Superpowers for industry: How Siemens and NVIDIA are transforming AI

Speaker A: Foreign. This is Talking Digital Industries, the podcast for technologies and trends that drive sustainable digital enterprises. I'm your host, Alex Chavez. This episode is all about industrial AI solutions, available now thanks to a collaboration between Siemens and and Nvidia. Plus we'll also explore how AI is set to shape the future of industry. To break it all down, I'm joined by two experts. Alvin Clark, Global Developer Relations Manager for Industrial manufacturing at uh, Nvidia, and Dr. Michael Schrapp, head of industrial AI at Siemens. Welcome to both of you.

Speaker B: Thanks Alex.

Speaker C: Thanks for having us.

Speaker A: It's great to be together in one room. Oftentimes we do our podcasts online, so I'm really excited to have you both here today. Nvidia is almost synonymous with gaming, but a partnership with Siemens, that might surprise some people. Alvin, uh, can you walk us through how this collaboration began and what brought the technologies from the two companies together?

Speaker B: Yeah, happy to. Nvidia is well known for gaming, but our technology has always been about accelerated computing, whether that's accelerating graphics, AI or simulation. And that's really where our partnership with Siemens started or it comes from. Siemens is a leader in industrial automation and digitalization and they've been at the forefront of transforming industries like manufacturing, energy and infrastructure. As the industrial processes became more data driven and AI powered, the need for that accelerated computing that Nvidia provides for digital twins and simulation has grown significantly. So the collaboration started with a shared vision to bring AI and digital twins together to optimize industrial operations. Siemens accelerator platform and Nvidia's OMniverse, along with AI and edge computing solutions, enable manufacturers to make industrial processes more efficient, sustainable and intelligent. So yeah, while Nvidia might be famous for gaming, our uh, core technologies that AI simulation high performance computing are solving some of the biggest challenges in industrial and manufacturing sectors. And that's really where that partnership with Siemens is such a natural fit.

Speaker A: Michael, when you look at these technologies that Nvidia provides, was it always clear to you from the beginning there's something we can do with this? This will benefit Siemens and its customers.

Speaker C: Yeah, absolutely. Nvidia's advancements in digital, uh, twins, industrial metaverse applications, uh, GPU advancements and of course as we have heard already from Alman in AI align perfectly uh, with our goals at Siemens. I mean in Siemens digital industries we provide hardware and software solutions and products along the entire value chain from product design towards automation of factories and, and for our industrial customers, ranging from process to hybrid to discrete industry. We need state of the art technologies and we need to Be fast on the market. And we see Nvidia here as a very powerful partner to do that. But it's not only a partnership where we integrate technologies, but rather set even new standards for industrial operations together.

Speaker A: A key word that I just heard from you was AI. So let's drill a little bit deeper into that area. So how is AI being employed for industrial applications? And maybe you can also talk about some of the challenges that go along with that.

Speaker C: Uh, sure, absolutely. More than happy to. So there are several drivers, key drivers why manufacturers are increasingly adopting AI. One is operational efficiency. So AI helps to improve operational efficiency throughput, uh, especially as traditional automation reaches its limits. And another aspect is, uh, risk mitigation. So AI addresses risks in, for example supply chain disruption, cybersecurity, energy efficiency, sustainability and cost management, uh, uh, to enhance the operational resilience. So AI can perform various tasks from the shop floor, such as enabling robots to handle differently shaped objects, analyzing quality imaging data through inspection technologies, um, and allowing workers to control, for example machines with voice commands. But uh, to come to your second question about the challenges, uh, there are challenges, especially when it's about how to fulfill the strong standards about reliability. Uh, the requirements make it very difficult for manufacturers to turn AI's potential into business value. And in manufacturing you do not have the right skillset of people on the shop floor, uh, to run AI models or retrain models or take care of things if they do not work. And there are several statistics which highlights exactly these challenges in manufacturing. For example, there's uh, a study from Gartner, a research firm from last year where they state nearly 40% of people think AI is not trustworthy. And trustworthiness is a very important topic in manufacturing. Uh, there's another study from BCG, also from last year, uh, which states that only 16% achieved their AI related goals. 16%. So there's still a long way to go. And to address these issues, these challenges, our focus is on developing industrial grade AI as we call it, which is robust, trustworthy and reliable to meet the demands of industrial environments.

