Sustainability Symposium 2025 AI & Sustainability: Game-Changer or Apocalypse Machine?

00:00:10:21 - 00:00:45:10
Mike Collignon: This next session here might be covering the hottest topic out there right now. And that is AI. Now here with us is Sneha Kumari. She is an accomplished supply chain operations leader with over 15 years of experience, specializing in driving operational excellence across diverse industries. As the founder of Merlin AI, she leads a team dedicated to integrating the latest advancements in AI and technology into the construction sector, revolutionizing project efficiency and effectiveness.

00:00:45:13 - 00:01:09:23
Mike Collignon: She's also the Head of Industry Trends at Circular Supply Chain Network, which is a global community of professionals who are passionate about making supply chains more circular and sustainable. Now, her session is titled AI and Sustainability Game Changer or Apocalypse Machine? Yikes. I certainly hope it's not the latter. Sneha, welcome to the Symposium.

00:01:09:25 - 00:01:38:12
Sneha Kumari: Thank you so much, Mike. I am very, very thankful for the opportunity and hopefully we have it in person someday so we are able to see each other as well. Excited for this and sharing more from my experience for the next 30 minutes. And thanks Sarah for helping keep me on time, because I tend to really talk a lot about when it comes to this topic specifically.

00:01:38:13 - 00:02:13:17
Sneha Kumari: I'll be going over a few slides and we will talk more about AI. A quick intro. Been in supply chain, done everything under supply chain, which is strategic and tactical. And really it's been it's been a great journey just being in supply chain touching every aspect that our product lifecycle really touches. And because of that holistic understanding I really have a good understanding, I would say from an impact standpoint on how things could be changed, things could be made circular, how what are those sustainable practices?

00:02:13:19 - 00:02:36:03
Sneha Kumari: And then finally, being a founder of an AI company, I myself am very, very conscious of every single word that goes out there with all the different elements we are dealing with and how expensive it can be. Awesome. Well, thank you all for taking the time. Good evening, good afternoon, good morning. Whichever time zone you're logging in from, I am super excited to take the next few minutes.

00:02:36:03 - 00:03:00:29
Sneha Kumari: Talk about this. Would love to answer some questions towards the end. And of course, feel free to connect with me on different social media channels. LinkedIn preferably. That's where I find myself most and would love to. Definitely not on this topic sometime if you would like to. Let's get into the slides.

00:03:01:01 - 00:03:12:26
Sneha Kumari: So let's talk about let's talk about AI the and let me know if I guess we are able to share. See my presentation. Right.

00:03:12:29 - 00:03:13:26
Sneha Kumari: Guessing. Yes.

00:03:14:01 - 00:03:16:13
Sara Gutterman: Yes I'm confirming.

00:03:16:15 - 00:03:18:28
Sneha Kumari: Thanks.

00:03:19:01 - 00:03:49:09
Sneha Kumari: All right. So again thank you for joining for this discussion. As the title of the opening slide definitely pretty much captured the essence for you. Right. Is is it a double edged sword that we are talking about? We stand out a fascinating and probably somewhat terrifying inflection point. Right. Today we are talking about I would like to walk through some key points, which is, you know, exponential AI growth opportunity risk we are witnessing.

00:03:49:12 - 00:04:23:25
Sneha Kumari: Immense amount of computational capabilities that's needed. That's probably doubling approximately every 3 to 6 months in our space in the AI space, by the way. And this acceleration brings unprecedented opportunities for innovation across sectors. Right. It could be former medicine, agriculture, building construction, energy everywhere. Hence, you know, and the same exponential growth creates risk of course, right. In terms of energy consumption and resource allocation and potential.

00:04:23:28 - 00:04:51:17
Sneha Kumari: Exacerbation of existing inequalities. So those are definitely some things that I definitely want to talk about. Of course climate crisis. We there's a need for an urgent and scalable solution. Right. The there was a later IPCC report that we have a narrowing window to avoid the worst impacts of climate change. And we need solutions that can scale rapidly and effectively.

00:04:51:17 - 00:05:28:08
Sneha Kumari: So the question becomes, can AI accelerate our response to these challenges, or will its own environmental footprint outweigh its benefits? And, you know, are we solving the problem or in some ways really becoming part of it? That's the central tension we'll explore today. Just to just for some background, like these systems require enormous. Again, as I said, computational resources and a single large language model training run can generate the carbon equal to hundreds of, you know, round trip flights.

