#133 AI driven Smart Infrastructure
Show notes
We hear a lot about artificial intelligence, whether as a key enabling technology or company strategy. But where are we with AI in the buildings and infrastructure industry? What are some examples of AI driven applications? Find out with Jon Lester and Ulli Waltinger, Associate Vice President for Artificial Intelligence and IoT at Siemens Advanta Consulting, Fellow of the Siemens AI Lab and Host of the Human & AI podcast!
Transcript
Hello and welcome to Buildings of Tomorrow. My name is John Lester, and in today's episode, we are talking about artificial intelligence. Everything AI and how we are using AI and what are the next steps for AI in this industry around infrastructure and buildings. I'm very happy to be joined today by Uli Waltinger. He is the associate VP of AI at Siemens Adventer. And the technical fellow of Siemens AI Labs. On top of that, he also hosts and creates his own podcast called Human and AI. So I really do suggest if you're interested in this topic, go and check it out because he has great conversations with colleagues and compatriots around the industry as well. Uli, thank you so much for joining us today. Hey John, thanks very much. What a, what a welcome. Oh man, awesome, right? That, that, this is how a day should start. This is how a day should start, you know, it should make you feel good, but, uh, you know, and also at the same time, a lot of words that only say such a small piece of, of, of, you know, let's say the knowledge and the expertise you bring to this topic. So thank you so much for joining us. I'm really excited for the chat. So am I. So, so let's, let's start a little bit with AI. It's something that we've spoken about before, of course, AI is, is something of a buzzword or, or a topic that we hear a lot about, um, you know, whether it's from, you know, from companies talking about what they do, or whether it's from, from other technology that we use in our daily lives. If we look a little bit at the, The the AI space today, uh, and, and artificial intelligence in this corporate space, in this technology space, in this industrial and infrastructure space, where are we up to? How are we traveling in implementation and really using this technology effectively? Yeah, that's a big one, right? AI has been quite, quite, quite on hip and hype actually, you know, on both levels, and I think it's, it's, you know, established itself as a key enabling technology. Uh, on the board level and the company strategy, which we see on AI first strategies and AI first companies narratives around that. And if, if we look a bit closer, then we can see that we are somehow on the convergence of value chains, what I call the AI and IUT which is uh from my perspective, separated in two different pillars, one aspect of the enterprise AI aspects like harnessing AI and automation to transform somehow the business processes, sales forecasting, demand forecasting. Autommation scenarios and the second one is the industrial AI perspective, which is more harnessing AI and control technologies, you know, to thrive for the transformation of industrial applications. And, I would say that the combination of it, the connection of both sides, enterprise and industrial world, right, is that what enables us, the business models and the scaling across the value chains in that. But obviously, somehow the AI magic happens in, you know, in those kinds of nice hotspots where, you know, human needs are somehow, you know, positioned, but also human expertise is leveraged, so this kind of combination makes it quite nice in in providing. value. And if we look down on a bit of an enterprise level, then we see that most of the enterprises, you know, start the AI journey, right, towards an AI operating model, right? But if you then see how many of these initiatives, PUCs, prototypes, you know, projects um have been actually then ending up at the, at the value, at the the deployment, at an operationalization that we see that only 10% actually of these kinds of aspects are actually. Uh, leveraging that value out there. That's a really interesting fact to see that only 10% are making it. And do we see a difference across different industries or different applications that people are trying to achieve some things and, and they work more often than not and trying to achieve other things and they never work. How do you see that perspective? Yeah, that's a good one. So, um, I see two different sides. So you see the small and medium enterprises, I think they struggle rather on, on the ROI what we call the return on invest, you know, investing a couple of 100 cases or tens of cases of more than five digit numbers or six digit numbers. And something to explore is a hell lot of money. So um here, it's, it must be clear, you know, what is the return on that invest in these kinds of aspects and one of good partner set ups in helping, you know, joining that because they may not have an own data science and AI team, an expert, the competence center, right? But, uh, you know, demand on, you know, motivation, but also capabilities and clear structure on the return on site. Large operations, you know, and enterprises struggle rather on the end to end solution, I would say, you know, there are, you have all this. Massive, you know, of IT infrastructure, you never give up an ERP system and then whatsoever system, right? And therefore, you know, going end to end in an API first paradigm is of whitest importance, but still struggling on that, right? Uh, overall, I think on, on an executive level, right, we see a bit of a foundation level which has three different elements in that. The one is the aspiration of scalable returns, um, you know, if you, if you draw, I'm a bit of consulting now here, right, but McKinsey's, you know, values AI around 3.5, maybe to 5 trillion US, right, which can be unlocked to AI. Obviously board members want to have part of the shares and say like, hey, how does that contribute to our boards and bills, right? The second aspect is the industrial great aspects and design aspects of that. You know, um, most recently, the European Commission, you know, launched the seven key requirements, um, that AI systems should meet, that means it's also not only benefits, but also it demands some requirements in terms of safety, explainable security in these kinds of aspects. So industrial grade is a difference than just deploying a consumer tech product. And the third one is, I think we'll see a lot is meaningful attention, you know, and most recently, the Nature paper, you know, um motivated um that 9 79% of all SDG targets sustainable development goals, right, can be positively impacted by AI and it means technology and let's say, uh, And or with purpose, as you know, Siemens