Jorge Maia and Rebeca Sousa with Crazy Tech Labs

SUMMARY KEYWORDS

ai, iot, problem, twin, solve, omg, good, data, put, digital, solution, copilot, work, industrial, Rebeca, man, prompt, important, listen, device

00:00

Scott, welcome to the Industrial Talk Podcast with Scott. MacKenzie. Scott is a passionate industry professional dedicated to transferring cutting edge industry focused innovations and trends while highlighting the men and women who keep the world moving. So put on your hard hat, grab your work boots, and let's go all right once again, thank

00:22

you very much for joining Industrial Talk, and thank you for your continued support of a platform that celebrates industry professionals all around the world. You are bold, brave, you dare greatly, you innovate, you collaborate, you solve problems, and that's why you're making the world a better place. Thank you very much for what you do. We are, once again, speaking of problem solving, we're broadcasting on site. Reston, Virginia is the location, and we are broadcasting from the OMG q1 meeting. It's not a conference a meeting, just keep that in mind. And it is a collection of problem solvers, and they are passionate about getting things done right, to make the world a better place, in the hot seat, we have two Jorge. Did I say Jorge, right?

01:05

Yeah, that's okay.

01:06

Oh no, no, it's not okay. That's No.

01:09

Tell me. In Brazil, let's talk

01:12

Jorge. Jorge, yeah, Jorge and Rebeca, I got Rebeca. Rebeca,

01:16

okay. Rebeca, is right.

01:18

Is it okay?

01:19

Is there? Okay?

01:20

Okay.

01:21

You gotta keep me. Keep me honest, right? Crazy Tech Labs. We're gonna be talking a little bit about AI, how it applies to IoT as well as digital twins. So let's get cracking All right. Are you having a good meeting? Yes, yeah. What makes it good?

01:41

I guess, listen, listen to real problems. And it's a good thing here, because every, every aspect of a congress or a conference is about things, PPTs, Dax and other stuff. Here we see the real thing and how it's going on. Hey, I have problems here. I try to do this that's so exciting. It

02:07

is. Every time I go to these events and I I look into the sessions, right? And everybody's just around a flip chart, or they're looking at something, and everybody's just scrambling to write things down. I just think it's such a energized environment. I really think it's pretty cool. Before we get into this topic, I need, I need to know a little bit about, okay, Jorge, Jorge,

02:37

that's good. Give

02:40

us a little background on who you are, and then you're next.

02:44

Rebeca, okay, I've been working on on it since 1995 so just a little bit later, man, I'm a computer scientist and a master in mechatronic systems, now pursuing a PhD on mechatronic systems, also doing IoT since 2014 so I started IoT in Brazil when the people said, it's crazy. It's really crazy. What are you doing? You leave all the software behind. And I said, No, I'm not doing this. I'm applying software with hardware and doing some good stuff for digital transformation. Man and we founded the Crazy Tech Labs over there. Now we are in Brazil, here in us, and with projects in Europe, also doing IoT, digital twins, AI applying AI in real things. So computer vision, normally, detection, prediction, machine learning, some good stuff. Man. Rebeca,

03:53

little background,

03:54

okay, my background, I studied a little bit after jars. You know, I'm a little bit younger, but probably, and my degree is in mechatronic engineering. And just after I finished my graduation, I find, I found Jorge, you know, I decided I wanted to work with IoT. I did a little bit of IoT stuff things in the graduation process, and I liked that very much. And just after I finished that, I found Jordan started working together at Chris tech labs, and it's a very nice that opened my eyes for a lot of things that I didn't know, including digital twins. And here we are now, exactly in the event that's talking about that and bringing and theory to real stuff. You see a lot of real projects that take theory in putting on action. You know, it's very cool. See,

04:58

I love that. Definitely. Yeah, it must be nice. How long you guys have been part of OMG,

05:06

since beginning of last year.

05:08

So you're relatively new, yeah, yeah.