00:05:28:08 - 00:06:09:26
Sneha Kumari: And I'll talk about some statistics eventually in our slide. But, you know, there since we are talking about so much data that exists out there that we are computing processing, these same systems have the capability to optimize our energy grades, reduce waste, develop new materials, right for renewable resources as well. So we'll we'll go go through both the sides of this double edged sword as I talk about this and explore ways where, you know, this can become a net positive force for environmental sustainability rather than another burden.

00:06:09:28 - 00:06:35:28
Sneha Kumari: But there are steps that needs to be taken that needs to be, you know, as users and also founders and whoever is, you know, working in this whole processing space needs to be cautious about and be looking at. Now, let's explore the the potential that, you know, in general, AI offers as as a tool for, you know, sustainability.

00:06:35:28 - 00:07:01:28
Sneha Kumari: And I want to highlight some key applications that are already demonstrating, you know, some promising results. You know, we're talking about let's talk about the smart energy grids, like we are talking about real time optimization that is reducing emissions. There are AI systems that are revolutionizing our energy infrastructure by predicting demand patterns. We are also doing that within our product.

00:07:01:29 - 00:07:31:18
Sneha Kumari: Like, you know what the what our customer demand patterns and probably getting ahead of it with pretty darn accuracy. You know, it's pretty accurate. And you know, the specially the machine learning algorithms can balance grid loads in real time and also, integrate intermittent renewable sources as well. Like we are talking about the the wind and solar energies more efficiently, like I do want to mention about Google's DeepMind.

00:07:31:20 - 00:08:00:02
Sneha Kumari: You know, they reduce the cooling energy in their data centers by about 40%. And I'll be able to share the report through just, you know, optimizing that. And we are talking about a lot. I'm going to refer a lot to data centers here, because that's where the whole computational resources and the load will be seen. And in construction, for example, like, you know, if if we are looking at AI powered building management systems, they are dynamically adjusting the operations we are all pretty familiar about.

00:08:00:02 - 00:08:21:12
Sneha Kumari: But then that's reducing the energy waste, right? I do want to touch upon the circular economy where and I don't know if how many of you are really familiar, but there's a difference between, you know, overall looking at sustainability versus just circular economy. Because and I often say this like, you know, when you're talking about economy, you're talking about monetizing.

00:08:21:18 - 00:08:46:03
Sneha Kumari: So how do you actually be are you how are you able to bring the waste back in your value chain, supply chain, whatever you call it, drag the materials and you know, how then are you taking measures to cut that waste? Our industry, the building industry alone generates about a third of all the ways globally, right? We are a massive industry trillion dollar industry out there.

00:08:46:03 - 00:09:15:27
Sneha Kumari: And AI will definitely play a transforming role in, you know, changing or transforming this whole material tracking system that can monitor the entire life cycle. Think about having a passport for your material at some point. For example, AI powered computer vision can identify and sort construction based on site. We are definitely looking at innovations where, you know, we want to increase your recycling rates dramatically.

00:09:15:28 - 00:09:41:19
Sneha Kumari: We again, as I said, digital material passports can really be a thing enabled by AI. And there are many other, many other fields. I do want to talk about design for disassembly, and I know there are a lot of experts talking about that. That approach in itself will be a fundamental shift from, you know, for a building industry from demolition to being able to recover the materials.

00:09:41:22 - 00:10:11:06
Sneha Kumari: I know agriculture is seeing a lot of AI driven. We'll be seeing a lot of AI driven precision farming. You know, drones, satellite imagery combined with machine learning can really detect exactly what areas of field would need water. And just in past year, a couple of years, I have been to some startup pitch competitions, and there are some really amazing ideas out there that can be utilized specially for the agriculture and farming industry.

00:10:11:06 - 00:10:45:20
Sneha Kumari: And lastly, touching upon the climate modeling, you know, how do we predict risks faster, smarter? Our ability to understand and respond to climate change is dramatically enhanced through AI. And I have to say that there are models that at a at a time would take weeks to run, and it could now be processed in hours. Of course, there's a lot that goes behind the scenes to be able to reduce that time, but that really allows us to simulate more scenarios, improve accuracy, right?

