Latest / PwC's accounting podcast / Replay - Talking ESG: Can AI be sustainable?

Transcript

Replay - Talking ESG: Can AI be sustainable? · PwC's accounting podcast

• 0:00
  SPEAKER_00: Thought Leadership from PWC's National Office.
• 0:07
  SPEAKER_02: Hello, and welcome back to Talking ESG.
• 0:10
  SPEAKER_02: I'm Heather Horn, and as always, thanks so much for joining us.
• 0:14
  SPEAKER_02: Today I'm especially excited because we're reissuing a great episode, which we originally recorded last fall during September's New York City Climate Week.
• 0:23
  SPEAKER_02: And the reason we're reissuing it is to highlight an exciting update.
• 0:27
  SPEAKER_02: My guest, Scott Likens, PWC's global chief AI engineer, and I recently had the opportunity to join The Atlantic and their podcast hosted by Nicholas Thompson.
• 0:39
  SPEAKER_02: We were on the other side of the mic as guests ourselves to continue the discussion of whether AI can be sustainable.
• 0:45
  SPEAKER_02: Check out that podcast, which is linked in the show notes.
• 0:52
  SPEAKER_02: In today's reissued episode, Scott and I explore AI's impact and sustainability and the costs and benefits of using AI to achieve sustainability objectives, including how the use of this technology may actually be able to, in some cases, reduce our carbon footprint.
• 1:09
  SPEAKER_02: We'll also discuss practical steps companies can take to integrate AI responsibly and effectively into their sustainability strategies.
• 1:17
  SPEAKER_02: And you don't want to miss what Scott's seen in practice from companies leveraging AI to help optimize sustainability in areas such as decarbonization and waste management.
• 1:27
  SPEAKER_02: So with that, here's my conversation with Scott.
• 1:30
  SPEAKER_02: Scott, welcome to the podcast.
• 1:32
  SPEAKER_02: So nice to have you on, and it's so nice to be recording while we're both here in New York for Climate Week.
• 1:37
  SPEAKER_02: And today we're actually combining two very hot topics, and that's the intersection of AI and sustainability.
• 1:44
  SPEAKER_02: And we're going to talk a little bit more about how AI can uh accelerate decarbonation and that's um or decarbonization, something I'm very interested in.
• 1:54
  SPEAKER_02: But maybe before we get into that, I think it might be helpful for the listeners to understand a little bit about your background and how you got into this topic.
• 2:02
  SPEAKER_01: Sure.
• 2:02
  SPEAKER_01: Uh going way back, I was actually going to be a pilot in college.
• 2:06
  SPEAKER_01: And to pay for flight school, I started programming and software engineering.
• 2:10
  SPEAKER_01: And it was on this thing called the browser in the internet, which dates me, which is uncomfortable.
• 2:16
  SPEAKER_01: But I just found the internet to be fascinating and uh jumped all in uh many, many years ago.
• 2:22
  SPEAKER_01: So for me, it was always about maybe chasing the next emerging technology.
• 2:27
  SPEAKER_01: And ironically, AI has been around a lot longer, and we can talk about that later.
• 2:30
  SPEAKER_01: But uh my journey through emerging tech is in you know undoubtedly based in in things like AI and digital technologies uh over the past many decades.
• 2:41
  SPEAKER_01: So really uh it was a curiosity.
• 2:43
  SPEAKER_01: It was just something I found super interesting, and now I fly a lot just in the back.
• 2:48
  SPEAKER_02: I was gonna ask, so no, no, you did not finish flight school, you did not become a pilot.
• 2:52
  SPEAKER_01: I did not.
• 2:52
  SPEAKER_01: I I did not.
• 2:53
  SPEAKER_01: About halfway through, I said I'm going all on computer science and engineering and mostly you know engineering translated aeronautical engineering to computer science.
• 3:01
  SPEAKER_01: There was there was good overlap.
• 3:02
  SPEAKER_01: So I didn't have to start over luckily to.
• 3:04
  SPEAKER_02: Exactly, exactly.
• 3:05
  SPEAKER_02: All right, well, very interesting background.
• 3:07
  SPEAKER_02: And so just let's jump straight into the topic.
• 3:09
  SPEAKER_02: And I think often when people um consider AI and the amount of power that's being used by AI, and there's been all the talk of data centers and everything else, and all the issues that are being caused by them, there's this immediate reaction that AI is actually going to add to our use of fossil fuels and otherwise.
• 3:29
  SPEAKER_02: But I think you are looking at it from the angle of where AI can actually be used to help from a decarbonization perspective.
• 3:35
  SPEAKER_02: So, what are some of the examples of how you're thinking about that?
• 3:39
  SPEAKER_01: Let's first start with the fact that we're talking about this.
• 3:42
  SPEAKER_01: You know, what some consider to be a new technology, but you know, seven decades of research and development.
• 3:47
  SPEAKER_01: You know, you have a bunch of scientists that are trying to create something new, and AI has been studied for decades.
• 3:53
  SPEAKER_01: But I think it's really exciting that we're we're starting the conversation early around, you know, the real social impact, the environmental impact, the energy impact.
• 4:01
  SPEAKER_01: But I think really it's being driven by the the last two years of just um amazing success when generative AI itself, a specific type of AI, came about.
• 4:13
  SPEAKER_01: It just was such a uh wild success that we realized maybe we didn't have enough compute to actually support the you know initial wave of consumers and now of enterprises.
• 4:24
  SPEAKER_01: So I think it's great that we're starting with the environment in mind and not just trying to go after the wild success of the next big technology.
• 4:34
  SPEAKER_01: So that's that's one thing.
• 4:35
  SPEAKER_01: But I think again, connecting to 70 or more years of research, AI could actually help us with some of the hardest problems that we see.
• 4:43
  SPEAKER_01: And I think the environment is one of the hardest problems we have to solve.
• 4:47
  SPEAKER_01: So I think the balance of innovation happening at a pace really that's it's hard to explain.
• 4:52
  SPEAKER_01: You know, and having seen decades of of innovation, this feels a bit different.