05:11

I met Dan in a conference in San Francisco, and we are in the middle of this trip, and I talked to him and said, We have to talk more about digital twins. I heard your session, and I believe that I can share some good stuff with the consortium. And he said, man, come in. Let's go. Let's do good stuff together. And

05:36

since that's a that's a table being moved around so disregarded. It's not somebody booing him or anything like that. Jeez. Let's get into that. Because when we first just spoke offline, we were talking a little bit about IoT. IoT, that's that's important the other areas, of course, digital twin. And there's been a lot of conversations around digital twin. I think that there's a real interest in and how do we align these, these very important innovations in technology, with AI. Give us a little sort of rundown, Jorge, a little bit about how you see that playing out, that AI component. Yeah,

06:21

this first thing Scott is, every time we get into a solution or a project or a problem, we have to talk with our clients about some basic stuff. Like IOT is not only connecting things. Digital twin is not only creating a 3d representation model. And AI is not the silver bullet. That's that's the truth. Don't

06:47

you find that people, they're looking for that silver bullet. I think that that's a great example. Everybody's

06:52

looking for something that solved all the problems with one line of code. Yeah. And, and cheaper. That's the that's the crazy stuff. And one, one thing that we we saw on the field, is about man, we really have to solve real problems, and in some cases, real problems are solved in simple way. So we do not have to change anything. We do not have to exchange protocols. Just take it all together, see how we can merge the ideas and how to put it on a line. So we have to prove results. We have to get results with that. So one, one thing that we we are doing is too much research before each each solution is problem, because there is some some good stuff that the people is doing on the market and in academia also, and we have to listen the others first. And I see in other, in other ways that the other companies doing some good stuff also, but they start doing during hands on the problem. And we take a step back and said, Man, we have to just take a look on the market. Let's see what's going on. Let's listen about what digital twins consortium is talking about that. Let's listen about open AI. Let's listen everyone on the table and said, Okay, we have this kind of technologies that we can use, and in some cases, we just apply a simple solution to solve a problem, and do not change PLCs at the very end, not change protocols. We just put an edge computing and solve a problem on the on the field. And we do not have to use Cloud in some cases, and in others, we process everything on the cloud and do the AI inference and the cloud, and it's cheaper than I put a GPU on the on the on the edge. So that's that's what we do in our daily job at Crazy Tech Labs. Look at, how can we solve the problem without putting costs ahead and look into our customers, our clients at the very end and say, I have to solve your problem. You know that? I know that. How can we do it together? How can you help me to put this line ahead a little bit? How can I solve this using a technology that we already on the market, and we do not have to rewrite some code, and we have this today. The problem that I, that I see today, is everything is really simple to do at the first vision of the use of technology. So open AI. Open. AI can solve everything. I can just plug the data and it solves that's okay, lies very, very good. But maybe it can give me the wrong answer. And if give me a wrong answer about an OEE on a factory, if you give me a wrong answer about a. I stop time on a convey man, what I'm doing is worse than I have technology. So it's better become humans looking the field and detecting patterns, as we had since 60s, 40s. It's better than I put a wrong algorithm on that and solve a problem on the wrong way. That's, that's, that's interesting,

10:24

because you bring up a really interesting challenge that that needs more attention. And is that, I know that everybody is talking about AI, right? Yeah, it's just all of a sudden last year, it was like a switch, and everyone's like, Yeah, this is going to solve it. But nobody really understands or watches. What is it? What is it, right? How do you how it works? How do I trust it? Because, because I know that OMG is all about that trust. That's why I like it, because there's this human vetting of information you agree with that there. Yes, yes,

11:00

I agree. Like you said last year, AI just became something that everybody talks about, not just in universities or big companies industries, and that is a little bit dangerous, you know, because people say whatever they want, and that's why it's important to have consortiums like this one, then we can put some standardizations. And I believe that AI is not going to solve everything, or maybe became conscious, you know, yeah,

11:41

yeah. Not tomorrow, yeah. Let's go. But

11:46

you you need to know how you can use that in your daily basis. It can help you, really, will save you time to solve a lot of tasks and bring you insights in your project, in your solution. So let's put our fingers or foots on the ground.

12:04

You have to be engaged. You have to figure it out.

12:07

Let me contextualize an example I saw, in some cases the people talking about computer feature. And let's put the camera and using a convolutionary network to do the understanding of patterns on the very end, blah, blah, blah, blah, blah, just a second, what we are looking to solve. What is the problem? Oh, the problem is, if I have a box this size, and if I have a box this size, a smaller one should be not in this convey should being the other one. Is that? Okay? Good. Why I need AI to do this? If I have the frame of the camera, I can apply a simple algorithm using CPU in a small computer at the very end, over there, and then I can detect that the size of the boxes is more than the other one, yeah. And I do not have to train a model to use that. But the problem is, everybody is listening at tons of information, YouTube, Tiktok, all those. Oh, yeah, open AI, the chat, and then when they come to the table, let's use open air. Yeah. So why?

