Latest / Technology Now / How can you balance AI and sustainability?
Transcript
- Michael Bird Hello and welcome back to Technology Now, a weekly show from Hewlett Packard Enterprise where we take what's happening in the world and explore how it's changing the way organizations are using technology. We are hosts, Michael Bird. Aubrey Lovell and Aubrey Lovell. And this week we are diving into one of our regular topics, sustainability and artificial intelligence. We'll be asking how AI can be used more sustainably, we'll be exploring how AI can make other processes more sustainable, and we will be discussing how waste products from AI processes could be utilized in the future. Michael Bird Yes, that is right Aubrey. if you're the kind of person who needs to know why what's going on in the world matters to your organisation, then this podcast is for you. And if you haven't yet done so, do make sure you subscribe on your podcast app of choice so you don't miss out. Right, let's get into it. Aubrey Lovell AI and sustainability have a very complicated relationship. On the surface level, AI is run on servers which are built from computer chips and wires, which in turn need rare earth elements and metals to build them. Not to mention, every step of the supply chain produces emissions and waste. The 2024 Digital Economy Report from the United Nations revealed that a single two kilogram computer requires 800 kilograms of raw materials to make. Michael Bird That is a lot of kilograms to make a computer. Aubrey Lovell It is very much so. And not only that, but the UN Environment Program predicts that within two years, AI infrastructure around the world could consume up to six times more water than Denmark. That's six times more water than an entire country, Michael. And we will link to both of those statistics in the show notes. Michael Bird My goodness. Michael Bird But Aubrey, that is only half the story because while AI consumes huge quantities of resources to create and use, it's also able to find new and innovative ways to reduce resource consumption in not only its own production, but other processes as well. AI has brought a lot of good into the world and balancing up the good and the bad is not just incredibly important, but also vital for those who use AI on a regular basis. Sustainable AI is quickly becoming a topic of global interest. So how can AI be created and run in a sustainable way? And how can AI improve sustainability in other sectors? Aubrey Lovell Well, back in January, Michael, you went to Davos. It was very cold and you talked to some of those at the World Economic Forum. And I know we've already kind of played a few of those conversations on previous episodes of the podcast. However, we have one left and I think that this interview highlights one of the most exciting things about the technology sector, which is the speed of progress. And some of the hypothetical solutions discussed just a couple of months ago are already being put in practice as we've seen. Aubrey Lovell Well, back in January, Michael, you went to Davos. Michael Bird I did, and it was cold. Aubrey Lovell It was very cold and you talked to some of those at the World Economic Forum. And I know we've already kind of played a few of those conversations on previous episodes of the podcast. However, we have one left and I think that this interview highlights one of the most exciting things about the technology sector, which is the speed of progress. And some of the hypothetical solutions discussed just a couple of months ago are already being put in practice as we've seen. But We aren't going to give the game away just yet on what those could be. If you're a long time listener, right? Keep your ears peeled for a familiar topic during Michael's chat with our guest this week, Monica Batchelder, Chief Sustainability Officer for HPE. Michael, let's take a listen. Michael Bird Monica, thank you so much for joining us. Can you just please quickly explain what a Chief Sustainability Officer actually does? Monica Batchelder Yeah, it's a good question because it's sort of an evolving role. So at HP, my job is to really guide the company towards one sustainable business strategy that helps us reduce the impact of our products and our value chain, to apply our technology to make the world a better place, and also to make sure that we're capturing the market opportunity of our products and solutions. Michael Bird So why is sustainability so important to you and HPE? Monica Batchelder Yeah, so first of all, it's a moral imperative. But more importantly, it's important to our customers and to our stakeholders. So for our customers, increasingly they're getting these climate targets set by their corporations. They're also trying to grow their IT at the same time. And sometimes those things can be in contrast. So really our customers are looking for more sustainable solutions that will allow them to grow sustainably and meet those goals at the same time. Michael Bird And this sort of board imperative for sustainability, that's like a fairly recent development. Monica Batchelder Yeah, it's been growing over time, the investor community has really sort of accelerated this need to report and understand the data and the impacts of what's actually happening. And then we're using those