• 4:57
  SPEAKER_01: You know, it feels like the organic growth and innovation happening, but also the funding coming from the top, the the technology firms, the enterprises, they're all thinking about, you know, not only the innovation, but the impact.
• 5:10
  SPEAKER_01: And I'd say no conversation I have on AI doesn't talk first about responsible AI.
• 5:15
  SPEAKER_01: So how do we look at the risks and understand what's ahead of us?
• 5:19
  SPEAKER_01: But then second is what's the impact?
• 5:21
  SPEAKER_01: And that's probably not most technologies I've been through, the, you know, the hype cycles.
• 5:26
  SPEAKER_01: It wasn't about the impact.
• 5:27
  SPEAKER_01: It was about give us payback.
• 5:29
  SPEAKER_01: And I think it's great that we're starting there.
• 5:32
  SPEAKER_02: So let me ask two defining questions before we get uh into some more of the details.
• 5:37
  SPEAKER_02: So you mentioned generative AI.
• 5:39
  SPEAKER_02: Now, this is not going to be a podcast all about AI, but at least at a very high level, can you explain when you say generative AI what that means in case our listeners have not been that focused on this and some of the terminology?
• 5:52
  SPEAKER_01: Sure.
• 5:53
  SPEAKER_01: I'll try and oversimplify the topic artificial intelligence has a broad set of underlying technologies.
• 6:00
  SPEAKER_01: And there's a like machine learning, I'd say the past decade.
• 6:04
  SPEAKER_01: We've focused on machine learning.
• 6:05
  SPEAKER_01: So can we create algorithms that are learning either through human feedback or not?
• 6:11
  SPEAKER_01: And generative AI to me is a set of deep learning that actually generates content versus just making a decision or finding an answer and a massive amount of data.
• 6:22
  SPEAKER_01: It's actually transforming learning that it has, generative pre-trained transformer GPT.
• 6:29
  SPEAKER_01: There's generation on pre-trained data using the transformer.
• 6:34
  SPEAKER_01: So it's a very specific area.
• 6:35
  SPEAKER_01: But I think it's the one that made this very open and accessible to everyone.
• 6:40
  SPEAKER_01: Now we can actually use natural language to interact with the model versus a technical language, like a programming language.
• 6:47
  SPEAKER_01: So generative AI is able to generate, not just create a machine learning or an algorithmic or statistical analysis of data.
• 6:55
  SPEAKER_01: So it's a bit different.
• 6:56
  SPEAKER_02: And I think that's a perfect lead in then to my next question because you also referenced responsible AI.
• 7:01
  SPEAKER_02: And again, that's probably an entire podcast on its own.
• 7:04
  SPEAKER_02: Absolutely.
• 7:05
  SPEAKER_02: But at least to level set for this conversation today, what are some of the highlights when you think about responsible AI?
• 7:12
  SPEAKER_01: So, of course, we have our framework.
• 7:13
  SPEAKER_01: Uh, we've been thinking about this.
• 7:15
  SPEAKER_01: There's m much smarter people than myself that can talk in depth.
• 7:20
  SPEAKER_01: But the the the simple description I'd give is creating the guardrails around how and when and and what we use around AI, not just generative AI.
• 7:30
  SPEAKER_01: But what is our strategy and understanding around the regulation, around the implementation?
• 7:36
  SPEAKER_01: How do we understand bias and the models themselves?
• 7:40
  SPEAKER_01: So there's a level of understanding from the high level of our strategy and the regulation around it to the lowest level of the mathematics we use, the data we train on, and really structuring the guardrails around how we think this will be done responsibly.
• 7:53
  SPEAKER_01: And there's no one single answer, but putting those in place early and adapting them to the world that you're working in, whether you're a consumer or an enterprise, using it personally, making some decisions and building trust in by design versus adding it on at the end, which I think a lot of technology is added at the end, right?
• 8:11
  SPEAKER_01: Let's get the technology out there and then let's make it safe.
• 8:14
  SPEAKER_01: I think in AI, this this era of AI, we're talking about being responsible from the beginning because it's very impactful to humans.
• 8:21
  SPEAKER_01: I believe in an accretive way.
• 8:23
  SPEAKER_01: It's it's positive, it gives us a superpower.
• 8:25
  SPEAKER_01: But of course, there's there's negative ways, and people could use it for negative things.
• 8:30
  SPEAKER_01: But I uh putting a responsible AI framework in place, I think, is one of the most important things you have to start with.
• 8:34
  SPEAKER_02: All right.
• 8:35
  SPEAKER_02: Well, again, I think that's a topic for another podcast, but I wanted to at least level set a little bit before we got into it.
• 8:41
  SPEAKER_02: And then again, before we get into the benefits, which you give us some very high level and I want to dig deeper on sustainability, I do think there's a perception that AI and sustainability are sort of opposing forces, not forces that can be joined together for good.
• 8:59
  SPEAKER_02: So when people ask you about the negative impacts, what are some of the examples that you give and any sort of perspective on those?
• 9:09
  SPEAKER_01: So the last few years with the generative AI wave, what we fit found is that we have to train models at a scale we we really haven't had to do before.
• 9:18
  SPEAKER_01: So there's the first impact is the initial pre-training that's done is being done on a massive scale.
• 9:25
  SPEAKER_01: And without getting too technical, you have to have a specific type of processing unit, a GPU, to do this training, and you have to have tens of thousands of them running for days, weeks, or months, maybe longer.
• 9:38
  SPEAKER_01: And we continue to see that increase.
• 9:39
  SPEAKER_01: And with all of those chips, there's two big things that happen.
• 9:42
  SPEAKER_01: We have to power them, so the energy cost of them, we have to cool them.
• 9:47
  SPEAKER_01: So the water to cool them that then evaporates.
• 9:51
  SPEAKER_01: So the initial wave we started to study actually a few years ago to understand the true cost in carbon of not only the pre-training, but then the inference or the using of those models for an enterprise.
