HARDWARE PLATFORMS

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 IoT advancements in manufacturing with Ricky Watts from Intel, and Teo Pham, an expert in digital trends. But before we jump into it, let’s get to know our guests a bit more. Ricky, I’ll start with you. Tell us more about yourself and what you do at Intel.

Ricky Watts: Thanks, Christina, and welcome everybody. Yeah, Ricky Watts. I work for Intel. I am in the Federal and Industrial business unit. My role, to put it simply, is looking at Intel technologies and architectures and how do they apply into the industrial environments—the industrial market segments with the energy or manufacturing process and industry. What are we doing with our silicon platforms, our software-enablement platforms, to bring new and exciting innovations that I think we’re all starting to see in the industrial environment we’re going to talk about today? How do we make them real for our customers? And obviously for those people that are producing goods in the factories today, that’s my role.

Christina Cardoza: Great. And Teo, I introduced you as an expert in digital trends, but for those of our listeners who haven’t been following you, tell us more about yourself and what that means.

Teo Pham: So, I used to be a startup founder and a professor of digital marketing. And in the past couple of years I’ve been hosting a German podcast about business and technology. And so I’m always very curious about the latest digital trends. And so in the past two years I was very curious about things like blockchain, about Web3, and also NFTs. But I think in the last six to twelve months I’ve been mainly giving talks and trainings about artificial intelligence. And I think this is also going to be one of the topics of today’s podcast.

Christina Cardoza: Yeah, absolutely. Like I said, we’re going to be talking about IoT advancements and overall technology advancements happening in the manufacturing space. And I think this is a good time to have both of you joining us today, because Ricky and Teo you guys were both just at the Hanover Messe Conference in April, and there were a lot of new things going on. Ricky, I saw you do a lot of interviews with a lot of different Intel partners out there, so I wanted to hear a little bit more about where you see this manufacturing space headed, based on what you saw at the event and just other things that you’re seeing in the industry.

Ricky Watts: Yeah, good question Christina. And, yeah, I think when I went to Hannover Messe this year, I was six foot two and I came back as six foot. So, I certainly covered a lot of area at the event; it was very exciting. We’re all, I suppose, forgetting it’s post-Covid, so manufacturing was back and forth at the event. I think there were up to about 3,000 exhibits at the event.

What did I see that was exciting? So many things, starting with just the excitement of the people, the interaction between the various people that were there to visit the event and learn from each other about what they’re doing. In terms of technology, I think two areas that kind of absolutely really excite me, and I suppose to some extent concerned me a little bit, I will say, was the rise of AI, particularly around ChatGPT, and how AI and ChatGPT are being integrated into traditional manufacturing to some extent—what is the impact of that and how that’s going to be implemented in the manufacturing execution systems that we have out there today in new technologies that are coming out.

And I saw various booths with some of the larger companies in this space, how they’re using that technology already to integrate it and drive outcomes from the data and drive different things that they want to do in manufacturing to be more efficient, etc. So I thought that was particularly exciting and, to be honest with you, I was really surprised to see how much of it was out there and how advanced it was. I think that was a bit of a surprise to me, I’m going to be very honest with you.

I think—the other thing I think that interests me as well is a couple of things. This thing called 3D reality, omniverse, all this metaverse and things like that—I did manage to see a few examples of how immersive technology is going to be used in the future. Manufacturing industrial is a very complex environment, very complicated in terms of how you operate it around OT principles and IT principles and all the things that I’m sure me and Teo look at regularly. But using that type of immersive technology to really make it much more accessible, and how do I design and build a factory to change the outcomes—can I do that in a 3D virtual reality, which is very visual, and then use ChatGPT and AI to work in that environment to create digital twins that can create obviously what we call physical realities in the manufacturing as well? So that was pretty exciting.

And the last one I’ll mention before I hand it over Teo is, which made me smile, was there were a lot of robotics at the show. As always, robotics is everywhere in manufacturing, for good reasons of course—the logistics and the repetitive tasks that we often see in manufacturing. But one thing particularly interesting to me was robots—building robots to drive outcomes. I thought that was really interesting that a robot is given a task, and another robot builds the robot to drive the outcome of that task. So, I don’t know—let’s say I’m working on a PCB, or the manufacturing of an electronic board: I design the robot, the robot builds the robot to do that task. What was particularly interesting then was using AI for the robot to learn what it needed to do to send off a command to build or design a new tool for the robot. So the robot builds the robot, which builds the robot, which is linked by AI.