somehow puts us, that like the combination of a meaningful intention of technology and innovation, and a good responsible business um has entered not only as a narrative, the boardrooms of corporates, but AI is playing a vital part in that and that's super exciting. Yeah, that is really exciting, especially when you start to talk about that scope and, and. And when you discuss the enterprise targets and the business functionality, then the industrial targets which are more on processes, when you start to put it in the scope of sustainability goals as an example, which we know impacts every single piece of this, it starts to tie these two worlds together somehow, which, which also brings up some of those challenges like you described that end to end requirement, you know, the challenge to really take an approach that broad and that holistic for a small to medium sized enterprise, you know, this is a These are not small challenges to undertake, especially as you, you bring in that point of, of once you start deploying this in a, in a more productized manner, the requirements, the security, the, the, the, the consistency, the reliability all start to come into play more and more. How, like when we take a look at the industry in general, so, you know, this is where we're up to today, and And we know that this is something that we've been trying and doing more of, and you mentioned even with all of the pilots and all of the proof of concepts that we have, 10% really make it into something that works. Where are you seeing improvement? You know, where are we having some wins and where are we getting better as an industry in deploying and leveraging this technology? Um, yeah, so, um, I think we, we can separate again to 33 different kinds of paradigms. The one I call digitalization on steroids, the second one is automation on steroids, and the third one is correlation on steroids, you know, digital. Digit digitization on steroids, right? I think the current pandemic pose a very effectively and efficiently as a magnifying glass where the improvement spots and levers in digital technology can be there, right, gaining through smart automation. Automation obviously is something which once you have digitalized processes, which, which helps you enabling it further and gaining more out of it. And correlation of steroids, right, is something which adds probabilistic uncertainty estimation in the vital role, right, identification, classification, forecasting, prognosis and semi or autonomous control. Um, so that's, that's quite, quite an awesome, um, aspect and with us along, right, the major challenge that we had prior was, you know, uh, limited few of the portfolio, limited few of the assets, limit visibility, let's say in building on carbon footprint or on asset performance on asset tracking, a lot of data was in silLAS and these kinds of aspects. Now we enter more and more uh the level of, you know, uh of the AIOT and that means, you know, connecting devices, and we are moving from connecting the people paradigm, you know, social media and social networks, uh, which I call the embodiment stage of machine intelligence, that we have smart devices which has something in there, right, which uh guides us through the daily journey now to an increasing level of. Machine autonomy, right? Currently, we estimated 21 billion active connected devices worldwide. It's estimated at 30 billion of IOT connections. So that means, you know, 4 IOT devices per person on average by 2025, right? Which leading use case are still predictive maintenance, quality inspection, manufacturing process, supply chain and uprising will be a hot stack, AI driven cybersecurity. And that means from the data silos perspective and the data availability, which is pivotally key in leveraging, you know, value out of data, we had some improvement and not only the connectivity, but also on the semantic side, understanding domains, understanding vocabulary and the semantic connection to each other. And I think this brings us to the, to the point that we are currently, obviously would say we are living in a data driven society. So that means for the first time in history, we can see enough about ourselves and our places and products, where, you know, we can, you know, rethink buildings, where we can rethink infrastructure, where we can rethink problems and and products around it. And I think Jeff Merritt from the uh the BF actually put it very nicely, right, connected device um makes our life easier and we need to see AI and IOT as a design language, right? When we suddenly have the capability in having real-time data. So you're not actually taking guesses, right? What, what could be an option, but you can just act in real time and this is, this is pretty dope, right? And this is pretty awesome because that means um complexity, you know, is a lot of complexity can be automated. Um, you know, lights out in the workplace or, you know, smart, smart conditions, smart experiences in, in the buildings and stuff like that. And some of them, you know, will lead to a dramatic shift in improved and awesome experience. And this is where, you know, data semantic and AI, especially predictions come into place where I say like, that's pretty dope. And, and what you just described there, I guess is a challenge that That whether it's AI related or just data related, this is something that the industry has been struggling with for a long time, right? We're so much better at collecting information, but actually using it and turning it into something valuable, turning it into actions has always been a challenge. You mentioned semantics or semantic data or tagging and approaches to understand and automatically, you know, you know, get some insight and start to leverage this information is such a big step towards that. And that last point that you mentioned where we start to shift away from some of the standard applications where we talk about predictive maintenance and, and making estimations to actually driving this stuff real-time, how, how do we make that mindset shift when we're, when we're still not there yet in, in the applications that we have already defined as a focus? How do we get ourselves then to start putting our our, our minds in the place to start talking about real-time. Uh, activations like this and, and moving past the, the estimations, the predictive approach, and starting to talk about real-time automation. Yeah, so that means more into the actually applications aspects, right? And then that, you know, I think the motivations are there. If you, if you open up the world and look out there, right, then you see that, you know, traditional industrial technologies, you know, um, come, you know, increasing more important and are fused with digital