• 10:03
  SPEAKER_01: And there's actually, I think, some positive news is because the pre-training is done once and shared by many, it's it's not something we have to continue to do.
• 10:10
  SPEAKER_01: So while there's an enormous amount of effort to train them, everyone using them doesn't have to do that again.
• 10:16
  SPEAKER_01: So it's actually helping.
• 10:17
  SPEAKER_01: In the era of machine learning, we all had to train our machine learning algorithms individually to get the benefit.
• 10:23
  SPEAKER_01: With pre-training or the GPTs, we get to take advantage of that massive amount of training.
• 10:28
  SPEAKER_01: Now it's a lot of energy and water, there's no doubt.
• 10:30
  SPEAKER_01: But I think that's also triggering some other innovations, which I'm really excited about.
• 10:35
  SPEAKER_01: Just the last few weeks, there's been major announcements that data centers are being placed next to nuclear facilities, even ones that are no longer running that may come back online.
• 10:45
  SPEAKER_01: So the the reach into new ways to generate that power in a more effective and effectively green way, I think is exciting.
• 10:53
  SPEAKER_01: The chips themselves are being uh redesigned to be more effective, lower power.
• 10:58
  SPEAKER_01: We're looking at new ways to create chips in, you know, not just the GPU model.
• 11:04
  SPEAKER_01: So the innovation around it is exciting, but there's no doubt there's a huge impact on power.
• 11:08
  SPEAKER_01: And as this grows globally, what does that do to our overall power consumption?
• 11:12
  SPEAKER_01: Data centers are in the low single digits, and if that starts to rise, what does that mean?
• 11:18
  SPEAKER_01: And water.
• 11:18
  SPEAKER_01: Water is a tough one.
• 11:20
  SPEAKER_01: You know, that's not as easy to figure out.
• 11:22
  SPEAKER_02: Well, and even from a power perspective, I think if you dig into it, you're gonna see in a lot of cases where maybe other uh customers of that local utility are finding their power bills rising and otherwise.
• 11:35
  SPEAKER_02: So I I I hear what you're saying about putting it next to a nuclear plant or trying to hook it up to renewables or otherwise, but I don't think that changes the fact.
• 11:45
  SPEAKER_02: We're still generating more power.
• 11:47
  SPEAKER_02: And that means even if it's all wind, you have to create those wind turbines.
• 11:52
  SPEAKER_02: And, you know, there's you can read articles about the negative effects of that, or you have to create the solar panels, or you have all the nuclear effects.
• 12:00
  SPEAKER_02: You know, one of the reasons we stopped producing nuclear, and I am a supporter of nuclear, but what do you do with the nuclear waste?
• 12:06
  SPEAKER_02: That's right.
• 12:07
  SPEAKER_02: And so I think in a way, it's a little oversimplification to say that, okay, well, it's great because we only have to train these once, but then you still have these huge data centers that are continuing to be built.
• 12:20
  SPEAKER_02: And so I I just think that there is a balance here that you really need to think about.
• 12:26
  SPEAKER_01: I agree.
• 12:27
  SPEAKER_01: And and it's almost a a little bit of a circle.
• 12:30
  SPEAKER_01: Can AI actually help us optimize all of those things you mentioned?
• 12:33
  SPEAKER_01: How we create solar panels, AI is helping optimize in a better way.
• 12:38
  SPEAKER_01: How do we optimize the energy distribution, right?
• 12:41
  SPEAKER_01: Can we use AI to optimize that?
• 12:42
  SPEAKER_01: Because we have more powerful AI models that we can solve harder optimization problems.
• 12:47
  SPEAKER_01: And there's a whole nother pattern that we're looking at is quantum computing, which out without getting into quantum mechanics and you know, multiverse, uh, can we compute in a different way that is actually less power and being able to solve larger problems, which most of the things you mentioned around energy are very hard problems to solve.
• 13:05
  SPEAKER_01: So AI, as it gets more intelligent, you know, it's going to help us solve those problems better.
• 13:11
  SPEAKER_01: So I'm I'm hopeful, but I think the amount of innovation around energy, because of the need of AI for energy, has triggered innovations faster than we typically see.
• 13:22
  SPEAKER_02: Maybe, Scott, one last question on this topic.
• 13:25
  SPEAKER_02: Do you, from your experience and as you're working with clients or even as you're thinking about PwC, are you seeing that companies are focused on this part of AI and put aside sustainability and what AI can do for sustainability, but thinking of the impact of AI, or are you seeing more companies that are just racing towards we have to be the fastest, we have to be the best, you know, we have to solve these problems sooner.
• 13:52
  SPEAKER_02: So just curious how, in general, you're seeing companies think about that.
• 13:56
  SPEAKER_01: Yeah, as I mentioned, I think for the first time, it's on the front of the strategy and the design.
• 14:01
  SPEAKER_01: So as we think about our value uh flywheel, so as you start with a value hypothesis, as part of that, cost and carbon is in there.
• 14:08
  SPEAKER_01: And you maybe could categorize that as part of responsible AI.
• 14:11
  SPEAKER_01: Not only do we look at our strategy and things like bias, and we look at cost and carbon.
• 14:16
  SPEAKER_01: So at the end, end of the day, a return on investment, we have to understand the investment, which is energy and carbon.
• 14:23
  SPEAKER_01: So I'm seeing a lot of discussions.
• 14:25
  SPEAKER_01: We've built that into our models.
• 14:26
  SPEAKER_01: In fact, last year we studied as PwC just our own inference costs.
• 14:30
  SPEAKER_01: So the foundational costs plus the inference cost is look is what we're looking at from an AI perspective.
• 14:36
  SPEAKER_01: And we equated it to how many uh domestic flights did that look like, which we take a lot of.
• 14:40
  SPEAKER_01: But how many does it cut out?
• 14:42
  SPEAKER_01: Because I can use AI in a different way to do work.
• 14:45
  SPEAKER_01: Can I do less flights?
• 14:47
  SPEAKER_01: So it was actually um under 100.
• 14:48
  SPEAKER_01: I don't know the exact number, but in the 70s for us as a firm, which I I think is reasonable to understand that we can now act on.
