, obviously, Sybase, that was a really successful turnaround. He had been asked to inject the same magic at Blackberry. And they'd gone to an outsider by that I mean, a non Canadian, you know, it's a Canadian company, BlackBerry, and it was like, Okay, we're going to bring in some Californians here. And, and in the whole job was to pivot from hardware only into software. Right. Hardware is really hard. And smartphones, you know, Apple obviously had a kick butt and was quickly becoming the de facto leader. And you had a lot of Android devices and all kinds of all kinds of things going on. And BlackBerry is trying to find its way in that number. It had made its name on on keyboards. Right. And Steve Jobs had famously held his hand up and said, This is a keyboard, it's your fingers. Right. And so yeah, so So John had come in to for that turnaround. And he asked me to come in as Chief Operating Officer and kind of help him move away from the hardcore hardware manufacturing. So forth and get into cybersecurity software. BlackBerry was always known as a very secure device. The President used to the CIA used it, you know, it was the lockdown secure device. And so how to leverage that heritage into this cybersecurity software space and so that I was involved in that whole pivot, it was literally a hardcore 180 move from a vastly money losing hardware business into a money making pure software company and try to do that as fast as you can. So that was a interesting, fascinating, exhausting experience that, that I went through, but lucky, lucky to have gone through.

    Steve Statler 50:46

    Very good. So you are part of history in well, more than once in your high tech history more than once in your career. One, what is it that you think gives you this facility with strategy? What what makes a good strategist? And what's your approach?

    Marty Beard 51:06

    You know, it's I like, I like thinking about patterns that are emerging, maybe from different areas. And I love a lot of communicating about what's coming. And by that, I mean, it's not just like communicating like a blog post or speech. But it's, it's sort of putting into words, the ideas that people have about what's going on. In the future, I've always been very, very comfortable doing that. So I think you can think of people that might have a really deep technical background, and maybe have a hard time trying to describe where that technology is going. Or somebody who might be good at the description, but really doesn't have a strong background in in kind of the tech and how we got to where we are. Maybe I'm a good combination of those of those two things. I've always really enjoyed thinking about what's coming. And then talking about that, right. And in and describing that. I studied, I'm a very unusual United studied philosophy and classics and, and in rhetoric, and kind of, in my background, when I went to Berkeley, I have to tell you, honestly, that is far and away the most practical business experience I've ever had, by far. I mean, I got an MBA. That was fine. I learned a lot. I was probably, you know, one of the only MBA students that have literally never taken accounting or anything, it seemed like everybody else already had an undergraduate business degree. But yeah, so I absorbed that stuff, but honestly, not not even remotely as useful as, as some of the thinking skills and communication skills that I got coming out of Berkeley. So I think that helped me as well. Maybe I'm just naturally predisposed that way. I don't know.

    Steve Statler 52:50

    So the philosopher and then you're the first high tech Exec. Who is the philosophy graduate that I've spoken to, and it's really, my, both my parents studied philosophy. So I'm like, I never did, but I always had that respect. So is it that it requires you to think very, very clearly, you can't just kind of waffle around the argument.

    Marty Beard 53:15

    Yeah right. Right. Yeah, it's, well, first of all, I think it's, it's really, really hard. So people, people roll their eyes, if you say, if somebody says, especially today's STEM focus world, if somebody says, What are you studying, and you say, I'm studying philosophy, and they roll their eyes, those people rolling their eyes, which is flailed if they it's very, very difficult, and but also very analytical. So if you can make your way through a Platonic dialogue, for example. That is not easy. But if you can do that, and you can articulate it, that's, that shows that you can analyze a lot of different things and kind of bring it home. Which if you think about business, you're analyzing a lot of things and trying to bring in all right. So I think, yeah, it gives you that that thinking discipline. And, you know, it's a much, much, much more complicated subject that most people realize, right? So I think you get people coming out of there that are very analytical can take about super complex things and bring them home, they can communicate that deal. Yeah, so I'm actually not surprised. I haven't run across that many, but I know they're out there.

    Steve Statler 54:29

    So what's the difference between a Platonic dialogue and a Socratic dialogue?

    Marty Beard 54:34

    Well, they're one in the same because the main characters is Socrates. Right? So Plato uses Socrates as his character. And so the character notice Socrates is the one that's always annoying everybody and asking him questions and trying to try to arrive at some, some conclusion. He's annoying, he's arrogant. He's he's a he's a great personality to do literature. But yeah, he's he's a character in Plato.

    Steve Statler 55:03

    Not a quest, you can tell. I haven't had any questions for this podcast, because it definitely wouldn't have recruited that question. But it's fascinating. So I want to wrap this up to make sure that we have room for the rest of the discussion short term, going back to strategy. What are the key elements? You know, you talk to someone and they say, Oh, we have a strategy for this strategy is a word that is bandied around very loosely. What are the key elements of the strategy in your mind?

