193. AlwaysAi Transcript

Sharemore_horiz

AlwaysAI

SUMMARY KEYWORDS

marty, people, talk, ai, restaurant, computer vision, data, cameras, customers, love, zone, listeners, technology, work, computer, identify, world, long, learning, machines

SPEAKERS

Marty (61%), Jeremy (39%), Intro (1%) 

I

Intro

0:02

This is the restaurant technology guys podcast, helping you run your restaurant better.

JJ

Jeremy Julian

0:13

Before we move forward with the show, I wanted to share about a product that I came across recently. We’re in the middle of the summertime. And so you’re going through dads and grads and I know the holidays are just as bad. But it’s a product that’s trying to become the OpenTable. for large parties. The name is restaurant, Nick and his team have created a online booking solution to allow restaurants to book large parties and do them online in such an easy way. It’s a brilliant solution. And having just gone through graduation for my son, I would have loved to have had a solution like this, check out Nick and his solution restaurant, when you get a few minutes after the show.

JJ

Jeremy Julian

0:53

Welcome back to the restaurant technology guys podcast. We thank you everyone out there for taking part of your weekend, listening to my rambling and my nonsense, but we appreciate it nonetheless. Today, we are going to jump into a really cool category. And I’m gonna let our guests talk a little bit about it. But let me introduce him first. He’s the co founder and CEO of always AI. Marty. Marty, why don’t you introduce yourself? And then we can talk a little bit about your background. And then what always AI does? Yeah, yeah. Thank you. Yeah, like, like you said, name’s Marty. And started always AI about five years ago. And we’re completely focused on computer vision, which I know we’re going to get into and what that means for, for restaurants. And I, myself, I have a long background in Silicon Valley, and enterprise software and mobile software mobility, a lot of years of experience in those areas. I love it. I love it. So just because the words AI machine learning computer vision gets thrown around. Before we jump into a little bit about your background, I guess, give me your definition of how you would define AI, how you would define computer vision and how you would define machine learning because I just got back, like I said from a trade show recently, and it’s amazing how every single booth had some form of AI in their booth. And it’s like, Well, do you really have AI? And what does that really mean? So I’d love to educate our listeners on on those things. Yeah, I mean, at the highest level, obviously, AI is really about can we get machines to do work for us, predict for us be quote unquote, intelligent, you know, for us, machine learning is a technique within always AI that helps computers, cameras, cars, drones, you know, all kinds of things do do smart things, it’s really at the end of the day, a lot of math, and just kind of grabbing a lot of data and then making predictions off that. Computer Vision is a super interesting

MB

Marty Beard

2:56

subset of AI. Absolutely. And it is just completely focused on vision. So you can almost think about it as you know, one of the most powerful senses of the human body is sight. And, and trying to apply sight, you know, and leverage sight in computers. And in my case, particularly in cameras, I love it so, so that that’s what computer vision is,

JJ

Jeremy Julian

3:21

I love it. I love it. And funnily enough, I think we weren’t calling it AI, but there’s been AI inside of computer applications. For years, we just never really knew and, and now it’s kind of the big buzzword. And a little bit more about your background, you said you were in Silicon Valley, you know, talk to me about some of the previous projects you worked on that kind of were related to this, or that kind of the idea for for creating, always I

MB

Marty Beard

3:44

yeah, I started off in, in big enterprise software companies like Oracle and, and really went through the whole transformation of, of kind of the client server world into the web based world and, and then from there, I spent a lot of time in mobility that was really about taking the web out to our mobile phones out to our smartphones. And that was a super interesting. You know, we’re all living on our phones, obviously, we’ve all gotten we’re all managing our businesses on our phones. But there was a time when that wasn’t the case, right? And we had to kind of make make that transition. And now we’re pushing from the mobile phone out to the actual real world where the real world is actually doing that work for us, without us asking. Right? So now you’ve got a bunch of sensors and cameras and they’re doing stuff without you initiating that on your mobile phone. So it’s kind of the next wave of pushing intelligence closer and closer to real time. And we’re where we’re actually living and working.

JJ

Jeremy Julian

4:44

Well, and I think, for our audience, it’s less familiar but if you listen back to the last maybe 10 podcasts that we’ve had that have gotten posted, there’s been a lot of people that are using AI and machine learning to solve, solve some unique problems from food cost management to food In preparation management, talk to me a little bit about what what are you guys trying to tackle it always AI as it relates to restaurants, you know, talk to me a little bit about kind of, because there’s so many different areas of opportunity where we’ve had data, but don’t necessarily know what to do with it. And so you guys come in?