• 14:55
  SPEAKER_01: Is that too many?
• 14:56
  SPEAKER_01: I don't know, but we can at least the starting point.
• 14:59
  SPEAKER_02: And sorry, with the inference cost.
• 15:01
  SPEAKER_01: So as I use that model that's pre-trained, there's more compute costs, the inferencing of the model itself, once it's pre-trained.
• 15:08
  SPEAKER_02: Got it.
• 15:09
  SPEAKER_02: I just would like to make sure if we're using terms on the podcast, we understand what they're doing.
• 15:12
  SPEAKER_02: Hopefully I used it right.
• 15:13
  SPEAKER_02: Yes.
• 15:14
  SPEAKER_02: Okay.
• 15:14
  SPEAKER_02: No, no, that's helpful, very helpful.
• 15:16
  SPEAKER_02: So then if you all right, so then if we take a step back and we think about some of the benefits and you started to go into these and how they can outweigh some of these costs, what are some of the key points that you would focus in on?
• 15:29
  SPEAKER_01: So I think once you have your strategy in place, the there's obviously the value uh hypothesis that you have to measure against.
• 15:36
  SPEAKER_01: But think about the amount of superpowers you give your people.
• 15:40
  SPEAKER_01: Can you do a lot of the tasks faster?
• 15:42
  SPEAKER_01: So efficiencies are great, but where can we fundamentally transform things in the business that effectively use a lot of energy outside of AI?
• 15:50
  SPEAKER_01: So optimizing our carbon footprint, let's start there.
• 15:53
  SPEAKER_01: Let's use AI to optimize our carbon footprint.
• 15:56
  SPEAKER_01: So we used an AI model to calculate our AI footprint because it's complicated.
• 16:01
  SPEAKER_01: There's where is the energy produced?
• 16:03
  SPEAKER_01: How is it transmitted?
• 16:04
  SPEAKER_01: Where is it then computed and inferenced?
• 16:07
  SPEAKER_01: It's a very complex uh model.
• 16:09
  SPEAKER_01: We actually did this for blockchain a few years ago because it had this similar stigma blockchain is bad.
• 16:14
  SPEAKER_01: But we actually proved it could help decarbonize.
• 16:16
  SPEAKER_01: And that's where we're looking at um AI helping us optimize our footprint.
• 16:21
  SPEAKER_01: But think about using AI optimization to optimize waste management as we're producing products, whether it's plastic or metals, um, balancing and enhancing energy decisions of where we place our factories based on where the energy is better for the world.
• 16:36
  SPEAKER_01: These are AI problems because there's a massive set of data that we have to look at.
• 16:40
  SPEAKER_01: We're doing another study of AI's true impact on climate, another really hard problem for us to solve.
• 16:45
  SPEAKER_01: Lots of data there.
• 16:46
  SPEAKER_01: And these are problems that humans could probably not solve right now with the traditional tools.
• 16:52
  SPEAKER_01: But with the advances in AI, the the superpower of AI, not only generative AI, but good old AI, uh, we're seeing these opportunities to really fundamentally transform industries, which affects energy.
• 17:04
  SPEAKER_01: So while the net effect of AI, the energy it uses itself, might go up, the overall energy impact may actually come down if we do this right.
• 17:13
  SPEAKER_01: Again, responsible AI.
• 17:14
  SPEAKER_01: We have to build this in from the beginning.
• 17:16
  SPEAKER_02: So you talked to, you gave a few examples there of optimizing energy usage, moving factories, and a few things.
• 17:23
  SPEAKER_02: So are there practical examples then that we have actually seen maybe someone decide to cite a factory in a different location or change its energy use pattern, or just again, I think sometimes practical examples can be helpful.
• 17:37
  SPEAKER_01: So when I thought of is you know, thinking about creating whatever consumer product without naming a client, um, taking small percentages of plastic out by optimizing the design of a product is a massive impact on the environment.
• 17:51
  SPEAKER_01: So the creation of that plastic, the forming of that plastic, the recycling of the plastic is all very energy intensive.
• 17:57
  SPEAKER_01: That's an easy one that impacts the entire world across many different types of products.
• 18:02
  SPEAKER_01: And we're seeing AI start to be able to optimize those designs in a way that we couldn't before because of the power that we have in the optimization.
• 18:10
  SPEAKER_01: And that's actually an example of using quantum optimization, which is able to find a better solution than classical AI today.
• 18:18
  SPEAKER_01: Um, that could apply to supply chains.
• 18:20
  SPEAKER_01: So moving people around, the traveling salesperson problem, that's one that we're studying intently because there's a big carbon footprint to moving people around the world.
• 18:29
  SPEAKER_01: And we believe things like quantum optimization and the benefits of AI there are showing true promise.
• 18:35
  SPEAKER_01: Taking, you know, 10% or you know, yes, let's say percentages out of a footprint is massive numbers of savings.
• 18:43
  SPEAKER_02: Well, and microplastics is something we've been talking a lot about, and something any uh company that's reporting under the CSRD in the EU is probably thinking about in some ways.
• 18:54
  SPEAKER_02: So it's a very interesting question.
• 18:56
  SPEAKER_02: Before we get into though the specific benefits of quantum computing, perhaps explaining exactly what you mean.
• 19:02
  SPEAKER_02: Um one question though.
• 19:04
  SPEAKER_02: When you talk about, for example, this impact it could have on a product in the supply chain and disposal and everything else, how many like companies are really thinking about this?
• 19:15
  SPEAKER_02: Is it a handful?
• 19:17
  SPEAKER_02: Is it many?
• 19:18
  SPEAKER_02: And again, I know it's from your own lens and what PwC is seeing, but again, some order of magnitude might be helpful to understand this.
• 19:25
  SPEAKER_01: I'd say generally optimization using AI is pretty pervasive, different by sector.
• 19:31
  SPEAKER_01: So there's certain sectors, I can think of pharmaceuticals that have been doing this for years, um, material sciences, you know, these are big data sets that have hard problems, and trying to find connections in these data sets is not a human capable task.
