Welcome to the Washington AI Network podcast. I'm Timmy Haddad, the founder of the Washington AI Network, where we bring together official Washington policy makers, DC insiders, and AI experts who are challenging, debating, and just trying to figure out the rules of the road for artificial intelligence.
Hello everyone. It is a late summer day, so it's great to see such a big crowd here. I'm Sumi Somaskanda. I am in for Timmy Haddad on the Washington AI Network podcast. Timmy may not be here, but she does listen, so can we get one big cheer for Timmy and the podcast? Thank you. Thank you. I've done my job now. Welcome to the CTA Innovation House for a conversation about the infrastructure powering artificial intelligence. Very pleased to say my guest is Dr. Roland Busch, the global president and CEO of Siemens AG. Roland, welcome to the Washington AI Network podcast.
We are two Germans talking about technology. I promise you it's going to be exciting despite that, I promise.
So, Roland, when people hear the word AI, you know, they have all sorts of concepts that go through their head, but you talk often about industrial AI. What is it?
Well, AI as we know it when it started off, it's a general purpose technology. And it surprised us because when you just ask a question, you can ask it in natural language, and you get amazing results. However, when you ask something, you get an answer that can be right, it can be wrong. Sometimes, it's completely wrong, it's hallucinating. So, the quality is, let's say, questionable. And if you want to deploy AI technology on the shop floor, it has to be deterministic. So, AI is a non-deterministic technology, and by the way, this is not a bug, it's a feature. You want to turn this technology into deterministic outcomes. So, when you tell a robot to do something and you do it over and over again, you expect the same results on a shop floor and not that the robot does something the robot thinks it's the best to do.
What that requires, number one is that the models which we use today, that they are trained on industrial data. Most of the models didn't see all the industrial data yet. For example, design data, which we store in our software, which tell you how a product is designed and how it works. It is time series data, very special animal. It's videos and the like. And then, once you have that, then you have to have models which are really trained in a way that you have deterministic results, meaning they do exactly what you want to do. So, that means if you program a PLC for industrial controller, it has to do the job precisely as you want it to do. That's industrial AI. It has a huge potential to change the way how we work, how we design products, how we operate manufacturing sites or assets in the field. And eventually, this is the biggest way, the biggest lever to monetize all the investments which are currently spent on AI.
Which is a lot, we should say. So, when you say shop floor, can you help us paint a picture? What does a shop floor look like with industrial AI?
So, let's talk about a welding shop. So, you're running a car manufacturing plant and you want to weld, let's say, a side door of a car. You have a robot, you want to program the robot. And what one of our products does, which is one of the first really proven AI application, we call it Eigen engineering agent. You tell with a prompt that I want to do this robot a welding job for a side door of a car. What the Eigen engineering agent does, it reaches out for the documents, the design documents of the door, the robot controls, everything what it needs, and then it starts programming the robot or the PLC in order to do the job. Then, once the program is done, it runs it through the compiler to see does it work? And if it doesn't, or it doesn't do really the job as expected, it goes over again until it really works. And then you have a chance to push the button and say download the software, push the button, and the robot does the welding job.
How far are we away from that?
There you go. You heard it from Roland Busch at Siemens.
It's called Eigen engineering agent. It's on the market. We have more than 100 customers are using it. And that's really very powerful because normally you need a trained engineer. And this example what I gave you, it takes maybe a couple of days for somebody to program and to really get the right results. Now, you do that literally in minutes.
How does that fit in with the discussion on reshoring? Because there is a big push here to rebuild American manufacturing, as you know, it's a big theme here in the US. Is it real or mostly ambition? And where do US partners and allies fit into that discussion?
So, if you talk about reshoring and the industries which are in focus are, let's say, pharmaceutical, a lot of investment coming here, so life science. It is semiconductors. It is even some car industries or defense, any kind of things which can think of what you need in the future. So, in order to do that, obviously, you need labor, skilled labor. You need to have a higher productivity because the labor costs are higher than in low-cost countries. So, what you want to do is you want to reshore, but not reshore what you offshored in the past one to one, but really go with a much much higher level of automation and digitalization.
It creates new jobs. These are different jobs, and eventually you cannot re-industrialize this country with the same kind of labor intensity which you had in the past or which you have in low-cost countries. You want to change that. And the positive news is once you re-industrialize and you do it right, you create new jobs, but you create different jobs, jobs with higher level of automation, people who are dealing with robots rather than doing handish work, repetitive handish work. But this is the only way how you can do it, I strongly believe. And in the United States you have all the ingredients. So, you have number one, you have a lot of domain know-how, you know how to build semiconductors, you know how to build molecules if you want to pharmaceuticals. You have the AI technology, and you have a lot of good people to deploy it. Yet, you have to train the workforce.
And what about those partners and allies? If you're reshoring for the US, how does that work with some of the US's partners?
If you talk about Siemens technology, we need strong partners like Microsoft. One very strong partners of ours. It's not only that they run many data centers for us, but it's also about the AI capabilities. We need the whole stack. So, that means it starts with the data centers, it needs the middleware, and the middleware is basically connection between the silicon, the GPUs, and the data. You need applications, and many applications we are running an open system, so we have a lot of applications which come from either our customers or third parties. So, it's a whole ecosystem. There's one element in it which very often we forget, is whatever you talk about, manufacturing a car or let's say semiconductors or whatever you want or even pharmaceuticals, you need the ecosystem behind. So, the suppliers for components, parts which you need, this small and medium-sized enterprise layer is very often forgotten. Without that, you have no chance to reindustrialize. You cannot think about exporting everything from the outside. So, you need also to talk about the small and medium-sized companies. They need to increase their digital capabilities as well.
Another component in all of this is the people, of course, the workers. If you look at surveys, a lot of Americans are certainly worried about what AI will mean for their jobs. How do you deploy industrial AI in a way that keeps the people at the center still?
Number one, AI creates jobs. And we invested in the last couple of years a billion in manufacturing capacities, creating new jobs. I just come from meeting with a senator from Texas. So, we create in Fort Worth new manufacturing line for medium-voltage and low-voltage and the like. So, the point is that you have to think about AI not taking jobs away, but AI creating new jobs. In some places, jobs will be gone, repetitive jobs. Sometimes you even don't want to have them. I mean, heavy lifting of a cockpit into a car is nothing you want to do. It's anyhow replaced by robots today. You can do much, much more. But it creates new jobs because you industrialize, you bring more value added, and it creates jobs which are more geared for the capabilities of humans. In Germany, for example, where we are about to build one of the most industrialized and most automated plants in the world, we have workers which used to do labor work on the shop floor, and they are now programming robots.
And they had to be retrained for that?
Absolutely. We have dedicated training courses for our blue collar workers. Some training courses they go for couple of weeks, some go for 2 years, depending on what the starting point is and what they are supposed to do. And I recently had talked to one of them and he was running through a more than 1-year training and he's now dealing with robots. He can program robots and automate a whole cell on his own and that's the new jobs which you see.
That sounds like a really good model for here in the US as well and Siemens has actually committed a significant amount to US investment and jobs. Why bet on American manufacturing now?