So I thought that was fascinating to watch—effectively robots not only being used to do and drive a task, but even the way that they were doing that task using artificial intelligence to design a better robot. I’ll use the word “real time” —as the process is ongoing, it’s optimizing the robot. So there are three things that I thought were exciting. In addition, there were so many other things, but I’ll let Teo jump in. He’s probably got some of his own insights.

Teo Pham: So, it was my first time at Hannover Messe, and to be honest I was very surprised and also amazed at the variety of topics and also participants at this exhibition. Because when you go to a manufacturing trade show, you expect obviously robots, you expect hardware manufacturers, you expect semiconductor manufacturers. But then you also had all these software companies, you had consultancies like PCG, you had cloud-service providers like Amazon Web Services. And I think it just goes to show how varied this whole space is, and that it’s a lot more than just physical devices, but it’s fully integrated with software, with artificial intelligence, with the cloud. And I think this is also how you can really create these new exciting applications.

And, as Ricky said, there was lots of talk about artificial intelligence—we’ll get to that in a second—but also what you mentioned about the metaverse. So, I saw companies, let’s say like Siemens or Microsoft, that were promoting things like the industrial metaverse. So, creating new technologies that make production more immersive, but also a lot cheaper in the sense that you can create these digital twins that allow you to run these amazing simulations so you can really test out things in the digital space before even needing to create them as a physical unit. And so I thought that was pretty amazing. So, Ricky, you already mentioned that there was so much talk about AI at Hannover Messe, but which AI applications in particular got you really excited when it comes to manufacturing?

Ricky Watts: Yeah, Teo, I think I’ll talk about one in particular in the use of AI ChatGPT. In the world of manufacturing we have these things called manufacturing execution systems—MES, or programmaticlogic controllers. Basically it’s a device or an appliance that basically runs the manufacturing. So if I’m building something there’s a bunch of machines that work together to create an outcome. They’re the builders, if you like, the building blocks.

Now these PLCs, they have a language that operates and runs with them, it’s called 61131. It’s a PLC code—a way that you actually design and run that PLC. And I think one demo that I saw was ChatGPT being used to build that code. That is typically a manufacturing engineer or somebody in that advanced systems integrator that’s writing that object code to create that outcome which they then apply into the manufacturing. It’s the bit that controls the machines, as I mentioned—this is now ChatGPT building that code. So something that might typically take an engineer to build that type of code and build it out could take weeks, months to do, ChatGPT was doing it in, I’m going to say minutes, seconds.

Now, I was excited to see that as an application for me, because that’s something in—when we look at manufacturing, one of the big issues that’s been as we’re moving towards this digital transformation world is the skills that are involved in the digital transformation of manufacturing are different to the ones that have been there in what we would call the mechanical, if you look at the previous industrial revolution. So you’ve got lots of very experienced engineers that are starting to come out of the workforce; you need to replace them with new people coming in who broadly come from a different type of perspective. They want to be much more involved in technology.

So when you see ChatGPT being used in this environment there’s a couple of benefits. One is obviously the speed and the rapidity of being able to do that, but it also helps them address some of the code issues. So I was excited to see that. I’m going to stress that it’s early, what it’s doing. But the potential of that technology in AI to be implemented—and link back to what you said, Teo, about omniverse and things like that—the reality is manufacturing is very much a structured environment driven around a set of standards, as we know. So, but as we’re starting to go into this kind of new world, the ability to be able to do that with AI and link that to some of the integrated engineers I thought was really exciting.

As I say, very early use cases: they were at pains to point out that there were some mistakes in the code, but I thought that given where AI is going and the use of that code, it will not be very long before the accuracy and the ability to deploy that directly to those machines is really going to become relevant. So that was one AI use case that I have to say completely, I was like, “Whoa, wow, this is fantastic.” In manufacturing, Teo, this will have a huge impact in the years to come. And I think when we see next year’s trade show, I think we’re going to even see more advancements in the use of AI in these models.

