AI • IOT • NETWORK EDGE

Christina Cardoza: Hello, and welcome to the IoT Chat, where we explore the latest developments in the Internet of Things. I’m your host, Christina Cardoza, Editorial Director of insight.tech. And today we’re talking about industrial AI automation with Camilo Fadul and Curtis Thompson from Emerson, as well as James Green from Intel. So, and they’re all joining us from a studio, so we have a panel of expert guest to talk about this conversation a bit more. Thanks guys for joining and for getting together all there.

James Green: Yeah, thanks for having us.

Camilo Fadul: Thank you for inviting us.

Christina Cardoza: As always, before we jump into this conversation, just want to get to know a little bit more about each of you all. So, Curtis, I’ll start with you. What can you tell us about yourself and Emerson?

Curtis Thompson: Sure, yeah, so my name is Curtis Thompson. I’m one of the Product Marketing Managers here in our DeltaV Group. So, DeltaV, of course, is a distributed control system that is used across industries in the industrial-automation space. And my primary focus right now is around our AI strategy.

Camilo Fadul: Great, so if I go next, my name is Camilo Fadul. I also work in the Product Marketing team for the DeltaV solution with Curtis. And we have been working on many, many of these different initiatives, but at this point we’re putting a lot of priority and effort in what we are doing with our partners, and finding opportunities to apply AI technologies in the process-control industries, of course, through our DeltaV DCS solutions.

James Green: And I’m James Green. I’m with the NEX Business Unit of Intel, and my role is centered around edge AI applications and software. Been in my role now for just over a year and a half, and I partner with Camilo and team around the AI applications and software.

Christina Cardoza: Great. And like Camilo and Curtis mentioned, DeltaV platform, that’s their automation platform. So we’re going to learn a lot about that today, especially how that’s helping apply AI automation to some of these manufacturing industries with the help of Intel. So glad to have you guys all here.

James, I want to start this question off with you first. We’ve been seeing AI evolve and go into multiple different, if not all, industries, improving and transforming operations. And now it’s coming into the industrial space. It’s already been in the industrial space, but now there’s work being done to include it in automation platform. So, what can you tell us about the role of AI in industrial automation?

James Green: Well, right now the role of AI is really diverse, and it’s customizable to the specific use cases. So, we’re relieving a lot of pain points for customers in several ways, and that’s mostly around process control, quality control, and that kind of thing.

Intel is well known as a chip manufacturer, but we are in the software space and the AI space, and we’re directly innovating right now in the digital twin marketplace, along with the computer vision applications that we’re developing. And we’re also developing a lot of ML Ops–type applications as well, which is my role at Intel.

Christina Cardoza: Great. Now, Camilo, I’m interested in hearing from you the use of automation in these manufacturing industries. How have you been helping customers apply this and use this in their operations in their factories? And what is the benefit of making these solutions AI powered?

Camilo Fadul: Great. Yeah, so in the manufacturing and automation industries our software is used so that we can help our customers produce their products. So we help our customers help their customers. And in the case of applying artificial intelligence to this industrial automation, which is—it cuts across many, many different industry verticals. We are being enabled by the regular evolution of what started out in the industrial revolution.

So now we are in Industry 4.0—we’re talking about Industry 5.0—but really what has triggered all this explosive growth is access to data. So, data becomes the foundation that any AI model can consume and then turn all that data into actionable information so that we can find places where AI can help our customers in predictive maintenance, in quality control, process optimization, workplace safety, and many other areas that may require repetitive actions but can be empowered and enabled by the large availability of data that we are enjoying right now in the moment that we’re in in the industry.

Christina Cardoza: Absolutely. And you’re right, we’re certainly already looking towards Industry 5.0, so anything that can help us become more intelligent, or any other solutions that can help manufacturers —it’s always something that they’re looking for, so it’s good to see all these benefits that this space is bringing.

Curtis, I’m curious, because manufacturing—we’re talking about working with different machines on the factory floor and making sure that product goes out safely. At the same time we want to make sure we’re implementing these things in a safe and secure way, and there’s even some manufacturing environments that are more highly regulated than just your average production manufacturer, like the pharmaceutical industry.