• 19:46
  SPEAKER_01: So optimization as a topic is something that I see pretty pervasively.
• 19:50
  SPEAKER_01: Now, the extent at which you use it, I think, is where it starts to you know differ.
• 19:54
  SPEAKER_01: Um, quantum optimization is not being used very much, but we believe it's it's a technology that is something you have to plan for.
• 20:01
  SPEAKER_01: There's new sets of skills around this type of AI that we have to build a workforce around.
• 20:06
  SPEAKER_01: So that's a longer-term solution.
• 20:08
  SPEAKER_01: But I'd say optimization in general is starting to take advantage of all of this new compute.
• 20:13
  SPEAKER_01: A lot of the issues before is we could understand the math, but we didn't have enough power to solve big enough problems to make the big enough impact.
• 20:20
  SPEAKER_01: But there were advantages.
• 20:21
  SPEAKER_01: So but you're starting to see that grow on the backs of the AI, Gen AI wave has been massive investments now in AI, good old AI.
• 20:29
  SPEAKER_01: So optimization, simulation, systems, dynamics, models, they're starting to flourish because now there's access to data, access to compute, and board-level uh investment in AI.
• 20:41
  SPEAKER_01: So it's it's you know, AI as an innovation has been brought along on the Gen AI wave, in my opinion.
• 20:46
  SPEAKER_02: All right.
• 20:46
  SPEAKER_02: Well, our main audience is finance and controllers.
• 20:50
  SPEAKER_02: And so for that part of the audience, we are going to get to later in the podcast exactly what their role in all of this is.
• 20:57
  SPEAKER_02: So I do want to do some more of this background and information first.
• 21:01
  SPEAKER_02: And I guess, Scott, as you think about then, you know, this idea of optimization and taking 10% out or otherwise, then are we really seeing that at least some companies are are taking advantage of this optimization and then measuring that impact.
• 21:19
  SPEAKER_02: So you would see that maybe translating into a sustainability report.
• 21:23
  SPEAKER_02: Or again, maybe they don't know their starting point well enough to actually see the benefits of this sort of happening at the same time that they're developing their sustainability reporting at that level.
• 21:34
  SPEAKER_01: Yeah, I think it's a combination of that, along with we have to remember we've only had Gen AI about two years.
• 21:40
  SPEAKER_00: Yes.
• 21:40
  SPEAKER_01: So we've you know sometimes forget we ex we expect why why is the return not here?
• 21:44
  SPEAKER_01: Well, it it's barely been two years.
• 21:47
  SPEAKER_01: Well, AI has again been around decades, but the the huge wave that all of the investment, the first year, the investment wasn't in budget.
• 21:54
  SPEAKER_01: The second year might have been there, but we're starting to see the the re the results.
• 21:59
  SPEAKER_01: But it's It takes time.
• 22:01
  SPEAKER_01: And quite honestly, it's a cultural shift.
• 22:03
  SPEAKER_01: So people have to work differently.
• 22:05
  SPEAKER_01: Uh they have to think about transforming an entire process.
• 22:08
  SPEAKER_01: That's really hard.
• 22:09
  SPEAKER_01: And that's going to take years to actually embed these technologies.
• 22:13
  SPEAKER_01: So the the the techies enjoy it because the investment has is been there and been increased.
• 22:18
  SPEAKER_01: But to actually trickle into the business does take time.
• 22:20
  SPEAKER_01: And to your earlier point, you have to understand where you're at to measure the benefits against.
• 22:24
  SPEAKER_01: And I think we're in this massively changing environment, both from a regulatory perspective, what do we report on?
• 22:31
  SPEAKER_01: How do we understand scope one, two, and three, which becomes more and more difficult in a complex supply chain?
• 22:37
  SPEAKER_01: And then how do we measure benefits?
• 22:39
  SPEAKER_01: So I'm not trying to dodge the question, but it's a hard question.
• 22:42
  SPEAKER_01: Maybe we need AI to help us.
• 22:44
  SPEAKER_02: Well, you told me that it could help with my podcast.
• 22:47
  SPEAKER_02: So I'm definitely interested in talking about that further.
• 22:50
  SPEAKER_02: So okay, you've mentioned quantum AI, quantum computing a few times.
• 22:55
  SPEAKER_02: And so again, maybe we can explain what we mean and what we think the role of that is.
• 23:04
  SPEAKER_01: I'll take a simple example today.
• 23:06
  SPEAKER_01: I mentioned the GPU, the graphics processing unit, the way we actually train these models uses a GPU, which is based on silicone and bits, zeros and ones.
• 23:17
  SPEAKER_01: In the world of quantum mechanics, um we have qubits, which could be zero, one, both, or neither all at the same time, because they have a superposition.
• 23:24
  SPEAKER_01: And again, go online.
• 23:26
  SPEAKER_02: I think you've already lost most of our audience.
• 23:29
  SPEAKER_01: Go to your your favorite AI chat client and ask about um superposition and quantum mechanics.
• 23:34
  SPEAKER_01: Yes.
• 23:34
  SPEAKER_01: You'll get a better explanation.
• 23:35
  SPEAKER_01: Say, explain it to me like I'm in high school.
• 23:37
  SPEAKER_01: Uh it'll do a very good job.
• 23:39
  SPEAKER_02: Maybe middle school.
• 23:40
  SPEAKER_01: But think of it as the ability to process all of the algorithms in parallel versus in sequential.
• 23:46
  SPEAKER_01: So machine learning processes um serially, and we add compute to make it go faster.
• 23:52
  SPEAKER_01: Quantum mechanics or quantum computing's intent, and again, this is still mostly theory, is that we could parallelize that and we could do it all at once.
• 24:00
  SPEAKER_01: And we've seen early tests that show us we can solve problems faster, and the solving is better.
• 24:08
  SPEAKER_01: So there's difference in the math, in the linear algebra we use in AI versus quantum uh computing AI.
• 24:14
  SPEAKER_01: The math changes because of that ability to process in parallel.