Speaker A: And I imagine you also have to work on changing people's minds.

Speaker C: Absolutely, absolutely. Because it's a big topic, especially on the shop floor, for blue collar workers to make sure that they are really willing to use the AI and understand that it helps them rather than replaces them.

Speaker B: From a manufacturing standpoint, from a world perspective, right, we're making more things than ever before. And the things that we're m making are more complex than ever before. So whether it's electric vehicles, Next generation semiconductors, um, smart medical devices. Right. The manufacturing demand hands just continue to drive precision, efficiency, agility just at a completely unprecedented scale. And at the same time there simply aren't enough skilled workers to meet demand. Um, also there's a skills gap because of the complexity of the things that we're building. So AI, I think it's sort of stepping up to help not by automating or replacing workers, but really by augmenting that human expertise that the workers that are there with their uh, domain expertise. Right. Can come in and hopefully, you know, through the partnership and we're, what we're bringing in AI to the factory floor. Be able to be given sort of AI superpowers to do more.

Speaker C: Right.

Speaker B: They don't need to become AI experts, but be able to be given the tools to really leverage AI, whether it be an intelligent assistant or agent, uh, or something to do with physical AI.

Speaker A: I like that. AI superpowers. I'd like to get specific now. What are some of these AI superpowers that the collaboration between Siemens and Nvidia is able to give people on the shop floor?

Speaker B: Yeah, so one of the things that we're working on together is industrial copilots. Um, so this is something where you are now able to speed up and provide additional intelligence to the operators to make decisions faster and more efficiently on the shop floor. So that's making readily available all the data. One of the things that I love about manufacturing is just the sheer amount of data that's available on the factory floor is there and then now is how do we make sense of that data?

Speaker C: Right.

Speaker B: How do we create intelligence, generate intelligence, ah, out of that data and into something that's providing business value. So one of the things that it's talked about is very quickly being able to maintain a machine because you're able to ask the machine itself, hey, what does this error mean? How do I go about doing this maintenance step? Um, that can also turn around to going from real time error detection, which we can do with AI, you can look vision and then create an error, but then how do you turn that into real time failure analysis? So that one of the things that as I've learned more about manufacturing, it's not really the ability to detect the errors that's important, but how fast you can arrive at that genesis of what created that error, where it stemmed from. So the ability now to be able to, through this AI assistance, be able to say, hey, we detected this error, we went out and we talked to the telemetry and the Sensors and the settings and the PLCs of the machine. We see something off in this particular setting for maybe an extruder, maybe that's where the error is coming from. And then so that operator now has additional intelligence to understand how to quickly arrive at addressing that issue.

Speaker C: Yeah, exactly. So I mean this industrial copilot is like a uh, super experienced colleague, right? Um, that supports shop floor workers, service technicians or maintenance engineers around the clock. So you simply ask in case of errors, questions and you get a quick answer back to solve the issues Alvin just eluded on. And I mean it's a solution wolfing both software and hardware. And uh, Nvidia works to make all use cases possible. And we as Siemens we make sure that we bring in our industry domain knowledge which is very, very important to tailor your application to the needs of the people on the ground. And overall we want to make the journey from deployment to usage of the industry copilots as easy as possible for our customers. And our vision is to reach a uh, productivity increase of more than 50% for our customers. And in order to reach that we really work hard on creating a powerful offering with the industry copilot that delivers real value for the people on the shop floor that it really helps them. And crucial for that is the powerful collaboration on our sites with Nvidia.

Speaker A: Can you give me a really concrete example of what's possible with one of these copilots?