So, I’m just curious, what are the challenges or the obstacles that you see manufacturers face when they’re trying to implement AI in their processes?

Curtis Thompson: Yeah, it’s a really relevant question, and I think, again, the first challenge that Camilo mentioned is the, you know, do you have a data platform? Do you have a place to aggregate data? Is it clean; is it usable? So, I think once you kind of get past the: Okay, I have—I have data, it’s accessible, it’s clean, it’s usable. And we’ve made a lot of progress just in the last five years around software solutions that can help enable that.

I think you touched on, regulation is a really big part of it, and specifically in the pharmaceutical industry. I’ve spoken with some experts that say: Look, some of these deep learning techniques that are certainly at the core of AI—it’s kind of a black box, and you can’t really explain what’s happening inside that black box. And so, I think we’re in this situation now where at least some of these AI technologies are way ahead of regulation, and regulation is lagging.

And so, in a GMP facility that’s—they’re making some lifesaving drug, they can’t deploy those things. Now, can they use it around the process? Yeah, I think there’s definitely room and opportunity there, and certainly our customers are looking to do that as well.

Camilo talked about workplace safety and some of those applications. I think another big challenge is around data privacy and intellectual property. And then especially, again, in the pharmaceutical industry that’s highly sensitive data, a lot of it. So, I think that’s another challenge that has to be addressed.

Christina Cardoza: Yeah, absolutely agree. And there’s almost two sides of this coin here. The amount of data valuable—useful data out there for us to use, and to derive all these insights from AI really helps us sift through that data and derive that value out of it. Otherwise, it would just be an overwhelming amount. But, like you mentioned, you have to make sure that the data is available; it has to be clean and usable and it has to protect users’ privacy.

So, James, I’m curious, from an Intel perspective, how can we mitigate some of these risks, or what do manufacturers need to be aware of when applying AI to some of their automation systems, and how can they overcome those risks?

James Green: Yeah. The chief one that comes to mind is reliance on technology—that there’s this balance that manufacturers are trying to place in automating their processes and procedures and realizing a value for that, while at the same time placing so much technology into the workflows that you become: what if something breaks? So that’s why customers are really interested in mitigation in terms of adding that person in the loop with the technology so that there is a person that’s there to kind of oversee those processes, or there’s an alarm that goes off that causes the human to interact with the technology if needed.

And so that’s the biggest thing I’m seeing in the marketplace. And then there’s just out there, there’s concerns around job displacement, right? So, you see more and more automation, that usually is going to mean somebody’s job is potentially impacted by just the sheer focus of automating some of these processes where it used to be a person stopping, maybe a manufacturing process for a quality-control issue or something like that. Now it’s being done by a machine. So, a way to mitigate that is upscaling your employees in a lot of ways to gain additional skill sets. Maybe they’re involved in the technology somehow, and learning those skill sets to kind of manage the technology that these manufacturers are using.

And then there’s ethical concerns that a lot of organizations are facing in this space because of AI. It’s, is there going to be a balance of bias? How do you manage that? And so that’s why it’s really important to have that. Again, it goes back to the second point there of having a person in the loop helping with that to make sure that those things aren’t realized in an organization. So those are the top three that come to mind.

Christina Cardoza: Yeah, absolutely. And to your point about the job displacement, I know this is a fear to many workers when a company starts integrating or using AI, but I think at the same time there are a lot of industries facing a labor shortage and there’s not enough people to be able to do the work, or people are retiring. And so it’s such a specialized skill to get enough people to do this and to maintain it. Not to mention that manual efforts are a little bit error prone.

So automation, I see, is helping in these—helping address some of these pain points. And, like you mentioned, upskilling your employees. I know Intel has a solution, Intel® Geti™, where it’s actually bringing business users in to do some of this stuff with the manufacturers to really be the domain users of what to look for. And I want to get into that in just a minute.

Before we do though, Camilo, in the beginning you and Curtis talked a little bit about DeltaV, so I want to learn a little bit more about that automation platform. How has your—you’ve worked with manufacturers and users to apply DeltaV, and how are you bringing AI capabilities into this solution?