MB

Marty Beard

5:14

Yeah, I mean, I think right, because Silicon Valley always has the latest buzzword and trend. And, you know, it’s getting pushed and marketed. And what the heck does it really mean for me, like as a as a restaurant operator? Yep. And you know, at the end of the day, it’s kind of like, Can some of this stuff improve my margins? Right? So if I’m, you know, if I’m making 5% margin, or 10% margin, can I make 12% or 15%? By by using some of these technologies, right? And in our case, it really is about is there? Is there real time data that you would find useful? You know, about your customers? You know, what are they buying? Where are they going in the store? What are they doing? What kind of customers do I actually have? How many do I have? Right? That’s an example or operations. Okay, I’m, you know, I’m running a fast casual restaurant, and okay, am I, in my as productive as I could be? am I wasting? You know, resources? Is my is my labor as efficient as it could be, etc. So these are the things that real time data can actually help you act on. Right? And so forget about buzzword AI computer vision, it’s just can I get real time data, that I as a manager, as an operator can actually improve my margins? Right. And that’s, that’s what we’ve been, you know, we’ve been seeing that happen with a lot of different customers. And I’m sure we’ll get into it, but but that at the end of the day, if I can’t help a restaurant operator do that, then then I am just kind of selling a technology they don’t really need, right? Yeah.

JJ

Jeremy Julian

6:50

Well, technology for technology’s sake, we talk about it all the time. It’s like it technology for technology’s sake does nothing for us. You wouldn’t be living on your phone, if it didn’t make your life easier if it didn’t make your life better, if didn’t make your life. Some might. Some might argue that it’s also make us made us much more distracted. And there’s other side effects of it. But I think humans will adapt to that as well.

MB

Marty Beard

7:11

Yeah, no, I agree. I mean, well, there’s always again, despite Silicon Valley, saying, it’s all good. It’s all good. Yeah, there’s, there’s always consequences of, of all these all these technologies, you just mentioned, a couple obviously, in the case of computer vision, you, you worry about things like privacy, and you want to make sure that you can protect people’s identities and so forth. But But again, at the end of the day, this is business, right? This is about I’m trying to improve my margin, and I’m willing to invest in technology if it can do that. And, you know, I know the restaurant industry, he’s got a lot of different technologies, POS terminals, you know, back, you know, they’ve got all kinds of different technologies. It’s pretty fragmented. Yes, is pretty fragmented, very fragmented. Right? Yeah, it’s fragmented. And, and, you know, AI comes onto the scene. And I think, all we do is we focus on look, if you have some cameras are already installed, we can get those cameras doing work for you that they’re not doing right now, we can make them a lot smarter. And we can give you that information. We can deliver it to your mobile phone, we can deliver it to your laptop, however you want to consume it, but we can get you that information. By leveraging you know, these computer vision techniques that are now much more cost effective and easy to easy to implement.

JJ

Jeremy Julian

8:29

Yeah, no. And I think the last guests we had on that I had on, most recently, they talked about it, almost everybody has cameras installed in some way, shape or form. But typically, they’ve installed these cameras so that they can manage slip and falls, they can manage robberies, they can manage, you know, like really bad incidents. And they’re they’re really not necessarily using these to do anything. From a front of the house journey perspective. Talk to me about some of the areas that you guys are using cameras to tackle either guest engagement, guest traffic guest attention, attrition where people walk in and they see the lines too long. Talk to me a little bit about how you guys are solving there on the front of the house to drive additional revenue.

MB

Marty Beard

9:09

Yeah, no, it’s great, great question. So I think, you know, the main use case that we is, kind of just give me basic customer analytics, that’s a fancy way of saying, who’s walking into my Burger King, for example. And then speed of service. So they want to know from the time that Jeremy walked in and ordered a burger paid for it and waited. How long did that did that take right? And if you start understanding the average amount of time it would take then you can pick out anomalies like wow, that person’s been waiting for nine minutes or something this we got we got to take some kind of an action. So we just started understanding who’s actually in the store. And how long is it taking us to to serve that person. The other thing about the the front is Also, where are they going? Now this this bleeds a little bit, maybe outside of restaurants and more new, like classic retailers, but it’s kind of where are they going? Like, where do people tend to congregate? Where do people tend to go, etc. So you start getting that information. But again, back to our earlier discussion, it’s kind of like, okay, what do I do with that? Yeah, well, maybe you’re not serving people as quickly as you can. So we are working, for example, with Burger King and in Europe, and its speed of service, I mean, literally down to like seconds of how long it takes. And they want to know, by certain zone, so a zone would be like somebody walked in another zone would be they ordered another zone would be they paid another. So you can start getting down to that kind of granular level, where it’s like, wow, we really have a problem on the payment side, for example, then the other thing I would just mention is obviously, tons of people are never going into a store. It’s drive thru. Yep. Right. And so that’s a very similar analysis, how many cars are coming in? Are people you know, why? Why would somebody abandon a lane? Or people getting frustrated? Or, Hey, am I matching up the right car at the first window? With delivering stuff at the second window? You know, some basic stuff? Yep. And again, this is where one camera, you know, that you probably already have installed can help you analyze that. And again, you can understand how many cars what kind of cars, what are the demographics of the people in the car, you know, you can start learning that type of information. So really, at the end of the day, it’s like, who, who’s coming to my establishment? And how long is it? Am I am I serving them as efficiently as I could?