Christina Cardoza: Yeah, I love hearing all of these examples, especially with the ChatGPT use case, because, like you said, this is just something that’s coming out now, and I think a lot of the use cases there have been more writing articles or more content driven, but to see it actually writing code and that being applied to solutions on the manufacturing floor, that is something really interesting and exciting to look forward to over the next couple of years.

But, like you said, a lot of this stuff is still early days; these are just early examples or applications. So, Teo, I’m wondering from, sort of as an outsider looking in, where do you see that the manufacturing opportunities really are now, from what you’ve saw at the event? There was obviously so much going on, it could be very overwhelming. Where do you think manufacturers should be focusing their efforts now, or how well do you think they’re equipped to be taking advantage of some of these new opportunities available to them?

Teo Pham: So, again, coming back to artificial intelligence, I think it just allows you to speed up all of the processes to make them a lot faster, a lot cheaper. Obviously there’s always a lot of talk about ChatGPT—so this is text-based AI—but you also have AI tools that can help you generate images or blueprints, generate videos, even computer code that Ricky mentioned, complete websites or applications. And so I think this is super exciting. So I think the cost of generating an, let’s say, an 80% solution will go down to practically zero. But then obviously you will still need some very experienced people to go from 80% to 100%.

But I think oftentimes it just takes so much time and effort to go from 0 to 80. And so I think having artificial intelligence will speed up a lot of that. And I think there will be some very fancy applications—let’s say like AI 3D modeling and stuff like that. But I think even for fairly boring stuff, like documentation or translation, I think this will be so helpful because you can get all of that stuff within minutes.

And I also like the idea of being able to talk to the machine, to have some kind of dialogue. So, you don’t need to necessarily know everything from the start, but let’s say you want to understand something, or you want to understand some issue that’s going on, you can just start off with a fairly general question and then just dig deeper and have a real conversation, as if you were talking to a technical expert. And I think this is super exciting because it just allows you to go very deep into any type of subject matter, even if initially you’re not necessarily a top expert, but it just allows you to go deeper and deeper and really accelerate your learning.

Christina Cardoza: Yeah, absolutely. And sometimes those boring solutions are the most important solutions or aspects to manufacturing. But it sounds like we’re talking about all of these AIs doing all of these things too. It sounds like manufacturers are going to need to have a lot of hardware or a lot of power in their tool set to be able to make some of this happen. Would you agree?

Teo Pham: Definitely. And this is also something I wanted to ask Ricky. So, obviously there’s lots and lots of data, you need lots of processing power, and I think it would be really useful for our listeners to understand, okay, when do we need CPUs? When do we need GPUs? What’s the difference? What’s the difference between AI model training and inference? And also what solutions does Intel offer in those areas?

Ricky Watts: CPUs and GPUs both have roles and advances in the use of AI, but if we think of manufacturing and we think of AI, AI really relies on data, the abstraction of data, what they call data engineering effectively to get that out. And then effectively what you’re going to do is the learning part, and then you’re going to do the inference. So, and I think what you’ve got is different types of compute platforms and environments. So CPU, GPU or FPGA are always involved in those environments.

Manufacturing’s very interesting. A lot of the early use cases in AI have been really around visual use cases, video use cases. I put a camera into a manufacturing environment to analyze something, and then what I want to do is I want to train a model around those images that are coming up. Let’s say I’ve got a production line and I’ve got something that’s coming out of that production line. I don’t know—let’s say it’s a bottle with a label on it, and I’ve got a camera over that bottle and I want to know, “Hey, is the label on correctly? Is the label on the right way round? Where is it structured in there?” So we create images around that, and then what we would do is we would train models. That training is generally done in a GPU type of environment, because it requires a lot of intensive processing to do that parallel processing around those images.

And then what we do is this thing called inference, which is now I have something that knows what’s good, knows what’s bad, effectively. Okay, then I want to apply that in a manufacturing environment. I can’t keep learning all the time; it’s too difficult. So what I want to do is I want to do this thing called inference. I want to then take images as much as I can as I’m seeing that production line generate those goods and they’re coming out, I’m using the model and I want to apply that. And that’s really where you start to see things like CPUs come into account, because it’s very much tactical at the end. It’s about applying something very, very close to where the manufacturer’s coming in.