00:10:45:22 - 00:11:30:20
Sneha Kumari: For the building industry, this would translate to better predictive models for climate resilient design. I know a lot of peers have been working on such designs as well, so the fact that our AI models can allies thousands of these variables to determine how buildings will perform under various climate scenarios can also enable developers right for future proof infrastructure against all these extreme changing weather events that we are, we are seeing these examples just kind of represent and give you an idea of what's beginning to happen and its potential for a sustain to be an enabler, in this case, for sustainability.

00:11:30:20 - 00:11:58:01
Sneha Kumari: Some of these initiatives, the intelligent application of these technologies across these different sectors, can really offer a pathway to dramatically reduce our environmental footprint. But let's get into also the other side of it, right? We talk we are talking about the double edged sword here. So how can this accelerate the crisis. What does this whole training entail. The energy demand and and things like that.

00:11:58:01 - 00:12:13:16
Sneha Kumari: So I do want to touch upon that a little bit here. So now as we are, you know, talking about these amazing applications, we should also accept its potential to.

00:12:13:18 - 00:12:40:20
Sneha Kumari: You know, the impact it will have for these environmental challenges. So what I want to talk about is let's talk about the training here. Right. The computational resource as as I mentioned, just for, you know, training a large AI model like GPT, it's it's staggering, a single training run of one LLM can emit as much as 300,000 miles of car travel.

00:12:40:22 - 00:13:07:26
Sneha Kumari: Just to put something in perspective that if if we put it in perspective of a construction, we are talking about a thousand tons of carbon emissions. That is similar to, say, typical office building. If that's what the footprint is, that's equivalent to what just an AI one AI advanced training model would take. Now this debt is incurred before these systems even deliver any value.

00:13:07:27 - 00:13:29:22
Sneha Kumari: Right now imagine the net impact of when the complexity of these models grows, like we have to have. This footprint exists even before we have started using these models. And this only goes and multiplies from when we start.

00:13:29:24 - 00:13:52:18
Sneha Kumari: Really talking to these models and pulling data, doing activities in our company, ourselves. We know that every. I'm sure a lot of you would have heard like a please and a thank you and any every single word that you shift to a GPT to ask question has a significant impact, and it costs money. And I am at the forefront of that because we are an AI company.

00:13:52:19 - 00:14:25:13
Sneha Kumari: We have we we use LLM models, the OpenAI models, and we know that we pay for every single click and question that our customers ask on our model, with our and we, we interact with the LM software's. So and the more complex tasks you do, the impact is more right. There's growing energy demand for all these data centers supporting all these LM modeling, the physical infrastructure supporting is all these models behind this is expanding phenomenally.

00:14:25:14 - 00:14:54:22
Sneha Kumari: Like rapidly, data centers already consume about 1 to 2% of global electricity power demand. And the this demand is only going to double or triple by 2030, especially the power constraints. And if you have been to any of these data center webinars, you will be aware of, you know, how how the demand is going up. The lead times have gone up, especially for the hardware that that's required to power up these data centers.

00:14:54:24 - 00:15:25:28
Sneha Kumari: In our in the building sector, we are seeing a massive surge in data center building projects. Right. That's creating a, I would say, a paradoxical situation where AI tools are designed to optimize being designed to optimal optimize building efficiency, are housed in facilities that drastically increasing the energy consumption. Right. Many of these facilities still rely on, you know, essential fuels despite renewable commitments.

00:15:25:29 - 00:15:53:20
Sneha Kumari: Not everyone is going green and renewable. And then mining is one example, you know, rare specially for the chips and batteries. And I wanted to bring that up because these hardware enable AI system. That in itself also requires specific, very specialized materials with significant extraction costs. Right? A typical smart building might contain thousands of sensors and processors, each with its own material footprint and the risk of efficiency.

00:15:53:20 - 00:16:25:25
Sneha Kumari: Rebound. Doing more and not consuming Lexus Is. It's more it's concerning it. What's more concerning is what economists call it like they call it a as rebound effect. Like when a technology makes a process more efficient, we tend to do more of it rather than reducing. Right. And it's also very typical in construction where, you know, we are optimizing AI, optimize building processes that lead to larger buildings to more complex features rather than reduce resource use.