JJ

Jeremy Julian

11:31

Yeah, and I mean, historically, and I know you’re newer to the restaurant space, I mean, historically, that would be a company that would come in and study these things, it wouldn’t be able to happen through an analytics through a camera, it would, it would be a company that would sit here and look at traffic, or it’d be a door sensor. You know, we’ve all gone to a retail shop and had those sensors to walk through to get in or out. That was the only way they measured those things. Yeah, what other kinds of data because you talked about privacy, kind of on the front of the house almost on the front of house for a few more minutes. You shared the privacy aspect, but I you know, we talked about a pre show, I’ve got four kids. So sometimes when I go in, I’m just my wife and I to that Burger King. Sometimes when I go when I’ve got just my kids, sometimes I go in and kids and my wife and and my basket, you know that what I buy from the brand is going to be different? Depending upon what my use cases, are you guys able to get to that level of data to be able to say that this most likely is a is a male that’s in his mid 40s. Or, you know, you obviously don’t know who I am, or maybe you do. And this is all part of Skynet that nobody’s going to talk to you about. But at the end of the day, Facebook knows that on there probably. Or Google knows that I’m there most likely. So yeah. Talk to me a little bit. I know you guys how you guys are attributing that? And then and then what happens? Sure.

MB

Marty Beard

12:45

Sure. It’s a great, it’s a great question. And but yeah, well, let’s let’s just talk about it practically for a second. So the the old style, like the sensors you mentioned, yeah, that could that counts. But if Marty beard walks in, then I walk out, then I walked it back in, they say that’s three customers. Yep. Right? Well, obviously, it was it was one. So one of the things that computer vision needs to do well, is what’s called RE identification. In other words, don’t double count. Marty, make sure that you know, that’s just that one person. And so you get a real count, you really start understanding, and that’s those are computer vision techniques. And number two, what can I know about Marty, what’s pretty commonly used today? That’s possible, not super expensive? Definitely. Gender, you know, definitely age. You can get into, like, you know, we have one customer that’s looking into what are they wearing in for this particular customer? That’s important. And you can you can even people are starting to talk about emotion. And, you know, it’s all about just training a model to do what you want it to do. And the more data you can get into the model, the more accurate it is. Right? So but things like gender, and things like number how many distinct I don’t want to double count Jeremy or Marty, just how many distinct? Totally possible, you know, it’s male, female, excepted. totally possible. Rough age totally possible. Right. So all that stuff is is pretty easy, based on existing models out there. And we and we’ve done this in, you’ve done this many, many times. Your next question was was harder, which is, okay, there’s groups of people. Yeah, they congregate, you’re carrying one child, you have another child below. This is where you get into some of the complexity about okay, is your camera up on the roof? Looking down gets a little more difficult. Is it on the side? You know, do you have just one camera or two cameras? So then now you get into a little bit of the complexity of what can I see what can the camera actually see? And based on that, that will drive a lot of the accuracy of what you’ve been able to able to able to calculate but again, if you came in by yourself versus you came in with your family, the computer vision application should be able to know it’s the same you Yep, Right now privacy, that doesn’t mean that I’m revealing, like in the case of Burger King for us, there’s never a name associated with that. It’s just a, it’s a, it’s a numeric value that just says, I know that numeric value is a distinct human, it’s the same person, etc. We do have situations where we have one customer that opt in people opt their their employees opt in to be identified. Why? Because when they’re counting people coming in, they want to know, oh, don’t count that that’s an employee, not a customer. Yep. Right. So that’s, but that’s an opt in, that’s like, Yes, I’m allowing you to see my face and identify that Smarty beard, etc. So it’s kind of a long answer. But yeah, all that stuff is possible today, and not super gnarly. And we’re not talking about major, you know, months long projects, and so forth, or