    Marty Beard 55:34

    Yeah, that's a that's a? I mean, that's a great question. And I think, a strategy, it's kind of like, what is what is the modus operandi of the company? And by that, I mean, what is the pitch? And then why are you pitching that? Right? What is the product that supports the pitch? And okay, now you're at the product level. Why is that product, what the customer actually needs? What's the value, the ROI that the customer is deriving from that product that you just pitched? Right? So you're kind of getting through that those layers high low, you're sort of starting high level? Like, we're focused on computer vision, and then you're talking about the product, okay, how do you actually bring that to life? And then you get to the customer? Why does the customer actually need that? What's the business problem that's actually being being solved? I think if you only if you start the other way, a lot of people will just focus on the business problem, the distance problem, but which is fine. But you know, again, back to Steve Jobs, where he famously said, you know, if I ask customers what, what they want, then I would never have built the iPhone, or I wouldn't have built the iPad, because they weren't telling me that's what they needed. What they really needed was information more quickly, information on the fly, that was the real name. Right? So then he backed into, you know, the product and so forth. So you can kind of cut both ways. But yeah, I think it's an all it's a word that encompasses your pitch, the way that you talk about your business, the product that supports that pitch, and then the problem that you're ultimately solving for, for the customer, there's lots of different ways to cut it.

    Steve Statler 57:19

    So you know, the rest, and we've talked a lot about always AI, but I'm interested in, you know, looking back on this history, and you've seen a lot of success, and doubtless you've dealt with a lot of challenges as well. Where are you in that Steve Jobs in? Not contempt, but but kind of self belief in direction? Where are you in listening to the customer, and kind of the lean startup? And, and you you have an idea of this is the problem we're gonna solve? This is the mountain and you know, we're gonna build it, and they'll they're gonna, yeah, have a great success.

    Marty Beard 58:03

    I think we're, I think we're in the middle of that. Honestly, I think it's still a vision, deep learning AI and how that impacts vision. Let's see, we can call it computer vision or vision, AI is still really early, it's very early, nobody can tell you exactly how it's going to evolve, right. And if you think about a part of the challenges, human vision, just our eyes, and arguably the most complex system that we have in our bodies, that's and also really critical to our lives. I think, you know, the applications that you could build are literally what you can see. Right? Okay, that's a lot that's really big. So part of the challenge is, it's so big and so encompassing. How do you chunk it down into areas that are really worthy of more focus, where machine vision and so forth would really fit? So I think the industry is still figuring this out. We're still in the middle of it. Some people are willing to take the leap, and they're more innovative, and they're kind of ahead on the curve. Other companies are a little more conservative, they know it's coming. They know it can help their business, but maybe they're not ready yet. So I think we're, we're in the middle. We're in the middle of it. Very interesting.

    Steve Statler 59:26

    Very good. Well, we come to the show where I asked you the hardest question, it's not about Socratic dialogue. It's about your three three songs that have meaning for you and why did you able to distill it down just songs?

    Marty Beard 59:42

    Well, I can Okay, any anything by so but anything by Bob Dylan is going to be it's going to be number one for me by far. And I would take I would take anything from an album called Blood on the tracks. which is a famous album, but for you know anybody under the age of 45 probably. But yeah, that there's a song called tangled up in blue, which is just an amazing song. By by Bob Dylan, I picked that one for sure. And then I'm a big Miles Davis fan so I think I probably picked something from kind of blue like so what are you know? Yeah, anything love that good. You're on an island and you only had one album? That's probably one that you might. might tell. Yeah. You said three, three. It's a song anything by the Grateful Dead in anything, any Jerry Garcia jam that goes on for 25 minutes. I'll take that.

    Steve Statler 1:00:47

    That's the value. For money. You got a choice the money little to show it goes on for a long time. Exactly. I went to one Grateful Dead concert. But it was post Jerry Garcia, but But you don't remember. I don't remember very well. It was good. Very good. Well, Marty, thanks very much. It's been a real pleasure having you on the podcast. Appreciate it. Great. Thank you very much. Well, I hope you enjoyed that as much as I did. It was really a great conversation. From my perspective. I want to thank you for staying with us. And listening to old thing and listening to the adverts. If you add a new, just to remind you all of the money from our advertising goes to the Monarch school for kids who have in homeless families. And I really appreciate the fact that you've listened and then put the hours in, I think it's an incredible industry that we work in. And so, but it's important to keep up with what's going on. So thank you. Thanks to Aaron Hammock and Brooke Ellsworth for all the work they do on helping me get the podcast out and stay safe. Until next time.