Camilo Fadul: So, DeltaV has been from the beginning a product that has revolutionized the industry, because it has always been designed around the premise that it has to be easy. And this is fascinating to me, because DeltaV has existed around in the industry since 1997 or around 1996. And from those early days of DeltaV we started working with life-science industries, with the oil and gas industries, with metals and mining, pulp and paper, food and beverage—you name it.

So, the applications are so diverse, so it is the potential for our customers to find helpful solutions in applying our software. Early on we did introduce some technology that we call Neural Network. And we call it a Neural function block. So, back in the day, it was, to Curtis’s point earlier, truly a black box that it was difficult to understand. And now that we have so easy—so much easier access and better understanding of the potential of artificial intelligence, we continue working on evolving those capabilities and finding additional opportunities. So I think that, Curtis, you have also very specific examples of where we have been able to expand those capabilities, and maybe you can tell us a little bit more about it.

Curtis Thompson: Yeah, so now we’re all experts on what DeltaV does, but more specifically our customers are deploying soft sensors, using AI to deploy soft sensors. And so, what is a soft sensor? A soft sensor is basically a way to replicate and predict what the actual measurement is of a real sensor. And so what that does is you’re modeling some process variables, as few as four or five process variables, to then be able to predict what a measured value is.

And so we have a customer, a particular customer, who’s deployed 20 or 22 or so soft sensors that are basically replicas of real sensors. And in their application they’re making a product in a continuous process, where if a sensor goes down and they’re not actually taking lab samples, running it back to the lab every six to twelve hours, they could be running for hours or even a day. And if they’re out of spec, they’ve got to throw all that product away, which is a huge waste, huge cost.

And not only that: there’s also the ability to validate the lab samples and measurement data. And also, in some cases, they’ve actually gotten rid of those sensors altogether because the models are so robust that they don’t even replace the sensor. Like, we have such confidence we haven’t updated this model in five years. They just deploy the soft sensor in that spot. So, that has huge savings on maintenance, huge savings on not having to be, to have downtime, and just kind of having confidence and validation of product quality.

Camilo Fadul: Yeah, I think that it’s important also to highlight something that you just mentioned. They have been using these soft sensors for five years, and it’s important to clarify that these are proofs of concept. So, when we think about the speed at which AI is evolving and progressing these days, and you have to think that applying that to process-control industries, we are running a proof of concept for five years that for other industries is unthinkable, because what we are trying to do is to go as fast as we can.

Well, we need to make sure in our industries that when we put these models out to the general public, to all our customers, and they can be trained and apply their own processes, they’re going to be there for the next 20 years. So that is a different challenge to what we see in other, maybe less restrictive, industries where AI keeps on speeding ahead.

James Green: Yeah, I’ve seen proofs of concept be 30 days, 60 days, and five years is a totally different ball game. That’s a long time.

Camilo Fadul: Right, right.

Christina Cardoza: Yeah, and it’s great to see that DeltaV has such a long history in the market, because manufacturers or businesses—anybody who is applying or investing in these types of solutions—they want to make sure that they’re future-proofing their investments. So it’s great to see how the company and the solution have been able to evolve along with the industry and still provide the capabilities that the customers need.

You guys were beginning to talk about some customers that have been using the solution and the proofs of concept. I’m just curious, are there any other customer examples or case studies that you guys want to talk about that really highlight the impact of using DeltaV, having it be a future-proof investment and asset to these manufacturers, and that really highlight that impact and benefit of this AI-powered automation?

Camilo Fadul: Right. So, DeltaV has been used in—I mentioned all those industries before—but, for example, in the process of refining petroleum, in the process of manufacturing lifesaving therapies. I think that it’s a really good tidbit of information that not everyone is aware of, but three out of the four Covid vaccines that were manufactured for the Western world were manufactured using DeltaV software and equipment. Because software is a solution that is not just about the software itself, but also the hardware platform where that software is running in.

So the impact that our company, Emerson, and in particular the DeltaV solution is making in the world is fantastic. And I feel proud to come to work every day knowing that we are changing the lives of millions of our customers’ customers. Some of those are ourselves—employees of Emerson, right? So that is a great example that I think drives home the point of the great benefit that DeltaV is bringing to our customers and our customers’ customers.