JJ

Jeremy Julian

15:49

privacy. I mean, and it’s funny, because the more you know, over the last, probably a year or so as we’ve gotten back out of COVID. And I start to meet with people like yourself and others in the space, it’s, it’s amazing how far a lot of this stuff has come. And to your point earlier in retail, you know, the models are able to be built, and you’re able to see that and the more data you get, the smarter they get, the smarter they get, the more accurate they are. I’m gonna stay on the front of the house for a few minutes. And you talked about real time alerting. So Marty’s unhappy, and I’m starting to starting to understand state. Because Marty is gone. Now he’s ordered and my normal window time might be six minutes to prepare the food and get it bagged and call Marty, but he’s now 10 minutes. What are we doing with that data? Are we are you at a place and as always AI at a place where they’re gonna be able to give some alerts to the management to do something about Marty being upset? And oh, what’s that look like?

MB

Marty Beard

16:43

Yeah, so a lot of apps are just integrating SMS messaging. Right? Some of your listeners might have heard of a company called Twilio. Right. So you, you just basically integrate messaging into the application. So what does that mean? You just set it up to say, you know, I’ll give another example. Like, I’m a construction company, there’s somebody up on a roof without a they’re not roped in. Right? You know, right. Now I need to, I need to know that, right? In the case of a restaurant, you would set up parameters to say, Okay, if somebody has been waiting longer than x, then then I want the the manager to get that message right away and be able to act on it. So yeah, this is this is very common. This is being it’s mostly text. And, you know, just messaged what I would call just messaging based alerts, but easy to easy to do, easy to set up as triggers within the computer vision app. And, you know, and then you have to, and then obviously, it’s up to the management to say, Okay, what do you want to know, you don’t want to spam the manager, so they can’t do their work? Right. So you got to be careful, like, we all experienced that in our consumer notifications, you know, it’s like, please, please stop

JJ

Jeremy Julian

17:52

eating out half the time, which, which is a bad thing. So ignore them. Right? So I had somebody actually asked me asked me a question for this. And I’d love just to get your thoughts because I know that some of our listeners are gonna want to do this as is oftentimes managers in a casual dining environment, are managing bathroom cleanups. And they’re managing table touches, and they’re managing drink refills, is machine learning and computer vision to a place where you can see the person leaving the code stand, and tracking that they went to the bathroom to do a bathroom check to make sure that it was clean, like are we at that place where we’re able to do that, or we’re able to see that Marty beard is down to his last two ounces of beer. And he’s he’s on his first course. So we probably should go offer him another beer before he leaves.

MB

Marty Beard

18:37

That one is pretty sophisticated. The first one, let me address the first one. So the first one, all you do is you have a data set of things that you’re interested in, you know, interested in learning about. And what you do is you set up zones. So one zone is the bathroom. And one zone is you know, behind the counter and one zone is in front of the counter in one zone is where people sit down. So you easily identify these zones, and then you gather data in each zone. Right? Then you say, Okay, I’ll take my you know, Marty, the manager went from zone A, and then he went to zone B and, and yeah, you can you can easily track this this type stuff in a more classic retail establishment just think of like a shoe store. And they want to know, Well how many people go to the Nike zone? Yep. Like, you know, versus how many people go to Adidas where are we getting more interest? You know, maybe we should reorient our our items out in the floor to better match what people same thing with the restaurant like where you know, where people go, and so yeah, you can absolutely do that. Your second question is pretty tough, right? Because you know, you’re getting into stuff like okay, a glass, it’s clear. So cameras, you know, these are not human beings, right? The cameras like what is that? I want to know, Is it full, not full. So you would have to Build up a pretty sophisticated data set of glasses, and what does it look like when something’s not full? And, and you’d have to have cameras that would be able to judge that, you know, in various lighting conditions and stuff. So I don’t want to make it sound too easy. You know? Yes, you could, you could do that. I haven’t actually seen that implemented yet. I’ve heard it. I’ve heard it talked about. So I don’t have any personal experience of that, but, but my gut tells me that would be pretty tough. But you know, you could do it’d be pretty tough, though.