So you’ve got CPUs and GPUs, which both have an area where they have some expertise. But what we’re starting to see from an Intel perspective is starting to integrate some of these things. You’ve seen it with some of our new technologies, particularly around the latest Xeon^® chip that came out recently, the Sapphire Rapids chip, where we’re starting to integrate packages and capabilities into the silicon. Now of course that chip is being used in the data center. A lot of the training, etc. is done where you’ve got a lot of compute power, which has typically been in cloud environments. The inference in most cases is done where the manufacturing is; you want it done at deployed, so, at the edge.

So we’re now starting to see these compute platforms in these environments kind of go from edge to cloud. So you’ve got CPUs, you’ve got GPUs, you’ve got FPGAs involved in that. The last thing I would say is, from an Intel perspective, is if I start to think of these environments that I’ve got, there’s two sets of data: the video I mentioned, but the one that’s much more pervasive in manufacturing is what we call time series data. The world of manufacturing has what we would call fixed-function appliances: machines, a robot or a conveyor belt or a system that’s doing a press or something like that. That is generating data. It’s not vision data; it’s data that is coming off the machine. It could be heat, it could be pressure, it could be vibration, it could be performance—all of these things.

That type of data, from an Intel perspective and from a manufacturing perspective is much better and optimized to run on CPUs at the edge as well. So you can do the training and the inference at the CPU, at the edge, and where data integrity and data sovereignty is becoming very important. And I know it’s not an area necessarily that we’re talking about today, but a lot of our industrial manufacturers, they have a high consideration on the value of the data that they’re generating and where it gets analyzed and what they’re doing with it as well. So they want to bring that inference and that training right to the edge as well.

So, our integrated offerings of our CPUs and our GPUs.  We’ve got a new portfolio of GPUs coming out—I’m sure you’ve read a lot about that too. It’s early days for Intel in that space. But we are learning fast, and we’ve got more products coming out over the next few years. On our CPU side I mentioned Sapphire Rapids. We’ve got some new products coming out; they’re going to integrate even more AI capability down there at the edge. So I think, for us, it’s integrating the hardware solutions, and then on top of that providing a uniform architecture for people and developers in the AI space to access those technologies.

And you may have heard of this thing called oneAPI. Basically it’s an ability for developers using whatever code—Caffe, Python, all of those things that they are doing—how do they get access into that? How do they work within that data set—OpenVINO Toolkit? So, we’ve built a number of toolkits and optimizations on those integrated packages. So whether you are a developer who’s working in the environment from an OEM or systems integrator or indeed even with some of the large manufacturers who may have data scientists, we want to make our silicon accessible and easy for those developers.

So, irrespective of being CPU, GPU, or FPGA, we optimize underneath; you tell us how you want to run what the workloads are, and then we’ll deploy those workloads into the right silicon platform at the edge and then provide a uniform capability to take that to the cloud as well. So that’s kind of what we’re up to. Sorry—long-winded answer, Christina and Teo. Because it’s a deep question.

Christina Cardoza: No, absolutely. And there’s a lot going on. I think that ecosystem was certainly on display at Hannover Messe. You gave that, “If you had a bottle and a label on it,” example, but I saw that exact demonstration from Dell Technologies at the event. Those robots that we were talking about, collaborative robots working with robots, VMware was showcasing that too. And NexCOBOT was actually showcasing how you’re controlling robots and using AI and all of these things, all powered by Intel processors and Intel technology.

And, to one of the earlier points you made, this was one of the first conferences really out of Covid. And so talking about all of these edge use cases, this is one of the first times the industry was able to get together and start talking about, not only talking about edge, but showcasing how this is really possible in an industrial environment.

So I’m just curious, going further with this edge conversation, what are the benefits that you’re seeing the manufacturers really gain from now moving to the edge? And you sort of talked about this, but expand a little bit on how they play on some of the trends that you’re seeing where the space is going.

Ricky Watts: Again, Christina, it’s a good question. Look, manufacturing is a very competitive business. There are manufacturers all over the world that are creating goods, and there’s a lot of competition out there. So the use of data and the use of these things in these environments can often give a very competitive advantage to an end manufacturer.