• 24:18
  SPEAKER_01: And again, I'm I'm not gonna do a deep explanation, but think of being able to go after as big of a data set as you want all at once, versus in a serial motion adding compute.
• 24:29
  SPEAKER_01: So it's gonna save us time, but it's also gonna find better solutions because the math could be different.
• 24:33
  SPEAKER_01: And we're starting to see some of those benefits already.
• 24:36
  SPEAKER_02: All right, so let's go back to our topic of the day.
• 24:38
  SPEAKER_02: We talked a little bit about some of this, the benefits of AI, and I think you've hit some of them, but I just want to make sure we say in one place.
• 24:46
  SPEAKER_02: Some of the ways that you're seeing AI companies use AI to move towards a net zero or a carbon neutral future.
• 24:55
  SPEAKER_01: Yeah.
• 24:56
  SPEAKER_01: So I mentioned a few waste management, so optimizing product design to minimize waste and plastic fuel production costs.
• 25:03
  SPEAKER_01: Transportation optimization is one we're really interested in.
• 25:06
  SPEAKER_01: Obviously, moving things around the world uses fuel, uses energy.
• 25:09
  SPEAKER_01: There, there's a lot of impact.
• 25:11
  SPEAKER_01: So, can we optimize better transportation networks, uh, interconnectivity, as you mentioned?
• 25:16
  SPEAKER_01: What does that scope two and three look like from a partner perspective?
• 25:20
  SPEAKER_01: Can AI help us simulate or model that?
• 25:22
  SPEAKER_01: Uh mentioned supply chain, uh, massive opportunities to collect more data on a supply chain and then use AI to optimize the supply chain.
• 25:29
  SPEAKER_01: Again, traditional, mostly good old AI versus Gen AI.
• 25:32
  SPEAKER_01: Yeah.
• 25:32
  SPEAKER_01: But Gen AI gives us the ability to maybe to talk to that data in a different way.
• 25:36
  SPEAKER_01: As a business user, can I find new scenarios or new ideas in a very complex supply chain?
• 25:41
  SPEAKER_01: Because the generative AI may find connections and make it easy for me as a business user.
• 25:46
  SPEAKER_01: Um smart grid, I mentioned energy.
• 25:49
  SPEAKER_01: I think there's a lot of opportunity to use AI to redesign our grids, uh, how we produce, transport, and use energy and among other technologies to help stabilize the grid, but no doubt investments needed there.
• 26:02
  SPEAKER_01: And then just broadly integrating this into our financial systems.
• 26:04
  SPEAKER_01: As you mentioned, you'll talk later about finance.
• 26:07
  SPEAKER_01: There's so much data we probably already have, but it's hard to get to because we need a technical interface.
• 26:12
  SPEAKER_01: Is there ways that AI can simplify that interface?
• 26:14
  SPEAKER_01: So as a business user, I can actually understand more scenarios more easily through natural language.
• 26:20
  SPEAKER_01: So simplifying the data infrastructure that supports what we do in finance, uh in operations, in some of the core businesses that support the business, I think there's a lot of opportunity.
• 26:31
  SPEAKER_01: And some of the early opportunities we've seen have been in those areas.
• 26:34
  SPEAKER_02: Well, and Scott, it's interesting.
• 26:36
  SPEAKER_02: Most of those items you listed, the thing again that came to mind to me is that it's reducing carbon, but that also is having other sustainability positive impacts.
• 26:48
  SPEAKER_02: So less waste, less pollution, you know, less people, movement.
• 26:54
  SPEAKER_02: You know, there's a lot of different things that potentially could be helpful.
• 26:58
  SPEAKER_02: And so as you're talking to companies, are they mostly focused on carbon or are they looking more broadly of, oh, this can actually have a lot of broader benefits?
• 27:09
  SPEAKER_01: Well, I'd actually even flip it to say if you get some of these things right in optimizing, there's new revenue opportunities.
• 27:14
  SPEAKER_01: You can outpace your competitor by getting to a market faster or to more markets.
• 27:19
  SPEAKER_01: Now, can I generate, even in a regulated industry, a product launch in new markets in local language with you know local imagery and understanding faster?
• 27:30
  SPEAKER_01: So the revenue opportunity, so the efficiencies get us to transformation.
• 27:35
  SPEAKER_01: I don't like to focus just on the efficiencies.
• 27:37
  SPEAKER_01: Now, to your point, there's also other benefits to the environment, to our brand, and to thinking about being a trusted AI user versus an AI user.
• 27:46
  SPEAKER_01: Yeah.
• 27:47
  SPEAKER_01: We're looking at uh some research called Three Tomorrows.
• 27:49
  SPEAKER_01: It's it's really gonna, you know, there's different ways we can really approach AI.
• 27:53
  SPEAKER_01: And is that trusted way, which we believe in at PwC, the best?
• 27:56
  SPEAKER_01: I think it is, starting with trust, building our brand around it.
• 27:59
  SPEAKER_01: But it the the new opportunities are the exciting part.
• 28:01
  SPEAKER_01: So that supports business, but it also supports our sustainability goals.
• 28:05
  SPEAKER_01: So that's why we've built it right in the value wheel.
• 28:07
  SPEAKER_01: Cost and carbon is a key part, but there's also how do we solution around it, the patterns we solution and the value hypothesis that we start with.
• 28:14
  SPEAKER_02: And then if I'm a company, for example, that's focused highly on biodiversity, because that's obviously another huge issue from sustainability.
• 28:22
  SPEAKER_02: You mentioned water earlier, reducing water usage.
• 28:26
  SPEAKER_02: So as I'm designing or thinking about how I want to use AI, if I as a company have these priorities.
• 28:34
  SPEAKER_02: So it's carbon reduction, minimizing negative biodiversity impact, maybe actually having positive biodiversity impact, minimizing water usage.
• 28:42
  SPEAKER_02: You know, there can be multiple different drivers of what they're trying to achieve.
• 28:48
  SPEAKER_02: Are those all things then that can be considered in these AI models?