JJ

Jeremy Julian

20:32

Okay. So I’m going to ask you to double click real quick on the zones portion of things. Because should we think about a zone? And again, I’m a casual dining establishment, I got 50 tables in my dining room, can I set up all 50 tables to make sure that Jeremy has the server walked in the zone one, zone two, which is which happened to be, you know, because he’s walking through the dining room, but I want him to stop for two seconds just to ask, Hey, how’s your meal? Are you doing okay, and then I move on to table two or whatnot. We’ve all sat at a casual dining establishment. And right now it’s a management challenge, because one staffing is low. And, you know, everybody’s been having struggles getting getting staff, and then two, how do they identify that Marty did get touched? Because you know, what, if Marty is there, and somebody stops at Murray’s table, he’s most likely going to see that the beer is empty. And he knows because he’s got a value in it, that his steps gonna be higher if he sells him another beer, so you might as well sell him another beer. You know, so even outside of this other complex issue, now we’ve got the ability to to know that he’s been there. Is that something that that is? is possible?

MB

Marty Beard

21:31

Oh, yeah, yeah, I mean, we’re doing we’re doing so for your listeners, just imagine you’re, you’re on your computer, and you’re looking at a gallery of images, right of a restaurant, just imagine you’ve got you know, 1000 images of one restaurant location, you literally, you know, grab, grab your mouse, and you’re gonna drag over different zones. So and you just create a zone. So you’re gonna create zone one, which is table one, you’re gonna then drag it over another table, that sounds good over another table, zone three. And then what you’re going to want to do is you’re going to want to collect what’s going on in that zone. Right. So now you’ve got these zones identified the camera

JJ

Jeremy Julian

22:10

around that somebody? Is somebody that has a uniform hit this table for this amount of time,

MB

Marty Beard

22:15

right? Correct. So we’re doing some of the more granular level, it’s not exactly the use case you mentioned. But we’re working with a well known fast casual restaurant. So just think about, you know, they’ve got a couple of 100 stores. And they’re they’re making high end sandwiches. Behind the counter. They’ve identified different zones for where the bread, the bread, condiments is a different zone. Meat is a different zone. You know, so literally seven small zones that are identified as a sandwich goes from point A to point Z, right? And you’re making. And yeah, so this is a very common technique. It’s like I want to, I want to know what’s going on in a specific area. And I just identify that, call it a zone, give it a name, and collect data around that.

JJ

Jeremy Julian

23:03

And then create some reports and some alerting to make sure that people are doing what Yeah,

MB

Marty Beard

23:07

exactly, then you you’re on a dashboard, and you’re looking at all the different tables, and you can start seeing the information coming off in real time that you want to know.

JJ

Jeremy Julian

23:15

Right, love it. So the other thing that you talked about in that example of the beer and the beer not being full, and having a build this model is, I guess, this misnomer that a lot of people have when they think about computer vision and AI that, that it’s snap one picture of a pint glass. And I know what this looks like, and you chuckle and those that are just listening on audio only, you don’t maybe don’t hear that. But I know that’s not true. So I’d love for you to talk a little bit about why the more data that goes into the system, and the more analyze data that goes into the system gives you better results, because I think that’s a misnomer that people have that. You know, and and funnily enough, I talked about the fact that I have four kids, I use Google Photos, and oftentimes My daughters are, you know, are mixed up by Google Photos. And Google’s obviously got some logarithms and pretty big database. But my daughters look a lot alike, even though they’re, you know, seven years apart. They look a lot alike. But the system says, Oh, they must be the same person. And they, they label them the same way. So I’d love for you to talk a little bit about how labeling works and how the amount of data in the labeling can help you so that they don’t think it’s just hey, you’re coming install, and everything’s working great, because it knows

MB

Marty Beard

24:25

you got to train these things, right. You know, just give it a funny example, like, early on one of our we were identifying computer cords for and I kept saying that’s a snake. And you’re like, No, that is not a snake, a three year old. That is not as bad as a computer cord. And so you keep telling it that’s you know that it’s like it says it’s a snake it’s not sure. And then eventually as you keep feeding in more and more pictures, it starts training itself. And it learns that’s a computer cord and then the confidence level. that it has goes up in the predictive capability goes up, right. So this is literally just just training something, you start with a sample set, it’s called supervised training, right, I’m supervising the machine, I’m giving it a starter dataset of tables in a restaurant. And I’m asking it to identify tables in a restaurant. And in the beginning, it’s not sure. And then it starts learning. And the amazing thing that we’re experiencing right now, all of us in our lives in this, this AI revolution is they’re getting really, really good at learning. And then they start predicting better than a human would, in many cases. And you know, humans get tired, right, these things go 24/7 And don’t get tired, they just start. So so the so the more data that you can put in, the more, the more confidence the model gets, the more accurate the results are. And the happier you as a user, are, it literally just comes. There’s a fancy term I love. You know, I’ve been in Silicon Valley too long, I think. But it’s like, deep learning. It’s called deep learning, right? This is the fancy title. But it really is revolutionary. And it just means that the machines, it’s just math, the algorithm is deeply learning about the environment, it needed you to help it to get going. Once it got going, it picked it up, it picked up the baton, it’s like I’m good, I get it. That’s a computer cord, right? Good job. And then it gets really, really good. And it gets even better at it. And then you just keep feeding it more information. And in a long you go right, so I’m sure all your listeners have heard about chat GPT if they haven’t tried it yet, you know, these are, these are normal models that are just a bunch of data. That’s predicting the next word. Right? And just That’s all it’s doing. These are just prediction machines. I’m just predicting the next thing that’s going to happen. And so computer vision is doing doing the same thing. I don’t know it was was that. Yeah, that makes sense. That was great. That was great.