Christina Cardoza: Absolutely. It’s amazing to hear these types of customers or industries that you’re working in. It’s not only manufacturers delivering a product, but these are highly sensitive products, these are highly sensitive industries and environments sometimes. So it’s great to see how AI and this solution is—can really make a difference in everyday lives and in the world.

And so I mentioned a little bit about Intel Geti just earlier, James. I’d like to learn a little bit more. Obviously we have you guys all together in a room, and you guys said in the introduction that you work together to make these solutions possible and make—bring it to manufacturers and make them a reality. So, I’m curious exactly what that partnership is like: how Intel is working with Emerson, and are you using Intel Geti to deploy some of these?

James Green: Right now the partnership with Emerson has been two-plus years. And I mean, when you think about AI, everybody kind of considers AI and the start of ChatGPT. It’s really before that. So, we’ve been partnering with Emerson and providing—they’ve been an awesome partner in terms of providing us real-time feedback of what their customers are experiencing and what tools and applications that might enhance their own, and to be able to talk to their customers to help with their processes and procedures and that kind of thing. So from that perspective it’s been a long partnership, and we’re really excited about the partnership in the future.

Camilo Fadul: Yeah, to me it has been a lot of fun because I feel like I am a kid in Christmas that is enjoying a brand-new gift and brand-new toy. And everything is about; Oh, we can do this; oh no, but wait—but we can also do that. And let’s put a proof of concept together. And to this point, we have already a couple of demos that we have in our customer-visit center here. We call it the Integrated Operations Center, where we have been able to leverage Geti, and tightly integrate it into DeltaV so that we can resolve challenging applications that we typically have out in the field.

For example, some of our customers have products that are stored in vessels, but when they are filling those products produce a lot of foam in the surface of those vessels. That foam typically throws off the level measurement. It’s not—the liquid as a temporary condition is foaming up, and it’s challenging to detect it. So we have an exercise where we were using a standard camera that was already in the room, using it for conferencing. It’s like having the webcam that we’re using right now—actually it’s more sophisticated than a webcam, sorry—but using a webcam, using a conferencing webcam, to now turn it into a sensor that can tell us: Wait, there is foam in that vessel; you have to do something. We’re going to trigger an alarm in the control system that is going to call a human to take an action and see what’s happening with that foam. And like that we have other examples, right, Curtis?

 Curtis Thompson: Yeah. I mean, you talked about asking just: Well wait, can I do this? Or what about that? And so we’ve actually demoed this to quite a few customers, and they’ll ask, they’ll start asking: Well wait, can we do this? Or, can we use it to deduct this certain condition, or can we use a thermal camera to do this? And I kid you not, every single time an idea has come up, the answer’s been: Well, actually yes, you could do that. Right?

And so the number of use cases—I mean I think, and Camilo and I have talked a lot about this too, is the future of the smart plant. And cameras being already easily accessible—widely available cameras as sensors as a main input into your automation and control strategy is—I mean the—there’s huge amount of potential. And I think the most immediate and obvious one is safety quality, but even optimizing your process as well. So, yeah, it’s really fun to talk to customers and show them some of the stuff that we’ve got going with Geti.

James Green: It’s interesting to get a customer on a call with Emerson and say: Oh, can you do this? Having this discussion and brainstorm about ideas of using that camera as a sensor is really impactful. Sometimes there’s things that—Oh, that’s a great idea. Let’s see if we can figure out how to do that. So, yeah, we’re really excited about the partnership with Emerson and look forward to a future with them.

Christina Cardoza: Yeah, it’s great to see experts working together. Like you guys said, you’re taking feedback from end users and from customers, and you’re really trying to customize solutions or improve solutions that are really going to make an impactful difference. So it’s great to see that type of partnership working.

One thing that you mentioned, we talked about the benefits and the challenges of getting this together, and you mentioned how manufacturers, they have a camera that they already have in the floor that they can use to add some of this. So I’m just curious, what is—how difficult is it for manufacturers to bring automation and then bring AI automation onto the factory? How can they leverage some of that existing infrastructure equipment that they do have without it being a huge investment to start getting some of these benefits?

James Green: Well, I guess I’ll start with that. That’s essentially what our platform Geti does, is it takes existing cameras, things that are already there, and is able to work with the subject-matter experts because of the platform’s ease of use to be able to transfer their knowledge of what it is that they’re looking at—whether it be on a product or in the life-sciences field, maybe we’re helping researchers detect anomalies in the life-sciences side of things.