And I meet manufacturers of all different sizes. We are very aware of some very large manufacturers. Typically, in the markets that I cover, the auto manufacturers are discrete—in which “discrete” is about making things, as I call it, so, physical items, very much so. Whereas process manufacturer is about mixing things, chemical manufacturers such as BASF, etc. So, between those two environments what they want to do is they’re in a competitive environment. So if they can implement technology to give them a competitive advantage, to change an outcome, to improve a process, that in itself will give them an advantage that they can then pass on to their customer, or they can improve their margins, or they can do many different things in that.

So when I look at some of the things that manufacturers are doing with this technology today, it’s really can they apply that technology in a business-driven outcome? It’s very easy in our environment, particularly with what I do, to forget at the end of the day it’s not about the technology, it’s about what the outcome is. What is the benefit to a manufacturer? And I’ll give you some statistics that I’ve seen recently, which are really interesting, which is over 80% of use cases that are using data fail, because it’s very hard to deploy these things and deploy with an outcome that a manufacturer says, “Well I’m doing this thing, but I’ve not got the benefit that I expected.”

And so what we need to do in the technology industry is make it easier for those people to consume. You mentioned VMware and Dell, some of the things that we showed actively at the event—VMware, the robotics. So it goes back to what I said earlier: manufacturers want to use this technology. They’re not experts in AI, they don’t have data scientists, they don’t have people like me and Teo to kind of help them out. They’re really—they want something that’s simple, that they can use.

So what we are trying to do within that ecosystem is give them something that really is—to coin a phrase as much as possible—the easy button. And, again, that’s difficult in a sense, and still there’s the complexity, but our partnership is about bringing that technology to them, making sure that we’ve built the relationship with Dell and VMware. If you have these architectures you can deploy this. It’s relatively simple to go in. And then making sure that those optimizations for the use of AI—whether it’s video data or time series data—we’ve made that much more seamless, so that when somebody implements that technology they can access the benefits of that much quicker.

And that’s really where we’ve been spending a lot of our time as well, with that ecosystem approach, which is, I think everybody knows the advantages of AI. I don’t think you need to convince anybody. What we really need to get to is how can somebody implement it and truly get a benefit from what you’re trying to do? Do they improve the outcome?

You go back to my bottle example, Christina. If I’m putting through a hundred thousand bottles a day, and let’s say as an example 5% of it is inaccurate, okay? So I might be throwing 5,000 bottles away a day. That’s a sustainability issue, that’s a profitability issue. If I can reduce that failure rate to 1%, that has a massive impact on the performance of that bottling factory. He has less goods being returned, he’s improved his sustainability, and of course his profitability comes up. So we’ve got to make that as easy as we can to consume, and we’ve got to make sure that that technology is accessible for everybody in manufacturing, not just large-scale manufacturers that have huge departments of engineers and data scientists. So that’s kind of where we’re at. And I think you are starting to see a lot of that progress coming up.

Teo Pham: When we talk about the implementation of AI, I think one of the decisions we have to make is about whether to do it with edge computing or cloud computing. And I wanted to get your thoughts on that, because obviously there are some advantages to edge computing: it reduces latency. Also in terms of data privacy—we don’t have to send it to a cloud. So these are very obvious advantages. On the other hand, we need to invest more in hardware. This hardware could be costly, it takes up some space, it generates some heat. So what are your thoughts on edge versus cloud, and what is Intel’s approach on this topic?

Ricky Watts: Let’s take it holistically. I think it’s both. There are distinct advantages with both scenarios. The cloud has what I would call cloud-scale compute, elastic compute, but getting the data into that cloud is very expensive. These systems that are generating data in manufacturing can be extensive. The volumes of data are massive. The cost of transporting that data is massive to get it into the cloud, let alone then, of course, some considerations around regulation, data sovereignty and privacy, security, etc.