Speaker C: One of the things we are saying is very important is that we can deploy this copilot where we need it, basically so it can be deployed in the cloud. For more an enterprise level application where you can basically deploy, try to understand issues on different levels, on plant level to fix issues, to improve the operational efficiency, energy efficiency for example but really down also to a specific machine. And we can deploy on our industrial edge, uh framework and also on premise. What does on premise mean? It's basically hardware and software mix which is installed and operated directly locally on the shop floor. And uh, everything is hosted within the company's own IT environment. And, and why do we do that? Uh, what's the main benefit here? The primary advantage of this on premise deployment is the enhanced control over data and IT security. So m, as I said everything stays at the uh, customer's uh premise. And many industries such as, I don't know, healthcare, semiconductor, defense, uh, have strict regulations regarding the data handling and the privacy. And this on premise deployment allows organizations to comply with these uh, regulations by maintaining complete control over their data, uh environment. And second aspect here, operating on premise can also lead to improvements in terms of speed and performance for applications that demand real time processing. So if you get real time data, you need to act fast, you have low latency times and that's what you can enable if the um, application itself really runs close to where the action is happening.

Speaker A: So there's a lot of flexibility with how the co pilots are installed into the manufacturing sites. But suppose I'm somebody on the shop floor. What can I specifically do with this? What would be like a very specific use case that would benefit me in my work and my everyday tasks?

Speaker C: Yes. So let's suppose you are the operator of a machine and you are not really skilled because you just started the job. Right? Um, so you got a first training and let's say you are working in a night shift and then something happens on a machine. Maybe not a very critical error, but some errors are happening. It's not running as it is supposed to. And so far you really didn't know what to do. Right. And now you have this experienced colleague of the industrial copilot and you can ask him, hey, I'm recognizing xyz, there's a certain error, what should I do about it? And then you get recommendations what to do, how can you fix the error? What can you do to bring the system up and running instead of asking some people? And we did in one of our factories, a study together with our human resource department to understand how do the people on the shop floor really use our industrial copilot?

Speaker A: All right, so it almost has a training aspect to it too, doesn't it?

Speaker B: Yeah, yeah. Human in the loop, that operator assistance, that's one of the amazing things that AI can bring. And it really addresses the challenges in adopting AI in manufacturing. Massive data complexity. Right. The amount of data that are there. You see some of these dashboards for some of these plants. There's also a lot of legacy systems. And now we're able to connect these AI assistants to legacy systems there to then be able to provide additional insights. And then of course with the edge computing and where they can be deployed, then they can really be scaled across any type of facility to meet any of the regulations and the rules there, whether it be enterprise and so forth.

Speaker A: What are customers saying? What kind of reactions are you hearing?

Speaker C: There are different reactions depending to who you speak, but they are all positive because it helps the worker itself. You do not need ask other people. Uh, you improve your own efficiency basically as an operator, but you also improve central KPIs you have in manufacturing. So it's not only helping, uh, the operators and getting the positive feedback from them, but also on a supervisor level or even on a plant level that this is definitely going in the right direction. And we piloted that with several customers and they always want to have more and more features. Right. And they're not only using the system as we intended it for a specific use case. They always think ahead, have other ideas where we maybe didn't think about yet. So we have a huge backlog already of additional new topics, what could be done. And I'm sure also from a technology advancements perspective, there will be a lot of things in the next years happening which really helps them even more than today. Uh, I think today it's just a starting point with our industrial cohabitation.

Speaker B: And I think that's what really makes the partnership so special with Siemens because it's this insight, the customers that they speak with, these requirements, these challenges that they're bringing to us and we're collaborating, working on how do we take our technology, how do we manage AI, uh, and actually really kind of provide business value. Which at the end of the day is the critical thing is that we're making sure that we're addressing the right things, those needs, those really cool ideas. I love that a lot of the ideas that we're getting are coming from the end customer. And then we're coming back and we're jointly discussing them and we're saying, okay, how can we bring this to market so that we can actually deliver on those ideas and provide significant business value?