impacts to actually make an impact in what we're doing. Michael Bird So while you're here at Davos, you're hosting a session at the AI house titled, Can We Resolve the Paradox of Sustainable AI? Can you just quickly explain what that's all about? Monica Batchelder Yeah, so we think about this in two ways. There is sustainable AI, which is essentially how do we create the most efficient, sustainable and ethical solutions possible? And then there's AI for sustainability. How do we apply those same solutions to create a positive impact and transform industries? What the paradox is though, is the trade-off between those two things. So how do we understand the true costs of those environmental and social impacts? And then how do we understand whether the social benefits or the environmental benefits downstream, the application outweighs the costs? Michael Bird So it's sort of a balancing act. OK, OK. So clearly, AI is super power hungry. So it's the point that AI can both make the climate crisis worse, but also potentially solve some of the problems. Exactly. Right. Monica Batchelder Exactly. We need to understand when we're using AI and make sure that we're doing it in a really targeted and responsible way so that we're not creating negative environmental impacts or dehumanizing people in the process. Yeah. So it's dependent on the development and the deployment of the solution itself. But if you think about it, compute power equals electricity demand, equals costs and energy, right? And emissions. To go with that. Michael Bird Got it, got it. So just how carbon intensive are AI deployments? mean, is it really based on where the energy is coming from? So can you give me some stats on the sort of power usage that we're talking about here? Yeah. Monica Batchelder Yeah, so IEA, the International Energy Agency, they're projecting that data center energy consumption could double between 2022 and 2026. That's equal to over a thousand terawatt hours of energy. That's equivalent to the entire country of Japan's electricity consumption today. Michael Bird My goodness, okay, where's all the energy gonna come from? Monica Batchelder That's the question, right? An AI can help solve that problem, but also it's going to be a consumer of that energy. Michael Bird Okay. So what are some of the opportunities that we are seeing today from a sustainability perspective? Monica Batchelder Yeah, so AI is actually an opportunity to rethink the actual energy infrastructure as well as the IT infrastructure. So we've been designing and deploying AI in different technologies the same way for a long time. So how can we rethink the underlying infrastructure, move to do new technologies like silicon photonics, but also then how do we apply those very efficient technologies to transform industries? So just for an example, if you think about it, carbon emissions from AI could be 2 to 4 % of global greenhouse gas emissions. But what if you could apply that same technology to solve the other 96 %? Michael Bird Yeah, OK. Monica Batchelder So that's where the opportunities are. And we're not just talking about the big breakthroughs that were waiting to come, like nuclear fusion or fission. There's actually things happening right now today where we can apply AI to find new efficiencies, find new clean energy sources, and advance sustainability Michael Bird So what's the solution here? How do we solve this AI paradox? Is it less AI, maybe more AI, but in a more sustainable way? Monica Batchelder Yeah, good question. So I think it's a multifaceted answer. One, we need to be more mindful about when we deploy AI and how we develop and deploy it. So AI is not the answer to all of our problems. And I've heard the analogy that using AI for simple applications is the same as driving a race car to the grocery store. It's not as efficient, right? So we need to be mindful about when... Michael Bird Yeah, yeah, yeah. So we need to be mindful about when... You might look cool doing it. Monica Batchelder Maybe it's fancy. Maybe you get a more efficient answer, right? But it doesn't get us to the outcome that we're looking for. The second thing we need to think about is efficiency. I always hear this question start with, we need more energy. We're running out of energy. What's the supply? What's the availability? We first need to optimize the actual infrastructure and the solutions and systems running the AI. And then lastly, we look at energy source. So we need to be powering this with clean energy. We have an energy crisis globally where we need to be sourcing more energy to power AI, but we have to be doing that in a way that prioritizes clean and sustainable energy. Michael Bird We already sort of talked about the sort of energy usage needed to power AI. Where does liquid cooling fit into this? Monica Batchelder So for a long time, we were cooling our data centers and our IT infrastructure with air cooling. And that was fine. But fewer and fewer GPUs today can kind of work in air-cooled environments. So we need to be working on direct liquid cooling, is what really the reason that we are incorporating into our supercomputing portfolio, right? So the way that direct liquid cooling was explained to me that I actually think is helpful to sort of understand it is think about trying to cool