JJ

Jeremy Julian

26:59

Because I think, again, I think that that while I love the the advent of of Apple and Google and Amazon and all the things that they’re doing for our world, I think that a lot of people think that just because my Alexa can do it, you know, or my Tesla can do it, you know, I mean, Tesla’s really just, it’s a battery with some wheels. And you know, I happen to drive a Tesla, but it’s constantly learning. And it’s, it’s this AI model that’s constantly growing and building on his data set. But people don’t comprehend that the early version of the Tesla 10 years ago, was awful at that stuff. It didn’t have any of that stuff. And now I get into the car, I rent a car when I’m not at town, and it’s not a Tesla, I want to scream because the car does everything for me, you know, right? Well,

MB

Marty Beard

27:41

yeah, yeah, I think we get we get maybe we get a little impatient, we want it to you know, and some of this stuff is, is is pretty hard, right? But at the end of the day, like it’s, it’s despite it, it’s basically large datasets, in our case that we’re talking about images, just a bunch of images. And you’re just getting smarter about what’s in that image. Yes, you know, it’s getting trained, and it’s producing results.

JJ

Jeremy Julian

28:05

And it does take take time to that point. I want our listeners to understand that it’s not something that just you put it in, and next week, you’ve got everything there. It’s got to take 10s of 1000s of images of what’s going on in order to be to be correct. Yeah. What do you what do you think? I mean, you’ve lived in the Silicon Valley world, what’s caused the Advent? Is it that the models have gotten cheaper? Is it that the compute power has gotten cheaper? Isn’t that the data? Like, talk to me about why there’s such a surge? Whether it’s Chet GPT, or any of this AI? It feels like, you know, yes, it’s a buzzword. Nobody wants to throw AI and machine learning and computer vision on top of stuff because they’re humans. But but it really is truly changing. Every day I pick up my iPhone, there’s something new in an app that’s making my life easier and better because of these models that they have. What do you think that is? No,

MB

Marty Beard

28:53

well, you you kind of nailed it, you nailed it. And a lot of the a lot of the reasons so that the revolutionary discovery was was deep learning. So it just it machine vision, you know, this, this type of stuff has been around for a long time, right? But then all of a sudden, there was a way to make machines predict things a lot better, faster, easier, right? So so that that happened, right? And it came out of academia. And everybody was getting really excited about it. It took a little a little while to hit the broader market. But that’s, that’s number one. Now number two is remember these things have to get on something. Yes. Right. So you have this model. And okay, where does it go to do something does it go into a computer, a server, my laptop, a phone, a camera? And and that got a lot easier. And also the hardware this happens in every technology trend, the hardware gets cheaper, it gets more powerful, and it gets cheaper, except when all of us want to buy the next iPhone. Are you getting company? Yeah, exactly. But I mean, but in general, like now, like I have one customer where they’re buying $70 cameras on Amazon. And they, they buy what’s called a, an edge device. Yep, that’s about $300. And they’re good. They’re doing deep learning algorithms on those devices out in the real world and getting the data that they need. So my, my point being that the, the technique got better, the processing power, and the complexity and the cost at the edge got easier and cheaper. And you just combine those two things. And then now people are just, it’s obviously a major trend. And everybody’s putting a ton of energy into trying to figure out what it what it can do. Chat GPT helped in the sense that anybody who’s used it, you have a browser, you ask a question, you get this answer, you’re like, well, that’s, that’s really helpful. So I think it I think, in general, it just made AI look kind of practical for the average person that just sort of, so I think that helped in a way because it’s like, okay, this is real, you know, I thought it was some science fiction stuff. But you know, I can I can now

JJ

Jeremy Julian

31:08

I can interact it, normalizes it, it makes

MB

Marty Beard

31:12

it it does. It’s an amazing research tool. It’s like having your own personal professional Professor walking, you know, so that’s good. So I think that helped. Yes, all these things are combining and here we go New World.