00:16:25:26 - 00:16:55:03
Sneha Kumari: Right. So the danger of creating this vicious cycle, where efficiency gains are continually being offset by expansion of these activities as it's there. Right. And the challenge before us is, how will I serve this law of return rather than accelerating extraction and waste? So the construction industry, as one of the largest global consumer of resources, stands at a critical decision point right now.

00:16:55:04 - 00:17:41:25
Sneha Kumari: Like, will we be using it to fundamentally reimagine sustainable building practices or just, you know, do more of the same and faster? So those questions and concerns still lie. And because of these growing demand, there's a lot in tears that's happening to make these AI systems, to enable these AI systems, that has a footprint of its own. And what steps that do we need to take in order to control or at least work towards reducing that impact will be important at every tier and every user, persona or industry that will take that will participate in enabling that.

00:17:41:28 - 00:18:14:01
Sneha Kumari: Did want to touch briefly on the impact. Like as I said, training one AI model will be equivalent to five times a lifetime emissions. You know, of a car. I do note that there was a advanced AI system, that there was a research from the University of Massachusetts that they found that, you know, training a single LLM generate up to 626 some pounds of CO2, which is roughly the five, five times the lifetime emissions of an of a car, which including its manufacturing right.

00:18:14:03 - 00:18:38:09
Sneha Kumari: And in construction terms, I was looking at it could be equivalent to approximately 70 tons of concrete or foundation of a small commercial building. That's a lot. We are looking at global data center use being projected, as I said, could double triple by 2030. That can potentially reach increase our electricity demand by, you know, power demand by 8%.

00:18:38:10 - 00:19:06:16
Sneha Kumari: That's almost like that. That in itself is pretty jaw dropping and something that we should be concerned about, or at least immediately think about ways of how are we going to mitigate the risk, how are you building the infrastructure and, you know, for the consumption and, you know, what's our carbon budget? Are we even talking about that? Only 10% of companies, by the way, are using AI to prioritize sustainability goals.

00:19:06:16 - 00:19:34:21
Sneha Kumari: By the way, alongside performances, I do want to talk about the success stories. So as I said, Google DeepMind, definitely the Google has been a pretty big pioneer in terms of when it comes to data centers. They have been doing this a decade or probably more back where they have been using renewable resources to power data centers. This is not only cutting their operational costs significantly, but also, of course, reducing the digital footprint of their digital infrastructure.

00:19:34:22 - 00:20:01:10
Sneha Kumari: The implications can well extend beyond data centers. Right. But there are people, you know, companies doing that. We I did want to talk about Siemens. It talks about smart manufacturing, reducing material waste. It has I know that Siemens also is very involved in partnering with their contractors to bring similar predictive capabilities, and also being able to take actions to reduce the footprint.

00:20:01:11 - 00:20:27:15
Sneha Kumari: You know, as a bigger company, I can do that, and probably because I understand that we want to make sure that this knowledge transfer happens to the supply chain that exists, that we are all coming towards, and translating those cost savings and environmental benefits at all those tears as well. Of course, grid edge AI, you know, they are also optimizing real time renewable energy integration.

00:20:27:15 - 00:20:50:08
Sneha Kumari: This is in the UK. They their AI platform actually enables buildings to predict their energy needs in relation to the grid conditions. So I think that's a that will automatically then adjust the consumption and coincides with periods of when you absolutely need abundant renewable energy and make the switch. I think that is pretty, pretty amazing to be doing as well.

00:20:50:10 - 00:21:17:24
Sneha Kumari: I think I do want to, you know, it a sustainable pioneer here that the best technologies don't just succeed in the marketplace, they actually transform it. And these examples really suggest that AI has a transformative, you know, potential when guided by clear sustainability objectives. I think that's where that's what we are missing really quick. Wanted to touch upon in this slide where what's the green AI movement like?

00:21:17:25 - 00:21:41:29
Sneha Kumari: These are some of the steps that we should be looking at. There is an emerging green movement that's representing a fundamental shift on how we approach AI development, right. Rather than simply pursuing maximum performance regardless of computational costs, it will be prioritizing on resource efficiency along with accuracy. So this will mean designing algorithms that achieve results with minimal computation.

00:21:41:29 - 00:22:11:23
Sneha Kumari: And I know there's research, there's universities involved in this. And we need to look at the hardware requirements as well. I talked about renewable power data centers. We also need to talk about transparent and accountable AI systems. We cannot manage what we don't measure. And we all are pretty familiar with that. Now creating standardized frameworks for reporting the environmental footprint of the AI systems, from training to deployment to maintenance, that will be essential.