So in the use cases they’re analysts. But it’s that knowledge transfer from the subject matter expert and placing it into the computer vision model. Then, and then AI is then used to help with that automation process. That’s the huge thing that Geti is looking to accomplish, which is in a lot of ways it’s like the WordPress for building websites. It’s allowing an end user that really doesn’t have a background in this process to be able to train a model. If I can train a model, anybody can.

Christina Cardoza: Great. And since we mentioned Industry 5.0 in the beginning, it seems like we’ve just started talking about Industry 4.0, and it seems like there’s still a lot of work there, and there’s still a lot of manufacturers that are still trying to reach this idea of Industry 4.0. But of course we’re always looking towards the future, and Industry 5.0 conversations are coming out.

So, James, I’m just curious if there’s any emerging trends or technologies that you see coming that are going to push us to Industry 5.0 or really revolutionize the way that we are using AI in these industrial settings.

James Green: Yeah. Intel has been working on our software platforms, and they range from Geti—which is computer vision focused—to other applications that we’re developing in the digital twin space—using those cameras as sensors, creating 3D digital twins of manufacturing for spaces, figuring out where people are interacting in those spaces, and if people are in the right place or a place where they shouldn’t be in—those kinds of things we’re also innovating on. And also in the ML Ops space of making this, all this data, actionable in a way that we’re working on. And you’ll see more investment, more announcements that come from Intel in that space in the relative future.

Christina Cardoza: Awesome. Well, I’m excited to see what else comes out from Intel and also Emerson in this space. Unfortunately, we are running a little bit out of time, but before we go I just want to give a chance to throw it back to each of you for any final thoughts or key takeaways that you want to leave our listeners with today. So, Curtis, we’ll start with you.

Curtis Thompson: Yeah, sure. I mean, I just want to say I’m really privileged to be working in this space. I think it’s a revolutionary time—I mean, across many industries, but certainly for manufacturers and automation companies. We’re talking about technology that has the power to learn from data, that can react quicker than humans can. And you think about, well, the art of the possible and where we’re going: we talked about Industry 5.0. And so, yeah, I think AI is going to be—I mean, it’s on the roadmap to Industry 5.0, and certainly this collaboration with technology providers like Intel is what’s going to enable that. So we really appreciate this partnership.

Camilo Fadul: Absolutely. I think that if I could add a little bit also to it: when we started talking about Industry 4.0 and breaking the silos where data used to be stuck at to make it available to more parts of the organization and eventually becoming the foundation for applying AI models—we in industry, in the process-control industry, refer to that as the data-democratization effect.

Now, with the generalization of tools like ChatGPT making access to artificial intelligence technology and getting people to familiarize themselves with artificial intelligence, I believe is being a similar process where AI has also been democratized. So as we continue to see the technologies evolving, the possibilities of course are endless. The challenges will come alongside with them, but we are going to be there making sure that when we provide our customers with these collaborations with important partners like Intel, they are going to continue to be impactful, not just to them but also to their customers. So it is a very exciting time. And I, like Curtis said, I think it’s really cool to be part of it and making a real impact in how we can drive industry.

James Green: Okay guys, there’s not much to add there. I mean, it’s a really exciting time to be in this space. It’s easy to have conversations with customers in this space around AI and what business impacts that the AI can have on their businesses. And it’s usually focused around reducing risk, or reducing costs, which is really transformative to the industry. And Intel is really excited about our partnership with Emerson, and we’re looking forward to working together.

Christina Cardoza: Awesome. Well, thank you guys so much for the insightful and great conversation. I invite all of our listeners to visit the Emerson website—see how DeltaV can help their manufacturing operations. And keep up with Intel, because they’re always coming out with the latest and greatest technologies that are going to really help push some of these ideas forward. So thank you guys again. And, until next time, this has been the IoT Chat.

The preceding transcript is provided to ensure accessibility and is intended to accurately capture an informal conversation. The transcript may contain improper uses of trademarked terms and as such should not be used for any other purposes. For more information, please see the Intel^® trademark information.

This transcript was edited by Erin Noble, copy editor.