Speaker C: Yes. And I mean what we do a lot is kind of early piloting with customers with the aim to learn quick, also fail fast. So not each and every idea we have or our customers have are really implementable. They're maybe not feasible yet. Um, so we want to fail fast in order to deliver, as Alvin said, real tangible business impact for our customers. Simply according to the motto learn ahead of investment. So we want to learn a lot, we speak a lot with customers and then if we figure out, yeah, that's something which really helps broad range of customers, we then really go, uh, deep into it and focus on that and execute on these ideas.

Speaker A: This is co creation and practice, isn't it?

Speaker C: Absolutely.

Speaker A: Until now we've learned that the partnership between Nvidia and Siemens began with digital twins. Right now we're in the middle of industrial co pilots. And so my question is, what's next?

Speaker B: I think the Siemens Nvidia collaboration is accelerating. It's just getting faster. We're working on so many Things together we have the strong focus on enhancing the uh, industrial AI copilot. Taking all the ideas from these early pilot programs, the successes, the failures and then making them smarter, more adaptive, uh, and deeply integrated into the factory floor. Um, we're expanding the multimodality capabilities, enabling them to as we said, process sensor data, video, operational insights, all simultaneously providing those operators with even more powerful decision making support. Beyond the copilots, we see AI agents coming and being deployed or embedded throughout manufacturing, whether it be for helping to orchestrate workflows, optimizing production, dynamically, adopting in real time. And this is also going to involve advancements in physical AI. Right. How do we take the perception, the reasoning of these AI agents and turn that into real world action, whether it be through robotics or through just decisions that would drive the machines and the productivity. So we're working on many exciting things together really uh, kind of pushing AI deeper and wider into industrial operations.

Speaker C: Absolutely. And I mean as Alwin already said, the developments around AI agents are uh, just the next step and very critical for our continued collaboration. I think it's very important. But it does not stop there. Right there are as you said Alvin, physically I or real world models. And from my perspective they really have the potential to revolutionize how our industry and manufacturing will run in the future. And that's really exciting to work on such topics here with Nvidia.

Speaker A: On the consumer side there are already AI powered agents. One example is in teams. There's an automatic note taking function. So as my final question, uh, I'd like to know from each if you had your own AI powered agent, what sort of tasks would you hand over to it?

Speaker B: I would love a AI agent that loads the dishwasher for me. I am notoriously known for leaving plates and glasses all over the house. It drives my wife crazy. So mixing AI with physical AI, with throwing a robot in there that can follow me around, pick up after me and load and unload the dishwasher would uh, definitely get me some brownie points with my wife.

Speaker C: That sounds good. So real world agent, that's got real

Speaker B: business value for me.

Speaker C: So if you have that, please let me know. I would want to have the segment.

Speaker B: Yeah, sounds good.

Speaker C: And uh, I mean what, what I would like to have is a kind of, and I think we are not that far away, ah, a ah kind of travel agent. If you organize a bigger trip with flights, several flights and then you have rental cars and you need to know when do we have to drop off and all these. Tim, it's time consuming. Right. And you book the right hotels and so on. If your travel booking agent knows all that already from your previous, uh, experience, that would be very helpful. So I just say a few comments what I would like to have and he's doing all the rest and organizing booking everything. That would be cool.

Speaker A: Yeah, I want both of those.

Speaker B: Yeah, I was about to say. Yeah, if you make that one, I'll take it.

Speaker A: All right. Thank you very much, Michael and Alvin for taking the time to speak with me. I've learned a lot about industrial AI, what's possible and these AI agents. That sounds like a really exciting journey ahead of us to our listeners. Please stay tuned for our next episode. And if you'd like more information on our topic today, you can visit siemens.com industrial-copilot this is talking Digital Industries. I'm Alex Chavez. Thank you for staying with us.