00:22:11:23 - 00:22:35:19
Sneha Kumari: And several several industry initiatives are now actually existing that is going to develop the, you know, energy nutrient, I would say nutrition labels for AI models that will disclose their carbon intensity. But we need to talk about how this becomes a policy at some time. AI is so new. There has to be some regulatory measures also happening. So we are all making sure that we are taking the right steps.

00:22:35:21 - 00:22:59:13
Sneha Kumari: There has to be policy frameworks that need to evolve to guide this development, right? Specially startups like us and even OpenAI and purposely and topic like how are they dictating the whole industry per se. Right. There has to be incentives to be able to, you know, making sure that we are pricing there's some carbon pricing for computationally intensive systems and things like that.

00:22:59:13 - 00:23:23:08
Sneha Kumari: So people are pioneering this whole concept in the right way. I'm a fan of William McDonough. He was a sustainable design champion, and what he said was our goal is to delightfully diverse is a delightfully diverse, safe and healthy world with clean air, water and soil and the resources. But we need to do it economically as well. And we understand that.

00:23:23:08 - 00:23:42:29
Sneha Kumari: But and ecologically and AI will be if we are able to direct this in the right way, it can really be a powerful ally in creating that work. Just wanted to, you know, say that it's not going to automatically save us, but if it's used in the right way, it might just give us the time to save ourselves.

00:23:42:29 - 00:24:06:04
Sneha Kumari: I think it's it's time that it just encapsulates the core message that I want. I hope I get across today that it's me. It's neither inherently destructive, not beneficial as well to our environmental challenges. It's just a powerful tool, and the impact will be determined on how we choose to design, deploy and regulate it. As simple as that.

00:24:06:07 - 00:24:25:04
Sneha Kumari: Where do we go from here? Like, yeah, I mean, we need to talk about building AI solutions that align with profit and planet, right? We are all here to win. So we want to do that, but we also want to do that sustainably by lowering our costs. And we want to hold technology companies also accountable to climate goals.

00:24:25:04 - 00:24:54:10
Sneha Kumari: There has to be innovation happening in that sector. While lots of innovation happening on the AI sector sector, how are we innovating on the sustainability sector and making sure we come up with some frugal innovation designs as well to develop lightweight and, you know, efficient AI systems? And we need to, I think in general, like demand better from leaders, innovators, ourselves as companies like I know within our company itself, we have taken a conscious effort.

00:24:54:10 - 00:25:21:28
Sneha Kumari: We are building a small marketplace. And so every skew that we bring on our platform, we are looking at open data sources where we are consciously putting out the carbon footprint of all the skills that our customers see. I think that's a very small effort, but we wanted to bring that up for our customers. So if there was a sustainable or greener or circular option, our customers get access to that and probably choose that over, you know, going towards and looking at virgin materials from the planet.

00:25:21:28 - 00:25:53:25
Sneha Kumari: So a small effort from my company, especially because of my association on the green side of, of the world for a decade now, is something that we are taking. But in general, we'll all have to be, you know, understand what our own sphere of influence looking looks like, whether it's building industry or, you know, and you know, or policymakers or just as citizens, we need to just identify specific actions that we can take to ensure that we are making AI as a powerful ally in our journey.

00:25:53:27 - 00:26:06:22
Sneha Kumari: But yeah, I mean, that was I can literally not on this for as long as we want to, but I this was a quick presentation. I wanted to touch upon the key points, wanted to leave five minutes for some of the questions. So Sarah, over to you for any questions.

00:26:06:23 - 00:26:32:15
Sara Gutterman: Wonderful. Thank you so much, Nina. That was fascinating. I think that, you know, there's just so many questions swirling around right now. With respect, I ai I have two main questions, and I think that's probably just about what we'll have time for. The first is what is the AI application in the building and construction sector that you are most excited about right now?

00:26:32:18 - 00:26:59:05
Sneha Kumari: So I would I would say that there are a lot of lots of those applications. I'm definitely super, super excited about building AI applications in the material sector side of things. That's something that's close to my heart, and it also talks a lot about the circular construction enabler initiatives that I have been also talking and writing about for some time now.