JJ

Jeremy Julian

31:26

I personally love it. And I’m always on the tech forward side of things. I’m gonna ask you a couple more questions, Marty. And then I know we’re gonna get get close to wrapping up. But when we so one of the big stalwarts in the in the industry is a company called holo, you may be familiar with them. At their user conference earlier in earlier in 2023, they put out a video of kind of what they expect the restaurant of the future to look like. And it looks a lot like a digital version of cheers is kind of how I talk about it. And so Marty goes into the same his normal pub that he’s going to or he goes into is from a restaurant they know who Marty is doing, where his wife is, they know who Marty’s kids are, they know what Marty likes to eat, what he likes to drink, where he likes to sit in the restaurant, all digitally to the server giving the cheers ask experience, but using computer vision, the you know, something else to get that data into the server’s hands to be able to offer Marty that experience no different than Google does when I search for Facebook does when I search, or I’m browsing around on there. How far away from that? are we and what do you think the risks are to getting to that?

MB

Marty Beard

32:29

I mean, I love these, you know, companies put out a vision and kind of how they’re going to dominate that. And you know, I’m not I’m not picking on all I’m just saying that. It sounds good. Let’s back up for a second and just kind of what is actually practically needed. Yep. Right. So I think this, this is going to be user driven, not necessarily company driven. And so I think I think, you know if it’s useful, if it’s useful for me to walk into a Starbucks, and they know that I like a nonfat latte. And, and I’m okay with them. Knowing that I walked in, let’s say I opted in. And I’m okay. If they if they do facial recognition. I’m actually I’m getting value out of that when I walk in, then yeah, I think that’ll I think that’ll take off. Right. And you mentioned a bunch of other or all mentioned a bunch of other use cases. Yeah, I think all that’s possible. But like any technology trend, we always talk about, it’s kind of like at the end of the day, what is really practically needed. Yes. Right. And I don’t have a crystal ball to say exactly what’s, what’s going to be practically needed. I can kind of see early. Do I want to understand my customers win? Yes, I think, okay, to what extent Am I allowed to understand everything about Marty when he walks in, and there’s gonna be those issues? But I think yeah, understanding Marty, oh, he’s a preferred customer, all that that’s good. And then can I improve my operations? Through all this new data that I’m getting? Or is it just getting in the way of me doing my job? Like, because I’m looking at a screen all day, right?

JJ

Jeremy Julian

34:00

So where does this predict these things? And I think it’s going to also be be based on culture because I was talking to somebody at this most recent trade show and and in certain cultures, and they happen to mention and again, I’m not this is not me making a this was them saying, and Asia, they’ve been doing some of these things for a long time. They know who you are digitally. It’s not the cheers effect of just you know, Dutch Bros is famous for for knowing who their customers are before they drive through the drive thru line, just because they hire really good people. And they see the same ford f 150. That’s white with a black top that shows up in the drive thru every day. They know who that is. I know anybody that’s a regular Starbucks user. If they go to the same Starbucks every morning, the staff oftentimes know who you are. Do we need that digitally? And if we do get it digitally, does it end up helping? And so it’s just I asked, because it’s this vision that YOLO had put out and I think it’s going to be that juxtaposition, even as a consumer culturally in the United States. Is that something I want them to have access to? Yes, then I want them to be able to do what Right.

MB

Marty Beard

35:00

And at the end of the day, it’s usually not either or. Right the world, the world tends to converge to the middle. So I think I think I think digital will make a lot of sense for certain use cases. And then the human element will really be wanted and other use cases, right, it’s probably going to be a hybrid of the two, like, I’ll give you an example of no human involved. So contactless checkout, right? So all of a sudden, you know, we’ve had increasing amount of experiences with that can be cool. Or it can be really frustrating. If there’s nobody around and you didn’t, you’re trying to get something answered, and you see people abandon, right, and why well, that’s not good, that’s not good for the brand. That’s not good for the customer, etc. So I think this stuff is the exciting thing is we’re in the middle of all this working out right now, which is cool. And it’s gonna be fun to see how it shakes out, I always try to keep my eye on what’s actually needed. And what’s actually helping that restaurant improve their margin. That’s it.