• 28:52
  SPEAKER_02: Or do you normally sort of say, here's the main thing I'm trying to achieve?
• 28:56
  SPEAKER_02: And then the rest maybe is more peripheral?
• 28:58
  SPEAKER_01: Well, maybe a good project because you know the initial wave was about energy and before carbon.
• 29:03
  SPEAKER_01: Yes.
• 29:03
  SPEAKER_01: But the then quickly after that was water, more from a constraint perspective, to be honest.
• 29:08
  SPEAKER_01: So the question is, do we make that more of a positive?
• 29:11
  SPEAKER_01: Can we give back around water?
• 29:13
  SPEAKER_01: Can we you know be net positive there?
• 29:15
  SPEAKER_01: So haven't seen as much of that in the sense of planning for it, other other than from a constraint perspective.
• 29:20
  SPEAKER_01: But maybe that's something we need to work in.
• 29:22
  SPEAKER_01: It's it's such an important issue.
• 29:24
  SPEAKER_02: Right, right.
• 29:24
  SPEAKER_02: And you do definitely see companies that are very focused, but I think for so many companies, there's too many problems to focus on.
• 29:32
  SPEAKER_02: And so that's where, again, when you were talking about this power of AI, well, if you're trying to focus on a lot of problems, maybe this is the same thing.
• 29:37
  SPEAKER_01: And which ones can you measure?
• 29:38
  SPEAKER_01: Yes.
• 29:39
  SPEAKER_01: Right.
• 29:39
  SPEAKER_01: Exactly.
• 29:40
  SPEAKER_01: The water one's a little bit harder to measure.
• 29:41
  SPEAKER_01: Weather is a little bit difficult.
• 29:43
  SPEAKER_01: Yeah.
• 29:44
  SPEAKER_01: Maybe AI will actually help us there.
• 29:45
  SPEAKER_01: We're starting to see some indications we can model climate and weather better and better.
• 29:50
  SPEAKER_01: But that's a hard one to measure and predict versus energy.
• 29:53
  SPEAKER_01: We know how to measure and predict and understand.
• 29:56
  SPEAKER_01: So it might be about the measurement versus the intention.
• 29:59
  SPEAKER_02: You mentioned extreme weather.
• 30:01
  SPEAKER_02: So is that an area where we are starting to see that people are trying to use that from a prediction or to consider or how are they using it to think about these various climate factors?
• 30:13
  SPEAKER_01: I think climate modeling is one that is absolutely being invested in.
• 30:18
  SPEAKER_01: We're doing some really deep climate modeling as it relates to AI right now.
• 30:22
  SPEAKER_01: But again, probably one of the hardest problems to solve because of the massive amount of data sets.
• 30:26
  SPEAKER_01: But I'm hopeful with new breakthroughs in technology around AI, even how we compute AI, that we'll get better.
• 30:33
  SPEAKER_01: And we're starting to see some indication.
• 30:35
  SPEAKER_01: But I think climate and weather prediction in general is one of the harder problems.
• 30:39
  unknown: Yes.
• 30:41
  SPEAKER_02: I don't even have a response to that because yes, it clearly, if you can predict the weather, maybe earthquakes would be another thing from California.
• 30:48
  SPEAKER_02: So that would be very helpful if you could.
• 30:51
  SPEAKER_02: So if you know, I said we bring this around to sort of the controller finance perspective.
• 30:57
  SPEAKER_02: So if I am in a finance role and I either see my company is dabbling in AI and they're not thinking about it from a, you know, with this considering potential sustainability impacts, or even if they are, what role do you think that they have in their company's use of technology?
• 31:18
  SPEAKER_01: Well, first off, as I mentioned, if you don't have a responsible AI framework, which would inherently have some of these things in there, I would promote that as the first step.
• 31:26
  SPEAKER_01: And I have that discussion with every leader is start with some sort of guardrails.
• 31:31
  SPEAKER_01: And hopefully part of that includes the impact on the environment.
• 31:35
  SPEAKER_01: And to some of the discussion, that that could be defined differently based on the industry you're in and some of your strategic goals, but there has to be something in there at a minimum, the cost of the carbon, the energy and carbon.
• 31:45
  SPEAKER_01: AI is not free.
• 31:46
  SPEAKER_01: As much as I love it.
• 31:47
  SPEAKER_01: Right now it's not free.
• 31:49
  SPEAKER_01: Hopefully one day it will be.
• 31:50
  SPEAKER_01: But building that into some framework and then looking at the true cost of the AI solutions and modeling the benefit, you know, in as part of the strategy, truly understanding the new transformational opportunities, the new markets will make it to the speed to which we get to market, the quality of our products, that has to be measured against the investment in the AI that supports that.
• 32:11
  SPEAKER_01: So I think coming up with some formula or some pattern or some process that fits in that responsible AI framework.
• 32:16
  SPEAKER_01: And we're always happy to share ours.
• 32:18
  SPEAKER_01: It's it's open and free.
• 32:19
  SPEAKER_01: So I'm not selling it, but thinking about all the aspects of what AI does solve and what it doesn't solve and truly understanding the cost of that versus the benefit that you're seeing.
• 32:28
  SPEAKER_01: So I think that squarely fits in the finance and controllers and executives that are looking at the operations of the business.
• 32:34
  SPEAKER_01: And then I'd say a personal uh challenge is how are you you going to use it to give you these new superpowers?
• 32:41
  SPEAKER_01: How are you gonna look at data differently so that you can get at faster decision making, bigger strategies, find new areas to transform?
• 32:49
  SPEAKER_01: It's really applicable to any function.
• 32:51
  SPEAKER_01: And I think starting in the core of the business is a great place to start.
• 32:55
  SPEAKER_01: So a little bit of a challenge, where are you gonna use it for a superpower?
• 32:58
  SPEAKER_02: Well, so actually then I have a question for you.
• 33:00
  SPEAKER_02: And we've talked about this a little bit in the past, and when we talked about AI in the podcast, but I feel like for some people, it seems like, and I know you've said this multiple times, this is not the case, but that in a way it came out of nowhere that no one was talking about AI, and then suddenly everyone was talking about AI.