technology, distributed cloud computing, right, and stuff like that, and they passed somehow the way. What I would say the sustainable future, right? If you look on, on the cities aspect, right, Energy amount resulting in CO2 emissions, right? It's counting more than, I don't know, 70, 78% probably, right? of the energy consumption, 70% of our CO2 emissions come from there. Heating, cooling, ventilation, right, pivotal aspects actually a big lion's share in that kind of energy in a mod. So the other question is, you know, how can we use responsibly these technologies and the elements that we have connectivity, data, semantics, in order to enable that and. Um, that means, you know, working not only alone, right? And that means a mindset of not only task and try and, and failure culture which a bit of, you know, most frequently being said, but actually, you know, how do we, how do we shape communities together in for us as industrial player to make transparent what the challenges are and. For us also given other colleagues outside the partners, the students and the enablement with working with data, working on these challenges and making a difference and moving them from just a pure cat classifier on a, you know, synthesized data set to a real world champion, uh, uh, you know, challenge, and a couple of examples what I think is quite nice is. Most recently, we launched the city performance tool aspects that more on the, let's say asset transparency aspect again, so, um, you know, how what are the indication of infrastructure, what kind of decisions may influence what kind of implements or um growth or infrastructure sector, so that's. Getting to know and knowledge about that and this is also what I think the expert team has made kuda to you if you listen here, you know, what makes you the team around you makes really awesome um to see where, you know, on real time is which data and how we can leverage it. I think the second one is especially in compute, right? Data centers are quite, quite heavy being used, right? And I think there's, you know, using machine learning to reduce the cooling energy demand, right? This is a partnership actually with Vigilant, right, um, on power users effectiveness with a hard KPI, but I think you save energy by, you know, cooling the right aspects and cold islands and hot islands using machine learning technology. That is pretty dope actually, to be honestly, it has quite some impact because ICT communication and with the need for compute will just increase. So making smart AI in there is awesome. I thought, maybe a quality detection, right? Um, Obviously react and prognosis, but also, uh, you know, putting as we did it in City of London, right, um, maneuvering and getting transparency, but also obviously predicting. And, and improving the service procedures there, visual quality inspection for assets using drones or um you know, mapping risks in order to improve service fleets or, you know, investment spending. So vision, especially in the eye, gives us the eye, obviously in the field, that's pretty dope. And maybe last but not least, the building, uh I think that's a pretty dope project also in Siemens, kudos to Siemens technology and the Smart-Infrastructure folks here because it's awesome, um, building digital twin, which is what is exactly, so it's actually a knowledge graph, so a graphical representation of the assets which powers the relation of, you know, unified and semantic typified information. It's using design. Information like IFC format, you may want to know the information of the allocation of the rooms, but also in combination with time series sensor information. And from a data geek as me, as I am, that's pretty awesome because I've seen that's one of the largest knowledge graphs initiatives apart maybe from the Energy twin, right, which shows, you know, how we can enable applications, we call them companions on top of this repository. Yeah, awesome. Those 2 or 33 or 4 examples that you just gave in different applications, the best thing about many of them is they build upon. You know, so many of the technologies that are almost standard for the industry, but they add additional knowledge, they add additional capabilities which enable us to do those jobs better, and that's that shift away from the standard automation, the standard design of a control system, and start to enable that control system to take actions in an autonomous manner and make those chass. So you talked about. You know, the compute demand and data centers becoming more and more driven also by different data regulations, etc. We start to see data centers being put in many more places in many different ways from small to medium, not just, you know, the really massive scale, and all of them still have this challenge to try and tackle whether they're a huge piece of infrastructure or small, and You also mentioned the expo, which I really like because I know there there was a big part and I haven't had the chance to visit, so anyone out there that's listening that has had the chance, I'm a little bit jealous, but I do know the sustainability spaces where you can go in and you get some real-time information about how many people are, are in that certain space and then what the impacts and what's possible to really achieve using different technologies and learn from that information in real time and show the impacts of that. You know, I think that's, they're just really exciting examples where we do make that shift away from, from making some decisions on a theoretical perfect world, designing a system to run in a certain way and then setting it free, even though this is better than, than not having them, of course, bringing in this, this, this learning capability. And, and maybe a simple question, but an important one is From your perspective, is this the difference? Is, is this what, what those applications where we could say, listen, if you had someone sitting there, if you had the knowledge and the capabilities, and you would then make a decision, uh, in a different way or make a change, are those, those the applications and the opportunities where, where AI really. Uh, it is ready for us to take action to, to at least investigate those, those opportunities and check out what we can try and achieve when you get these different people together and we start to have this collaboration. Is that a reasonable assumption? Yeah, I, I would say so, right? This is, this is what, um, um, also the demand of the younger generation, um, I see. Uh, we launched different, um, uh, you know, formats from a residency program, research residency, business residency, residency program. Um, and a lot of the questions of the younger folks who can maybe choose their employees right away, choose the companies, right? …