Artwork for podcast The Industrial Talk Podcast with Scott MacKenzie
Michael Hollinger with IBM
14th December 2021 • The Industrial Talk Podcast with Scott MacKenzie • The Industrial Talk Podcast with Scott MacKenzie
00:00:00 00:17:42

Share Episode

Shownotes

On this week's Industrial Talk we're talking to Michael Hollinger, Distinguished Engineer and Senior Developer at IBM about "Real and Practical AI Solutions to solve Industrial Challenges".  Get the answers to "AI" questions along with Michael's unique insight on the “How” on this Industrial Talk interview! Finally, get your exclusive free access to the Industrial Academy and a series on “Why You Need To Podcast” for Greater Success in 2022. All links designed for keeping you current in this rapidly changing Industrial Market. Learn! Grow! Enjoy!

MICHAEL HOLLINGER'S CONTACT INFORMATION:

Personal LinkedIn: https://www.linkedin.com/in/mikehollinger/ Company LinkedIn: https://www.linkedin.com/company/ibm/ Company Website: https://www.ibm.com/us-en?ar=1

PODCAST VIDEO:

https://youtu.be/lwhAz8qEW-o

THE STRATEGIC REASON "WHY YOU NEED TO PODCAST":

OTHER GREAT INDUSTRIAL RESOURCES:

NEOMhttps://www.neom.com/en-us Hitachi Vantara: https://www.hitachivantara.com/en-us/home.html Industrial Marketing Solutions:  https://industrialtalk.com/industrial-marketing/ Industrial Academy: https://industrialtalk.com/industrial-academy/ Industrial Dojo: https://industrialtalk.com/industrial_dojo/ We the 15: https://www.wethe15.org/

YOUR INDUSTRIAL DIGITAL TOOLBOX:

LifterLMS: Get One Month Free for $1 – https://lifterlms.com/ Active Campaign: Active Campaign Link Social Jukebox: https://www.socialjukebox.com/

Industrial Academy (One Month Free Access And One Free License For Future Industrial Leader):

Business Beatitude the Book

Do you desire a more joy-filled, deeply-enduring sense of accomplishment and success? Live your business the way you want to live with the BUSINESS BEATITUDES...The Bridge connecting sacrifice to success. YOU NEED THE BUSINESS BEATITUDES!

TAP INTO YOUR INDUSTRIAL SOUL, RESERVE YOUR COPY NOW! BE BOLD. BE BRAVE. DARE GREATLY AND CHANGE THE WORLD. GET THE BUSINESS BEATITUDES!

Reserve My Copy and My 25% Discount

PODCAST TRANSCRIPT:

SUMMARY KEYWORDS ai, data, 29th annual SMRP, industrial, michael, talk, conference, model, professional, world, reliability, human being, people, assets, problems, industry, point, train, machine learning, driven 00:04 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 get right 00:22 welcome to industrial talk. Thank you very much for joining the number one industrial related podcast in the world. We are broadcasting live from the 29th annual SMRP conference here in St. Louis, Missouri. wonderful venue, great location, and an incredible amount of professional reliability specialists maintenance, you know, acid managers all here, put that on your bucket list, because that's an important place. If you want to be in the in this career, you need to be a part of SMRP. Also, I'd like to sort of thank Accruent and Fluke Reliability. If I go out to their website like I normally do, and I'm looking at Accruent, I'm looking at their website and its real estate facilities and asset management, gaining insights and transform how your organization manages its physical resources, you cannot argue against that. That is important and Fluke Reliability, great website full of incredible information, provides maintenance and reliability professionals the data to do the job assets by asset wherever they are an incredible organization. Great company, great people, great solutions. Alright, hotseat, Michael, this is take three, Michael, just FYI. I want to make sure that everybody understands that we've tried this three times. And now we've we're on the third time and I think we're recruiting on it. We I think we do. I think we're cruising right now. How're you doing, Michael? 01:38 I'm doing awesome. It's been a great conference. 01:40 Oh, hasn't it? It really, really has. Come on. It's it's been great to get together. It's 01:45 been nice to actually be in a place with human beings. You have no idea? 01:49 Oh, oh, I have idea. Type idea. And it looks just like it, you're right, you just realize how just humans just need that. That interaction. And especially when it comes to reliability. And what we're talking about here at this particular conference. It is it has to be collaborative, they they're dedicated to the education, they want to collaborate, they want to solve problems. It's it's a, it's a it's quite the collegial environment. I like it. Now for the listeners out there, Michael, give us a little background little 411 on who Michael is for the listeners. 02:26 So I have the privilege of being a distinguished engineer in IBM. That means that I helped to oversee the technology underneath the Maximo portfolio for our monitoring application, and our visual inspection application and some edge stuff too. So my, my background is actually computer engineering, I've always kind of walked that boundary between hardware, firmware or operating systems and systems of record and all that kind of building up. So through my career, I've always kind of moved up that ladder of just stuff to the point where now it's AI and it's just fun. And the hardware is like way over there somewhere. You know, 03:07 it's it is interesting, we're going to be talking a little bit about AI and and really try to get some clarity into that. But the reality is, is do you find just the business that you're in. There is a velocity, there's a speed, there's a in it, for me personally, human side trying to keep up with it. It's hard to see what's happening out there from that just speed perspective. I can't keep up with it. I mean, that feels stupid. Every day club 03:36 that I have on my on my desk. I have sticky notes for all the stuff I have to do. Right, right. And one sticky notes is the power of yet like that. Not kidding that that is good. I gotta write that down. And that that word yet? That's an empowering word, right? Because it means that I don't get this at this point in time, but I'm gonna get it eventually. That is literally how I live my life. 03:59 But it's great, because it's free. It's sort of It's nice, right? 04:05 I don't understand how this thing works or why it broke the way it did. But we'll figure it out. That is very much what my job is. 04:13 See, that is just absolutely incredible. Now you had a you, you were running a workshop, or did you do what do they call it? I 04:19 don't know. Yeah, I did a talk in the Yeah, and one of the tracks right. So the whole idea was I'm trying to show how AI can help you with leveraging the data. You've already gotten your enterprise because there is so much stuff we talked about in you know the the ability to leverage AI to go solve these great problems to even go just see commercials. You can see examples in movies, you can see all these things, and all the Sci Fi stuff that happens. And the reality of that. It's hard to tell sometimes what's possible and what's not. Right so it's hard to tell what she could even do. So what I actually shared was a couple of different frameworks to pick up and kind of decompose the AI 4.0 or industry 4.0, or this AI thing or these AI projects and say realistically, what are you trying to do? Right? What what is the intent behind why we're talking about this, it's not necessarily to fix a bearing it might be to improve the quality of water for a given area, it might be to deliver a higher, cleaner power signal and might be these different things. The bearing thing is like a means to an end for right, in the same way that AI model is a means to an end for this solving this problem. So that's a lot of what we do, it's trying to decompose it. So in the talk, we walked through the ways of kind of picking this apart, picking apart what the intent is for the projects, looking at what this whole system might look like when you put it together. I shared an example from one of our clients and partners have this rolling freight thing from a train, right? This is one of the coolest thing in the world. CN Rail, can run a train through a shed in the middle of nowhere in Canada, and high res images of that train that specific vehicle with a specific stock behind it will be taken, analyzed, run through a bunch of different AI models and much different rules. And it'll trigger work orders saying, Hey, we found this, go inspect it, or hey, there's this that happened right way really, it's not? I would not believe two years ago that was possible. Right? And that's, that's totally doable. It's and I love that thing. So a lot of my teams that I talk with, we try to figure out how to differentiate the science fair experiments, or the really weird dreams that someone might have is a fever dream, from what's possible, right? And to be honest, a lot of the things I start with, I'll say that, we can do that. And we'll sit down and maybe a day or two later, oh, well, you know, maybe that worked. But actually, that actually, you know, has some legs to it. So let's keep going. So we talked about the different stages of projects and that kind of thing. 