13:25

You have no idea what that means? Yeah, that's

13:27

the same as the people is talking about copilot for development. Copilot for development. Place of view is amazing, but you need to know what you are asking for. Copilot, if you don't know what is an objective oriented pattern, if you don't know what is a standard man, you are asking for something that you can have the ready answer and you don't know if the answer is correct or no, it will compile. You can send it to a via pull request, and you can put it but, but we should align, yeah, and it's working, maybe, but it's correct. No,

14:07

see, this is interesting. And we'll get, we'll get to the length of how AI has been at least leveraged. But even in my short period of time that I get involved with that generative AI, I'm interested in the prompt. Who? How are you writing that prompt to get the results? Because I might, I might think of it a different way, like, give me the summary of whatever. Somebody might write it a different way, and then the result is completely different, yeah, from the prompt. So I'm always intrigued with the prompt. It never the result is, is sort of secondary to that prompt. We

14:48

are living the same, the same problems that we have in the near past. I can talk about five or eight or 10 years ago. I. 10 years ago, the problem was, what kind of question and how can I elaborate a question to put on Google, to answer me and to have the first page and I can respond for you and for anything that it is the same problems that we have today about creating props and the people is talking about a prompt engineering, new job, blah, blah, blah to know we are talking about. You should know what you were asking for. Yeah, that's the same. The question is, what is a question?

15:36

Yeah, that's exactly correct.

15:41

Yeah, yeah.

15:42

I I just think, how do we now because, because AI is so I'll just say new it is. It's year two, year and a half, whatever. I don't care. It's still relatively new, and I still believe it's still at a sort of the tip of the iceberg where we're at. But, you know, digital twin and IoT has been around forever. You know, to from my perspective, I've been collecting data off of assets and then making decisions. How do you see that whole AI component impacting IoT and digital twin?

16:20

It's another tool, on my point of view, it's another tool that we compose a solution. Because if you, if you look at publications and articles and anything else in the past seven years, you will see that we do not use IoT term anymore. We are talking about connected factories. We are talking about real time data from the fields, smart fields and others, right? And it's becoming transparent, and it's good. The people said, oh, IoT is dying. MSM, no, I guess not. I guess now we need context, because AI is about context. So if you're talking about a copilot, you need context. You open all your codes. If you're talking about a copilot running on offset three, six, on Windows 365, or on Azure, or anything else, you need context. And context is IoT, digital twins is the idea of you can aggregate all of this inside services and AI, and is the cherry on the top, because it can make things that we take hours or days or months to do, and we can now do in a comment, in a prompt, but We can do it in a prompt if we know what we are talking about. Doesn't make sense for me. I keep talking about how to create a to d graphics to use on that it will generate a Yeah, it will work. Yeah, it accomplished. No, that's

17:58

Rebeca. How do you how do you do? And that's that's perfectly well explained, explained right there. How do you deal with data quality?

18:09

You know, in acquisition,

18:11

if I'm a, if I'm a deep, digital twin person, and I'm trying to aggregate all the data that exists out there, out there, wherever out there is, how do I even trust that that data is accurate, and therefore it gives me an accurate representation of my twin? What do we do there? Because if you're going in and helping them, that seems that's important.

18:36

It's very important. There is a lot of approaches to the time the data is correct. If there is something wrong on the position, no one thing is. You can use AI for that, you know, I IoT will bring you, you, you it will have only a lot of historical data, you know. And you kind of, you find a pattern on that. And by that pattern, you can see from deviating from, you know, attending on the time series, you know. And you can use AI to do that? Yeah, it's a way to

19:23

do that's, that's the key, if I can, if, because nobody likes nobody, and I think maybe there's a rare person likes to scrub data, yeah, another

