AI for sustainability – Ep. 2 transcript
Welcome back to episode 2 of AI for sustainability on the Scaling Sustainable Impact podcast. If you haven’t listened to episode 1 yet, you can find it here.
Victoria Carlos: It’s awesome to hear both of you, so passionate and excited about this topic and to see how much potential there is with developments in AI specifically for sustainability. What kinds of AI are essential for these industrial environments? And how does Siemens make sure its AI solutions are optimized for these demanding settings that you’re both talking about?
Pina Schlombs: So where we see what is very crucial in the industrial space is that we call it applying industrial grade AI and industrial grade AI really needs precision. And what does it mean for our understanding and how we apply it? It means it needs three key characteristics. So the first one being trustworthiness. This means that we need to understand the data and their sources that are used to feeding and training the models.
Now, we also need to understand how biases work in the industrial space and there are biases, but they’re different to the consumer space where we are more used to using AI, right? So there are even biases against different machines, depending on what the data of the training model might look have looked like. So correcting for that is a very crucial element of making it more usable as the output. The secondary characteristic is the accuracy. So that means that the answer that it gives us actually matches the real world. So, since we’re using the digital world right to make the real world better, it needs to match in the output the real world as best as possible. And that’s where the accuracy comes in.
Now the last and third one is reliability. Having worked with AI, people would understand that if you ask it the same question over and over again, it might give you different answers. And this is actually a phenomenon that in the industrial space is not very helpful, right? So we need to employ measures to correct for that or mitigate these kinds of behaviors of models. In other words: trustworthiness, accuracy, and reliability. I really like the key characteristics of industrial grade AI. And one other thing that I think makes industrial grade AI really powerful is that we can feed it all the domain expertise that we have in the industrial space. We’ve been talking about engineering, we’ve been talking about production, we’ve been talking about supply chains.
These are all very specific industry domains and expertise and all of that can be fed into the model so that they specifically understand and specifically can help us in these domains. We’ve been talking on hearing a lot about large language models, but we often need like the large language model sometimes we we can very much deploy like small language models, right? Right sizing as a keyword here. To the to the task at hand, so that also the impact of inference that we have with a training and using of the models is minimized.
Lastly, I would say that looking forward, I’m very curious to see and very excited to see how agentic AI is going to transform and take us really to the next level and giving us the agency flexibility but also the speed of having expert domain agents specialize in their fields, orchestrated by humans to really get us a jump ahead in the speed and scale of how we can deploy and leverage the results of use of AI.
Victoria Carlos: I want to circle back to the circularity and circular economy topic for a moment. No pun intended, but we are seeing companies realize that the potential for circularity to be a profit driver in addition to an enabler for sustainability. How do you see AI enabling greater circularity in industrial operations? And I think, Eryn, you can get us going with this one.
Eryn Devola: I think the first thing to always recognize is if we look across the planet, we have few regions with really abundant resources and we have a ton of regions with really abundant waste. So how do we turn the waste of today into the products of tomorrow, while also making the products of tomorrow last longer? And as Pina said at the beginning, make sure we’ve got a plan for that next life of that product of how do I disassemble it. How do I repurpose it? How do I replace as little as possible? In that particular process, really what it reminds me of so much is the manufacturing philosophy.
I didn’t mention this earlier, but I love factories. I grew up in factories and we’ve been working in that space for so long to eliminate waste, to work toward a lean program, and for me, circularity is really just the next level of lean and really looking at any type of waste you have in your production or. In your product as a design flaw that we should be addressing with that 80% up front. And so this is where we kind of go back to what Pina was saying about how important AI is to get us to design the right product in the right way, so that it’s easy to manufacture, easy to use and operate, and then easy to redeploy at the end of its first useful life.
And so this is where I really see AI as being a a significant tool there. If we look at products we’ve already produced, this is where I think AI might have a great way for us to turn some of that waste into the into more usable products. So often products are removed from service because they aren’t in fashion anymore in the consumer side, because maybe an entire production line is no longer going to be used. But there’s components in there that still have quite a bit of useful life left. And we can use AI to help us evaluate those products for how much life is left and then help us better streamline the specific repair upgrade that would be needed to return that to a full lifetime of service. So I see a lot of AI as a way to manage the ongoing complexity of the work outlined in front of them, an opinion I’m sure you’ve got some great examples or some great things to add here as well.
Pina Schlombs: Absolutely. I think the just picking up on the last point you made, there’s just figuring out when we get our products back that maybe have have ended their first useful life and then figuring out, OK, what are the best R-strategies we could employ, right? Balancing the the the effort we need to put in to make them valuable again for a next life, to OK what value can they create for us in that next life cycle and then deciding, OK, this is the route we’re going to take, right? Because that’s actually closing the loop here again to the engineering space.