00:26:59:06 - 00:27:21:16
Sneha Kumari: Material trans passports traceability will be key, and AI is pretty adept in being able to do that. We are cells are actually testing some models like, you know, piloting some blockchain enabled material tracking systems, as I said, like looking for footprints and finding out, you know, I'm talking about circularity. So I cannot go without saying that. How do you bring secondary materials back in your chain?

00:27:21:16 - 00:27:39:23
Sneha Kumari: An AI will play a key role in being able to bring that initiative, bring that intelligence to our fingertips. And of course, safety, right? Safety has been a big concern in AR. VR technologies is going to get us there. I think that's something I'm super excited about. Definitely.

00:27:39:25 - 00:27:56:15
Sara Gutterman: Thank you. Now my next question is about ensuring ethical AI deployment. So what ethical considerations should be taken into account when deploying AI in general, but particularly in construction? With respect to data privacy.

00:27:56:18 - 00:28:22:01
Sneha Kumari: I think there's a lot like, you know, data privacy in itself intrigues me a lot. We have been talking about it a lot internally and also with our with our customers. But ethical use of all of this data is is not a new question. We need to understand who owns the data, the consent, minimizing and collecting data that's strictly necessary for, you know, a systems function.

00:28:22:01 - 00:28:49:10
Sneha Kumari: No over reaching. Beyond that, I think transparency will be key. And that's why I keep saying that we have just started seeing all these amazing startups blow up in the AI space, including mine, right? And I'm super passionate about it. But we also need more initiatives. As companies, as myself, I want to be a true, you know, enabler while I'm showing that I know what my footprint was and I'm being transparent about it.

00:28:49:10 - 00:29:12:24
Sneha Kumari: So with my customers, we are very in a in a very close way in our contracts. Also like what data is being collected, how it's used, how these decisions are made, whether it's on job site or scheduling or safety alerts. Like, you know, we have to avoid black box AI models, right? Whenever possible and opt for X, I would say explainable AI in in high impact areas.

00:29:12:24 - 00:29:29:08
Sneha Kumari: So biases and fairness I mean there are many tools like there is a lot of bias in hiring. We have read about that, especially if you use automated systems. We need to, you know, make sure that we are constantly we have metrics to track the performance and evaluate for bias as well. I don't know, I don't think we do enough of that.

00:29:29:09 - 00:29:43:17
Sneha Kumari: We need to ensure algorithms are not, you know, disproportionately impacting vulnerable groups and especially, you know, workforce management. So, you know, accountability will be important. Let's just summarize with that.

00:29:43:19 - 00:29:59:04
Sara Gutterman: And then the very last question in 30s or less, because literally that's about all we have. You talked about using AI to reimagine sustainable practices as opposed to just regurgitating what we've been doing. So what's the best way to do that?

00:29:59:07 - 00:30:28:29
Sneha Kumari: So there there are many different. There are different initiatives or probably steps we can take. But I would say let's start with understanding and mapping our own value stream. Right. Really understand where we are at. Whether I can talk about a lot of like, you know, 30,000ft level initiatives and I don't want to if there was some takeaways, I would say the first and easiest step that you can take is go ahead and really understand you really understand what your value stream looks like.

00:30:28:29 - 00:30:47:08
Sneha Kumari: And when I say value stream, it includes everything, right? Interacting with your subs with your suppliers. How's your processes moving? What's the entire supply chain look like? Once you have the visibility, you will know what step. I think 50% of your job is done right there, because now you know where things are coming from and where things are going to.

00:30:47:09 - 00:31:12:02
Sneha Kumari: Once you know that there are enough resources that we can talk about, you can find online that you can take step by step to make impact. It's not all going to happen. We're not going to boil the ocean. You're not going to do all of this at all together. What we need to do is identify the what those, you know, bursts are in our value stream takes one small step at a time, which is doable.

00:31:12:02 - 00:31:29:28
Sneha Kumari: And it is. Trust me, there are companies doing it that you can take examples for. But in West Time and KPIs in it, I would say it's not a philosophy anymore. It's time that we take action and we have measurable metrics that we look at and keep keep up with it, as you know, as part of your leadership goals.

00:31:30:00 - 00:31:35:01
Sara Gutterman: Thank you. Thank you so much for being here with us and for sharing this incredible information.