19:36

good stuff, another good stuff on this, in this topic, in this subject is many, many people is looking for the data on and the data, in some cases, was generated by a device. And then on the device side, we need to look at the life cycle of the device. Is about the authorization of the device, about the authentication of this device, about the security of this device and the OT network. We had problems here in us on the oil pipelines, yeah, because the OT network was compromised. So that that was that was a mess. And man, in some cases, when we arrive into a client and we ask about who take care security on the devices, and in some cases, the people said, What is security on a PLC? And they said, who have access? Anyone. The login is on the on the PC, so you can just open the PC and over there, and then you can log in and strike the tape. It's not only for strike the data. So, no,

20:54

no, I don't want to hear this. You plug in your ear. No way. That's not secure,

21:01

and that's why the contextualization is so important. Because when you know where the device is attached with and how it should behave, you can know if it's going wrong or right. So just a bunch of numbers would mean anything if you have contextualization, it can help to find issues in all those stuffs. That's why digital twins comes together. See,

21:29

and I hate to I'll go back to AI real quick, if you can apply an AI solution to cleaning and scrubbing the data and identifying inconsistencies or consistency, or making sure that I have a greater confidence in the quality of the data, therefore my results will have greater confidence I'm all in, I don't know is that is, that's where? Is that? Where it's going? And

21:54

this is the good thing about the AI today, because you can use that to detect a problem that you take maybe months, years or never,

22:04

never. It's never, trust me, it's one year, then it's never, we've all been done that. Now, one final question, where do you see it? What's that future look like? You've been involved in, OMG. You've been listening to a lot of the individuals talk about their challenges and and how the you know, trust is very important. Where do you see all of this going?

22:33

I guess, I guess the the outlook of the development of IoT solutions, and I talk all it to solutions, as the whole bunch of of the solutions, using AI, digital twins and others, is become more simple because we have more tools to use on that and on a perspective of the designing solutions, it becomes more simple and a perspective of the owner of the site and the plant and the fabric, it will become more complex, because we have tons of new solutions coming daily to the market, and the choose and pick the right one will be the problem. See, I

23:20

think you're, you're touching upon something that's real important and and for me, I'll just use me as an example. I'm, I'm a manufacturer. I don't have the bandwidth to to make the right decision. I need individuals. And this is truly an education component, but it's a collaboration solution, because I can't find the right people. I need the right people, because I need to innovate in a way that helps my business succeed and create some level of resiliency in that business. All right, we're going to wrap it up here. Rebeca, how does somebody get a hold of you if they say I want to talk to Rebeca, what's the best way

24:01

someone talk to me? I believe the best way is maybe Instagram. That's the easy

24:09

way. Instagram and LinkedIn. Okay, got LinkedIn. I'm going, Oh my gosh. Now I'm just like, speaking of stumbling and going, you know, I have an Instagram.

24:18

Yeah, because it's easier to write the name, you know, but you can do this is the more professional way it

24:26

is. I'll have your link out there. Jorge, link, I think is the best one. There it is. Man, easy peasy. You guys were absolutely wonderful. I really enjoyed this conversation. Thank you for having us. All right. All right. Listeners, we're gonna have all the contact information for these two individuals, these professionals, that you need to connect with out on Industrial Talk, again, we are broadcasting from OMG. Go out to omg.org. You need to get engaged, because it's all about education and collaboration and innovation, and this is the location for you to be able to make that happen. A lot of people here. So stay tuned. We will be right. Back

25:00

you're listening to the Industrial Talk Podcast Network.

25:09

That was a great conversation. Really enjoyed that. OMG. Was the event Jorge and Rebeca Crazy Tech Labs. We were talking about merging AI and, you know, digital twin to solve real problems, throw an IoT in there too. Solving real problems, make it happen. It doesn't have to be perfect. Move forward. That's my, my call to action to you. Move forward. All right, reach out to them. They're going to have all the contact information out on Industrial Talk. Their stat card is stacked. All right, we are building a platform. You have a podcast, or have a desire to do a podcast, like have your own branded podcast, you need to talk to me. You go out to Industrial Talk, and we have a conversation. You have technology that you want to try to amplify, go out to Industrial Talk. Have a conversation with me. We want you here at Industrial Talk to be successful, because we're all about educating, we're all about collaborating, and we're definitely all about the innovation to help industry succeed. Be bold, be brave. Dear. Greatly. Hang out with you as two professionals change the world. We're gonna have another great conversation shortly. You.