down a can of soda. And you can either put it on the counter, and you can turn up the air conditioning in the room, or you can put it in a cooler. For a long time, we've been doing the former, and direct liquid cooling is essentially equivalent to the latter. Michael Bird So it's more efficient, but actually it's vital to be able to solve the problem of running all these data centers using more energy. Monica Batchelder It's more efficient, it uses less energy, and actually a lot less water as well, which we haven't talked about. Michael Bird and you can do something with the heat that's being generated from these liquid cooled components. Monica Batchelder Yeah, so heat is actually the biggest byproduct of data centers. So if you think about it, the heat that's coming out the back, you think about a refrigerator, there's always heat coming off the back of that, right? That's similar to the data center. And what we can actually do is we can repurpose that heat. You could plug it into like a local district cooling. You can use it to heat a local swimming pool. And actually, heat is expensive. So it's actually a source of financial value if you can find an end user to sell it to. Aubrey Lovell Thanks so much, Monica. That was really a great conversation. And I think it's so important to look at the complexities and nuances of any situation, right? Especially one like AI, because let's be real for a moment. AI isn't going anywhere. You see it everywhere. And the cat is already out of the bag. So it's really good to know that we have people dedicated to investigating how something like AI can be more sustainable in the future. Aubrey Lovell Okay, now it's time for Today I Learned, the part of the show where we take a look at something happening in the world that we think you should know about. And I think you've got something special for us this week, Michael. Michael Bird Yes, I do. And actually, Aubrey, I'll do something slightly different. I'm going to ask a question. What sort of music do you like? Aubrey Lovell ooooh. my goodness, this is a good one. And it's also a loaded question, right? Because I feel like I love all music. I listen to basically everything. I'd say, know, indie, pop, rock, you know, all the good stuff. New wave, what about you? Michael Bird I think I'm probably in about the same camp. You play guitar and you're a singer. I play drums. I mean, I know... why we haven't started a band yet. Aubrey Lovell I know, we need an HPE band. I feel like we've talked about this over the years, but we have to pitch it. Michael Bird We totally do! think we totally do! Anyway, the reason I'm asking is because a recent study has shown that our enjoyment of music is partially inherited. The study was conducted on both identical and fraternal, that's non-identical, Swedish twins, so the two groups could be compared. Now, if identical twins enjoyed the same types of music more than their fraternal counterparts, then genetic makeup could be considered to have had an effect on the outcome. The study found that genetics do have a significant influence on people's enjoyment of music. Up to 54 % of what they called the music reward sensitivity, basically how much you enjoy and get happiness out of the music, was inherited. Another study of Norwegian twins found a similar result. Now the outcome of this study adds another layer of knowledge to the classic nature versus nurture discussions, but the authors do acknowledge that both this and the Norwegian study were conducted on Northern Europeans so may not be applicable to everyone. Studies like this may not have an immediate impact on most of us, but understanding why seemingly random patterns of sound can cause us to experience such strong emotions is an important step to understanding how we as humans experience the world, especially with the importance of music in socialising and building connections with others. Oh, I 100 % agree. I don't know about you, one of my favourite things to do is just to play music. I get so much enjoyment out of that. Aubrey Lovell definitely. This is really fascinating because if I actually think about my parents, right, my dad was really into music, also played guitar, so did my grandfather. So inherently, know, nature versus nurture, I think they were more inclined to have that creative musical side. But then when you're seeing it in practice and you grow up with it, I think it's just easier, it's more accessible to you, right? What about you, Michael? Michael Bird Yeah, I mean, I have four elder sisters and my house was basically filled with musicals. We had musicals on vinyl and I think I was brought up on musicals. So there are many musicals where if I hear the first like three chords, I'll just immediately sing the rest of it. I'm the only drummer in my family, but all of my family, my dad played guitar, my parents both sing. So I think I sort of inherited the musical. I think there is bit of a musical gene, isn't there? Aubrey Lovell the musical gene. Definitely, really interesting story. Thanks so much for that. Aubrey Lovell All right, well, now it's time to return to Michael's conversation with HPE's Chief Sustainability Officer Monica Batchelder recorded at the World Economic Forum in Davos. Michael Bird So what other ways can AI become more sustainable