I mean in that next life would it be reused? Would it be repaired, remanufactured, or recycled? There are so many options, but we’ve not classically been trained to look at them to evaluate at which stage of a product, which kind of our strategy be most useful to us, right? And this is again where we can leverage AI in closing also the loop back to engineering and then deciding, what strategies are we going to deploy? And I think just taking example of from the the Siemens world, Siemens Energy is actually one of the great examples. These Gas turbines that they produce, these powerful products that have a huge material value also inherent to them, great engineering expertise that flows into them.
So you want to be keeping them alive for as long as possible. And what I find interesting in the projects that we’ve been doing with them is when they take these huge gas turbines back and they they can actually employ analysis technologies on determining which blades are still healthy, or where there might be footers that might not be visible to the naked eye but can be detected using X-rays or other technologies and then be able to determine this kind of defect we must employ our strategy X. So they have been doing additive manufacturing for these kinds of defects which is actually more helpful or more valuable than basically breaking down the whole thing and or exchanging one blade. This can actually minimize the effort that you need to be taking to extend a lifetime and repair a complex product such as a gas turbine and using the conversions of different technologies.
And this is what I think is fascinating here. What they’ve done is basically take a comprehensive digital twin. Then they gain the full understanding of how this product is is made, this is how a 3D model with characteristics on a specifications looks, then the understanding the health status that it comes back with. Then they’re able to analyze and determine to get it back to health, well, these are the technologies that we need to employ and operations we need to do to be able to use it again. And so multiple different technologies coming together here, as I said, from digital twins, multi physics simulations, AI right powered Internet and additive manufacturing coming together to be able to extend the life of such a such a big product here from our own Siemens family.
Eryn Devola: You know, one other example from our Siemens family, one of the factories we have within our DI portfolio in Fürth. There we have a bit of a mixed model where we make a lot of products in the same facility and we’ve been able to then extend we’ve got on site repair. We’re extending that to do that lifecycle extension in the same facility where you’re doing new product creation. And we’re doing that now for over 310 products and growing that by the end of 2026 to over 2000. So again, it opens up a whole world for us when we can do remanufacturing and repair to products and redeploy them at really a lower impact and therefore a lower cost by reusing those things that may have looked like waste to us, you know, 10 short years ago.
Victoria Carlos: Pina, you mentioned something earlier that AI allows us to see into the future and it can be a really great tool for that. So in your opinion, what is the best way for businesses to begin integrating AI into their operations, and where can they expect to see the largest sustainability gains, whether that be in the long term or in the short term?
Pina Schlombs: Yes, so there are four or five main steps that companies should follow. The first one clearly is and this is actually sounds trivial, but gets jumped over a lot that I see is clearly defining what you want to achieve with applying AI right? And then start small and scale right. Find once you have determined what your goals are within that. Find one that is small enough for you to start with where you are at today and then also reuse with pride. Reuse existing capabilities that are there today. Expertise. Know how that you already have? You might be surprised. There are so many colleagues that I have who’ve been tinkering with AI just in their own time in their private lives.
Tap into that potential that you have there in the in the people on your teams and let them use them for use cases that you’ve together defined as being useful for your goals. And then I would set to build the base, really ensure that you have the data that you need in the right quality that you need it at to feed the models to achieve the goals that you set out in the beginning. And this is a a point where really a differentiation in the speed the companies are able to adapt and deploy AI is often keep companies are very enthusiastic to start with a because, of course, all the promises that we hear are very tempting. But then they see that there is a certain digital maturity that is needed.
Certain databases can actually leverage AI from that base. So this is something that I would also look into.
Where do we find the use cases to start with? And do we actually have the data in that space to employ AI effectively? And then lastly, I would say to foster a culture of exploration. Get people to explore using AI again experience and then skill up, reskill and educate people in a broader scale, but make it
in a way that they can get to know this new and powerful technology. For some it might seem daunting. For others, it’s super exciting in a playful way, to let them let them explore and then also let them find new ways of doing things that they have done classically, maybe on a day-to-day that they thought mundane and now can use AI companion, a copilot, to fly with them.
Listen to the full episode here.
Siemens Digital Industries (DI) is a leading innovator in automation and digitalization. In close cooperation with its partners and customers, DI is the driving force for the digital transformation in the process and manufacturing industries. With its Digital Enterprise portfolio, Siemens provides companies of all sizes with all the necessary products, along with consistent solutions and services for the integration and digitalization of the entire value chain. Optimized for the specific requirements of individual industries, this unique portfolio enables customers to enhance their productivity and flexibility. DI continuously extends its portfolio to include innovations and the integration of future-oriented technologies. Siemens Digital Industries, with its headquarters in Nuremberg, has a workforce of around 72,000 people worldwide.
For more information on Siemens Digital Industries products and services, follow us on LinkedIn, Facebook, Twitter, and YouTube.