So as a manufacturer looks at what he’s trying to do, he’s considering a lot of things, he’s looking at what is the use case that I’m doing? What is the outcome that I’m trying to generate? What are some of the considerations that I need to do, and what’s the benefit? I think it’s use-case driven, to your question. There’s a lot of advantages for doing training in the cloud, doing inference at the edge. And then as more and more powerful compute comes down to the edge and we define that, not only the training and but the learning can be done at the edge as well, and you’re going to see more and more compute move to the edge, and then that in itself will then connect back through digital twins—things that we talked about earlier—into what we call the manufacturing execution system.

So I see a problem or I want to do something, I need to act on that, and I need to act on that in a very low-latency environment. So in my mind more processing goes to the edge. But I do believe that there is absolutely going to be a very integrated offering between the edge and the cloud for abstractions and things that you might want to do when you get this massive cloud-scale compute.

Christina Cardoza: Yeah, absolutely. And this has been a great conversation, guys. We’ve talked about AI, robotics, edge, and still so much more we could talk about, so many more different areas that we can get into. But unfortunately we are running out of time. So, before we go, I just want to quickly throw it back to you both one last time. Any final thoughts, any key takeaways you want to leave our attendees with today? Where the industry is headed, what needs to happen to get there, and what’s still possible for the future? So, Teo, I’m going to throw it to you first.

Teo Pham: So, some people are saying that currently we are witnessing the iPhone moment of artificial intelligence. What does that mean? Even before the iPhone came out in 2007, obviously we had regular phones, but we also had mobile phones, right? But still the iPhone changed everything. And today we can’t even imagine a world without the iPhone, without smartphones, without mobile apps.

And, similarly, artificial intelligence has been around for 50 or even 60 years. There had been research in AI in the 1950s, but I think currently we are in this kind of virtuous cycle, where a lot of things are just creating this kind of perfect storm. We have lots and lots of data, we have the necessary compute, we have the models, and we have very easy-to-use interfaces like ChatGPT.

And I think each player in those different areas, they’re making so much progress. OpenAI is making so much progress with their models and their user interfaces. Intel is making so much progress with their compute, and I think each of them is contributing so much to this virtuous cycle that maybe in even in six to twelve months the whole space might be unrecognizable because there’s just so much progress. I mean, just six months ago no one could even spell out “ChatGPT,” no one knew about it. And today hundreds of millions of people are using it. And so I imagine that even in the fast-moving space of technology, we’re in for a pretty fun ride over the next few months.

Ricky Watts: Well said.

Christina Cardoza: Yeah, absolutely. And I love that example, even beyond the iPhone. You think about when phones first arose, we’d never think that we would be using phones the way we are today, that there are these tiny computers in our pockets taking pictures, things like that. And I think in a couple of years from now, things that we see today, we are saying, “Oh, AI—that’s not really going to take off,” or “This is not really possible.” And in the next couple of years we’ll be laughing at ourselves, saying, “I can’t believe that we were doing it this way.” So, really love the way that you phrased that in that point. Ricky, before we go, any final key thoughts or takeaways from you?

Ricky Watts: Manufacturing has a duty to produce its goods, so technology is moving on. I love the way that Teo said it. So I’m going to say technology’s coming. It is changing. Teo is right: in twelve months’ time we could be talking about something we’ve not even heard of today.  And I think that that is something, but ultimately manufacturing has to continue to produce goods. So what I see is manufacturers, they are focused on the new technologies, but they’re also making sure that the manufacturing environments that they’ve got today are going to be there for the next few years. There’s a lot of operational standards and things that we need to do.

So, here at Intel, together with those, we’ll work with our partners, our ecosystem partners, and our customers and the industry as we go through that transformation. But what we want to do is make sure that we keep making sure that we keep the lights on, if it’s energy, and the manufacturers generating the goods that we need to do. So we want to make sure that that transition is smooth and integrated, and is as little disruptional as possible as we go for this industrial transformation.

Christina Cardoza: Absolutely, and to your point, none of these advancements would be possible or will be possible in the future if we don’t have the technology out there that makes it simple for us and easy and smooth, like you mentioned; if it’s complicated, no one’s going to really want to do it or be able to do it.

So, it’s great to hear about all of the technology coming out from Intel, and I know we have a lot more to look forward to. So with that, I just want to thank you both again for the insightful and informative conversation, and thanks to our listeners for tuning in today. 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.