and more ethical? Monica Batchelder Yeah, so I think that we're sort of at an inflection point where AI can actually be a catalyst instead of a risk to our electricity grid. And what I mean by that is the amount of energy that we're going to have to add to the grid to power these AI systems. The tech companies that are driving that are the same companies that have the most ambitious net zero goals and have the funding and the technology to transform our grid. So really this is an opportunity to sort of do a system wide revamp of our electricity systems and at the same time expand AI and meet our business goals. Michael Bird And so from an ethical standpoint, there's a lot of conversation around supply chain, ethical supply chains. Monica Batchelder Yeah, so first we need to be developing and deploying AI solutions in an ethical way and making sure we're not dehumanizing people. But also if you think about it, the materials that go into AI infrastructure today often are mined upstream by, you know, maybe 12 levels down our supply chain, but that is real human impacts. And so if we can move away from those materials, gold, copper, that sort of thing, to silicon photonics, to new different types of technologies, we're actually going to be able to reduce some of the upstream impacts, not just to the environment but also to humans. Michael Bird So is this where the global data partnership against forced labour comes in? Monica Batchelder Exactly. So there's an estimated 28 million people living in forced labor today. And there is really not much data, especially in the electronics industry, about those impacts. And so the aim of the partnership is really to advance new approaches to data measurement and accountability and data-driven action to eradicate forced labor. Michael Bird So is this something that us as individuals should be thinking about and taking responsibility for or is it something that we should just leave to our organizations? Monica Batchelder Even large enterprises aren't really going to see the true costs of AI, right? You're going to see that in the hyperscalers and the people that are actually building the models. But I do think it's important that we're mindful of how we use AI and everyone needs to understand that there are real physical impacts every time they use a tool of AI. Michael Bird Yeah. So is that the sort of concept of like you don't always need to plug a question into your, you know, AI engine. You can just search on the internet. Monica Batchelder Exactly. So one generative AI inquiry uses 15 times the amount of resources as just putting it into a search engine. So that's why we need to be mindful. Of course, that's small on the scale of what we're talking about here. But if you add that up over the scale of an enterprise, that can actually have a much larger impact. Michael Bird Yeah, and I think as a user you sort of don't get a sense of what the impact is that you sort of type this question into a box and it sort of just goes into the ether and comes back with an answer. You don't really know the environmental, the energy, the any impact of it do you? Monica Batchelder Yeah, and it's not just for the consumer. It's for the technologists building the technology as well, right? We understand how to design out a dollar of cost We don't understand how to design out a ton of carbon and that's one of our biggest barriers to making more sustainable AI So will we maybe start to see organisations not just listing the cost of things, but maybe also the environmental, the carbon cost of things. Maybe, I don't know, if you're paying for a service of a product, you'll actually see what the environmental impact is. Monica Batchelder Yeah, exactly. So we refer to it sort of like nutrition labels, right? You're going to see the calories, you're going to see the carbon impacts, and our customers are actually already requiring this from So even at point of sale at the RFP, they want product carbon footprints for us. And so we're actually an industry leader right now in some new tooling that we've developed to give that insights to our customers. Michael Bird So do you think we'll maybe start to see customers making buying decisions based on environmental impact rather than on sort of pure cost? Monica Batchelder I don't think we know yet whether customers are willing to pay a premium, but we think that when all other things are equal, when cost is equal, they may even trade performance for improved environmental impacts in some instances. Michael Bird Because it's something that as organizations, like it's important to them or they're being targeted on or they're having to report back to the board on. Monica Batchelder Yeah, I mean it's not just about being sustainable, they're actually hitting space, power, cooling constraints, and so they need more energy efficient products if they're going to continue to scale. Monica Batchelder Well, let's look five to 10 years in the future. How do you think AI models will look from a sustainability perspective? Monica Batchelder I think it's hard to predict. think one is the growth of AI is skyrocketing right now, particularly in the training, and that's going to level off. And then there's also going to be changes to the underlying infrastructure. But I think we can assume two