07:09 See, outside of the fact that I'm going to steal fever dreams. Danger, because we've all had fever dreams, and they are wacky, right? But But you live in a world like that it you know what you're saying the power of yet, you live in that world where you have to have those conversations. It's almost you have to be audacious in a sense, and then be able to sort of like, like, I think Kevin Clark was talking about Amazon, they have audacious, you know, goals, you have to whether you at least you begin to have that journey toward it. Right. But that's, that's what I see a 07:52 i I see. That's the entire thing, right? Everyone can kind of see that there's a thing there. There's there, there. There. There is a there, there's there. So then, how do you get through to that? Right, right. So a lot of our clients and a lot of the stuff that IBM does is is infusing AI into the systems, right. So when you talk about infusing AI, what does it really truly mean? That means I have to have the ability to collect data, I have to have the ability to analyze AI to be able to label it train models to validate most to say it's gonna be correct and trustworthy, to update those things. And all the infrastructure that goes around that is necessary for that to be more than just the science for experiment. Right? You have to have those parts to put this into production and make it valuable for that human being who's just trying to go do a thing. 08:40 See, I hear what you're saying. And I think I, I'll be the first to geek out on this stuff. Absolutely. Now, I, I take a situation where let's say, I apply an AI solution. Here's a motor and I'm down on the shop floor, there's a motor, and that motor is a unique motor. But this is another unique and that has other parameters associated with the data, and another, and then another and then another. And how do you take that AI mindset and be able to apply a solution across? Didn't I can't say disconnected, but different assets that have different performance, you know, standards and all that stuff? How do you do that? 09:21 You have to know that you have to know that they're that those are different, right? You have to know that you can't just take a model that you trained or a system that's built around this one asset and know that oh, yeah, it's gonna work. What I did in Iowa is gonna work great in California, Walt right. It won't work. It won't. And that's that's the pitfall people fall into is they'll they'll build something and they'll do a proof of concept. We'll do a pilot but with the other some some demo. And that demo will be such a tightly constrained environment. There's actually a word for that in the space, right? You'll actually hear this word. It's called overfitting. So anyone's data driven right. You overfitting your data set. So if I have a bunch of data points, right, and I'm trying to figure out good or bad, right? There's, there's a bunch of ways you can fit a curve to even draw straight line draw a second order polynomial, whatever, right? At a certain point, I can give you a 35th order polynomial that will directly skewed between all the different points. And it will be flawless for how you test it. But the reality of the world is not that No, right? So that kind of stuff comes up. And you have to know that this particular motor or this particular bearings, failure, so this system will have these characteristics, and that you should be able to understand these things about it. But I can't necessarily take the data from here and apply it to over here, right? I have to retrain the model from here to here. There's even a term for that to transfer learning. transfer learning is for learning. We're gonna make a smart, we're gonna make everybody smart about this, right? 10:55 You have to be okay, so I hear I hear what you're saying. And I understand and I get it. But there are zillions of assets out there. I don't even think I'm exaggerating, because and I think that for AI, you have to really want to target where you want to go. To your point, what do you want to try? What are we trying to do? But tell us or share with us what this is where I see machine learning. If I can get something to sort of learn, and then learn and then learn, then it's 8am I on something? 11:31 Absolutely. So we talk about AI and ML, right, right. AI is this general space, right? That's trying to get a system to understand and behave and act like a human being and simulate, like the kind of insights you might get out of a human being. Machine learning is a specialization in that, which is teaching a system about some data to make some observation or make some prediction about the future. That machine learning process that that thing is driven by data is driven by the ability to go pull data from the field, and build a model, have that representation, and then make predictions based on that model. Right? So if I take that take all the AI stuff away from this for a minute, right? I'm building a predictive statistical model that can predict something about something. So given some input, here's the output. It's basically a bunch of linear algebra, literally as much linear algebra. So when you look at it from that perspective, it's given that data, how do I create something that can given this information produce something useful? So now the question becomes the data. When I talk about like pump a pump, a pump B to pump Z, or pump, quadruples, Z key, each of those may or may not be different, each of those may be part of a broader data set. So managing that information, managing the flow of that information from the field, or into wherever you're doing the training, or doing the AI stuff, right, or in that platform, that's trying to help with predictive maintenance. That becomes such a key part. Because you have to be able to build it off of real life information. They can't just be simulated can't just be the one to tell you. Right. Right. Yeah. 13:13 Why is this important? It should industry? Do it? 13:19 It's about it? Well, let's take why why are we talking about AI in the first place? AI is to help deliver those insights that a human being might have, but at a different scale? Right. So why is I think so when you can answer with work? Can I drive efficiency or work? Can I help be a force multiplier for good here? That's very much how we define what we're doing. So when I look at, like, the stuff we're doing In Maximo around visual inspection, or on our edge stuff around our monitoring stuff, it's trying to say, We want to help be that force multiplier for good in this space, to say, Hey, by the way, I noticed this thing. So we have actually a client that's doing quality engineering there, they're doing, they're one of the marquee brands, right? That that produce high quality vehicles. They are using our stuff to try to notice and fix problems ahead of time, before they sealed up certain things, right. And they're doing it all from a smart device that, you know, has a bunch of sensors or whatever, right? And that's cool. I love that force multiplier for good kind of thing. But that's absolutely. Why this is important. Because if you look at it from a different perspective, for a sec, if, if you're not paying attention to that efficiency gain, then a there's a competition that might might be happening or is probably happening in the field, but then be you're losing out on a chance to try to improve things and I think that constant improvement, like continuous improvement, that's That's why people come to conferences, right? That's literally why you come here. Exactly. Yeah. So yeah, so like if you kind of buy We want to be improving over time. This is like, this is another way to do that. It's not one of the tools in your toolbox. It's not the answer for everything. But it's one of the tools you should consider when applying to solve a problem, right. 15:18 You know, the, the world that we live in this, this is so intriguing. I'll give you an example. I was at a conference, pre pandemic, a couple of conferences. So I'm at a conference, we were talking AI, and talking. There's a body of people talking AI.