JJ

Jeremy Julian

35:56

I love it. I guess the the last thing that I would throw at you is what are the risks? What are the risks of doing these? These kinds of things? I mean, obviously, there’s the the mega risks that says computers are gonna take over the world. And, you know, Elon Musk has thrown that stuff out there. But I’m not necessarily looking at that. But what are the risks? What are the risks that that people need to be aware of when they look at this type of technology? Is it driving behaviors within the restaurant that might be hurting margins that might be hurting guest experience that may be hurting? The ability to deliver on the brand promise? You may have?

MB

Marty Beard

36:28

Yeah, that’s a great, I mean, I think we got into it a little bit, which is if you’re just generating data for data’s sake, and it’s kind of like people that can never get their head out of their phone. Right? It’s like, you’re just, you know, at some point, it’s just too much. Right. And so I think I think it really gets back to being super thoughtful about, okay, what do I really need to understand, right? And focusing on that, so I do think one of the risks is this kind of data saturation, right? It just becomes too much for an average manager to be able to really deal with and honestly, why are you even asking me as a manager to deal with it. On the other hand, there’s going to be other data that’s absolutely critical. They’re going to they’re going to, they’re going to have to the other thing, I think, is we didn’t really talk about privacy much, but I’m sure your listeners have heard about, you know, you can get these kind of inherent biases that are built into the algorithms. And it’s not the machine quote, doing it on purpose. But it’s, it’s, it’s producing a result that we know is not fair, or we know is not accurate, or what have you. So I think we, you know, we really need to be thoughtful about that stuff. And I think we are going to see some incidents where stuff like that happens. And it has it backfires, right? It’s kind of like, certain customers are being denied service or something’s happening where you just like, Okay, that was really bad. So we gotta, we gotta be thoughtful about that. But I think a lot of people are thinking about that, and trying to make sure that that that doesn’t happen. Those are a couple of risks.

JJ

Jeremy Julian

37:59

Yeah, I love that night. I would. I love your your description, because oftentimes, we talk to customers about, you know, in listeners about how do we get to a place where you’re truly solving business challenges, that you’re truly driving the behavior, because, you know, too much data ends up being, you know, data printout, or paralysis by analysis, and they can’t go anywhere with it. And ultimately, it doesn’t enhance the guest experience or, or drive additional margin drive additional revenue and, and such. So, Marty, I know we got to wrap up, how do people get in touch with you? How do people learn more about always AI? Talk to your team and figure out if it’s the right fit for their brand?

MB

Marty Beard

38:36

Yeah, I mean, I think that honestly, always ai.co Just we have a good description of the of why computer vision, the uses, you know, we were talking about? And then there’s a good description of the of the product. So like, what is this thing? How does it How does it work? And, you know, we love talking to people, so it’s really easy to just ask to speak to an AI expert. It sounds like it goes into some black hole, there’s, you know, hopefully, AI responding to them. It might be Yeah, exactly. It might be me that actually picks up the phone. And it just talk to us. And then we can, we’d love to hear what you’re thinking about. And it’s pretty easy to do like a POC or something like that. And we’d love to, you know, we’d love learning about what people are doing. So

JJ

Jeremy Julian

39:22

I love it. I love it. I’m grateful to be at the forefront of this revolution. As you as you said earlier, I think back about, you know, if I was born 100 years ago, none of this stuff was even possible. I mean, who would have thought that they you know, are probably in the same age demographic, it’s like who would have thought that we would have had a phone in our pocket that had more power than then my kids and you’ll laugh about this that we used to have to go to computer lab to learn how to type and they’re like, you know, those back in back in elementary school, I had to learn how to type and then got to high school computer. So I don’t know. And now my phone is telling Like me what time I need to leave to get to my next appointment.

MB

Marty Beard

40:02

So there’s no it’s amazing. It’s kind of fun. It’s fun.

JJ

Jeremy Julian

40:05

Well, Marty, thank you so much for your time to your listeners, guys. We know you guys have got lots of other things pulling out your time. So I appreciate you guys spending time learning more about what’s going on in the world, the restaurant technology. If you haven’t already done so subscribe to the newsletter, subscribe to the podcast on your favorite podcast player. Marty, thank you for your time and to our listeners make it a great day. Thank you, Zack Spees if you haven’t already subscribed, go subscribe to the newsletter as well as wherever you want to consume your content. Rob And Ben, thank you guys so much for your wisdom and hanging out with us today and to our listeners. Make it a great day. Thanks, guys.

I

Intro

47:37

Thanks for listening to the restaurant technology guys podcast. Visit restaurant technology guys.com For tips, Industry Insights and more to help you run your restaurant better