things. One is AI is going to become more efficient because it has to be for us to be able to scale. Secondly, I think we're going to see more targeted applications of the AI. So instead of needing the world's information to solve something for a specific sector, we're going to have very targeted direct models that we can use with a lower environmental impact and a lower cost. Michael Bird Fantastic. Monica, thank you so much for your time. It's been a pleasure speaking to you. Monica Batchelder Thanks, Michael. Aubrey Lovell Again, thanks so much, Monica, for taking the time to talk to us. the idea of having nutrition-style labels but for carbon footprints is really kind of a fascinating idea. And I think it really showcases how much consumers care about these things, right? mean, change really is being driven from the top down and the bottom up. And if you want to find more of the topics discussed in today's episode, make sure to check out the show notes. And you can also find an entire episode about how heat from data centers can be repurposed. If you go to our podcast channel, just look for the title, How to Harness a Great Untapped Energy Source, where you will also see Michael in a bathtub. Michael Bird Right, we are getting towards the end of the show, which means it is time for this week in history, look at monumental events in the world of business and technology, which has changed our lives. Now Aubrey, you gave us the clue last week. Can you just remind us of the clue? I'm pretty sure it was something to do with taking a photo or something, but you and I were both stumped. So just remind us, please. Aubrey Lovell Yes, so last week's clue was it's 2019 and scientists have answered the question, how do you take a photo of something which is completely black? Michael Bird Yeah, no, still stumped. Aubrey Lovell And I was also stumped, I didn't have a clue, right? Well, turns out this would be the release of the first ever photo of a black hole. Michael Bird see what they did there. Aubrey Lovell Nice. Well, after years of planning and unbelievably large quantities of data needing to be processed, researchers from the Event Horizon Telescope revealed the first ever photo of a black hole. I actually totally remember this on the news. Now, I know what you're thinking. How do I take a photo of something that even light can't even escape? And the trick is to look at everything around the black hole as that is what is giving out the light. And as it gorges itself on stars and gas and dust, the area around the black hole gets so hot, it starts to glow. And this is what we can take a photo of. I'm getting very interstellar vibes here. So here's the really cool thing though. The Event Horizon Telescope uses a nifty technique called very long baseline interferometry, which sounds like a mouthful, but in reality can be boiled down to this. The scientists take a few telescopes based around the world and combine that data to create a virtual telescope the size of our planet. Isn't that cool? Michael Bird Oooooh, that is very cool. Aubrey Lovell Very cool. Now the first photo they took shows the supermassive black hole at the center of a galaxy called M87, which is 55 million light years from Earth. And that means this photo shows the black hole at the same time as the first horses, rhinos, and cattle that were starting to appear. I can't, like my brain cannot comprehend that, but that's pretty cool. Michael Bird Aubrey, our producers love putting stuff in the show that blows our mind and this is one of them. I feel incredibly small in the universe after all of that. Aubrey Lovell I feel very exhausted, you know, after reading that, but it's actually really amazing and pretty cool how our whole universe works. Michael Bird Yeah. Very, very out of this world. Right, Aubrey, I have next week's And the clue is, it's 1983. Ring, ring. You'll never guess where in the house I am. These clues are getting harder. Aubrey Lovell They really are, but I love a good challenge. I mean, obviously I feel like it has to do something with like cellular or, you know, telephones, something to do with that, but who knows. Michael Bird Oh, oh, like the wireless phones in a house. Not cellular, but you know like there's, yeah, yeah, they're all like, all the... Aubrey Lovell some sort of advancement in the telephone technology. Michael Bird Yep, that's what it is, yep. I'm gonna go with that. Aubrey Lovell Well, that brings us to the end of Technology Now for this week. Thank you to our guest, Monica Batchelder, and of course to you, our listeners, thank you so much for joining us. If you enjoyed today's episode, do let us know, drop us a review or a comment whenever you'd like to listen. Michael Bird Technology Now is hosted by Aubrey Lovell and myself, Michael Bird. And this episode was produced by Harry Lampert and Izzy Clarke with production support from Alysha Kempson-Taylor, Becky Bird, Alyssa Mitry and Natasha Naik. Aubrey Lovell Our social editorial team is Rebecca Wissinger, Judy-Anne Goldman and Jacqueline Green and our social media designers are Alejandro Garcia, and Ambar Maldonado. Michael Bird Technology Now is a Fresh Air Production for Hewlett Packard Enterprise and we'll see you at the same time, the same place next week. Cheers. Aubrey Lovell Cheers.