• 33:17
  SPEAKER_02: And so if I'm an individual that hasn't really been focused on it, and now I just think, well, I'm behind, you know, someone else is gonna deal with this problem.
• 33:28
  SPEAKER_02: How do you respond to those people then?
• 33:31
  SPEAKER_01: No, it's absolutely what we hear all the time.
• 33:36
  SPEAKER_01: But again, AI is a broad topic.
• 33:38
  SPEAKER_01: Generative AI, the transformer, which is actually five or six years old now, what that means is now any human can interact with AI.
• 33:46
  SPEAKER_01: You don't have to understand or program AI, but you can leverage and use the power of AI, again, done responsibly.
• 33:54
  SPEAKER_01: So the overnight success is now it's easily human accessible to any human, any human that could interact in in some sort of a language or um natural way.
• 34:05
  SPEAKER_01: So that means the power of AI is for everyone, not the programming of AI.
• 34:09
  SPEAKER_01: So you don't actually have to understand the mathematics.
• 34:11
  SPEAKER_01: You should have a team that does and puts the guardrails in place.
• 34:14
  SPEAKER_01: But for you, think of it as this intelligent digital friend that understands all of your data, likely understands almost any topic at a high level that you can interrogate and talk to and learn from.
• 34:27
  SPEAKER_01: Instead of thinking of it as a tool I use once, right?
• 34:29
  SPEAKER_01: And I ask one thing and I get an answer.
• 34:32
  SPEAKER_01: Use it as a companion that has a lot of intelligence that can help you make a judgment or help you reason better or help you learn a topic, not as a one-stop shop for all answers.
• 34:44
  SPEAKER_01: Because that's not how you should use it, at least right now.
• 34:47
  SPEAKER_02: Well, and so again, if I'm a normal corporate finance person and my company in some ways using AI, but they're not responsible for these data centers and the data sets and all those types of things.
• 35:01
  SPEAKER_02: If I come back to the concerns about the energy usage and some of those other items, is that something you would say sort of an average user of AI needs to be focused on, or is that a problem for someone else?
• 35:16
  SPEAKER_01: I would say that's much more of a large-scale problem.
• 35:19
  SPEAKER_01: So the individual user, the actual again, inference of them using a model, asking a question or interrogating, that's not that's not likely the issue until it gets to scale.
• 35:30
  SPEAKER_01: Now, if you're using that 12 hours a day with millions of prompts and tokens, it will start to add up.
• 35:36
  SPEAKER_01: But also the offset of what aren't you having to do that the AI is helping you do faster and transformed into a different process.
• 35:44
  SPEAKER_01: So that you have to look at both sides of it, not just the the cost of the AI.
• 35:47
  SPEAKER_01: But I think right now the average user that's not the issue we're we're looking at.
• 35:51
  SPEAKER_01: It's more of a scaled enterprise.
• 35:53
  SPEAKER_02: Yeah.
• 35:53
  SPEAKER_01: So we've rolled it out globally to 100 countries.
• 35:56
  SPEAKER_01: We have to look at that.
• 35:57
  SPEAKER_01: We have to measure it.
• 35:57
  SPEAKER_01: We have to understand the impact.
• 36:00
  SPEAKER_01: Um but in the early days, it the impact was very small.
• 36:03
  SPEAKER_02: Yeah.
• 36:03
  SPEAKER_02: So again, uh if I'm the average user, I can actually focus more then on how can AI help my company from a sustainability perspective.
• 36:13
  SPEAKER_02: And you know, you gave some good examples there.
• 36:15
  SPEAKER_02: I know this is very industry-specific where companies should start and otherwise, but if you were to take a step back and say where you can find it to be most beneficial or helpful from a sustainability perspective, what would your advice be there?
• 36:30
  SPEAKER_01: Maybe I'll approximate back to the originals, the work we did for blockchain was to understand truly the full cost of the financial system and then the fine the true cost of the blockchain and the connections it created.
• 36:43
  SPEAKER_01: I'd say the same thing for AI.
• 36:45
  SPEAKER_01: What is the process you use today?
• 36:47
  SPEAKER_01: Whether you're working in a spreadsheet or in other collaboration tools, you're using energy to some degree.
• 36:53
  SPEAKER_01: Can I shrink that energy by using AI to do something faster?
• 36:56
  SPEAKER_01: Again, the base efficiencies are very small and you you can't focus just on those.
• 37:02
  SPEAKER_01: Looking at an entire process, where can I take steps out?
• 37:05
  SPEAKER_01: Which again, they're using energy, whether I call it AI or not, they're using energy.
• 37:10
  SPEAKER_01: I want to take entire steps out.
• 37:12
  SPEAKER_01: I want to take the whole process the team works on and change it.
• 37:16
  SPEAKER_01: And I think about travel, the big energy uh users, those are those are easy ones to look at and say, where can I just do this differently?
• 37:24
  SPEAKER_01: The world of the video conference has saved business travel, although I'm not sure anymore.
• 37:28
  SPEAKER_00: Yeah.
• 37:28
  SPEAKER_01: It might have come back.
• 37:29
  SPEAKER_00: Yeah.
• 37:29
  SPEAKER_01: So how do we actually instantiate a process that changes forever in a good way?
• 37:34
  SPEAKER_01: Those are going to be really individual to your function, your sector, your industry.
• 37:38
  SPEAKER_01: Um, but there's massive, there's opportunities everywhere.
• 37:41
  SPEAKER_02: And I guess this goes back to them.
• 37:42
  SPEAKER_02: We were talking about product redesign, supply chain redesign.
• 37:46
  SPEAKER_01: Reinvigorating marketing, going you know, to new markets, product design, the quality of our products, the speed at which we deliver, every aspect, no matter what you're building.
• 37:56
  SPEAKER_02: Yeah.
• 37:56
  SPEAKER_02: And again, if I keep this focus on sustainability, it's almost like if I look at the processes in my company that have the potential to be the most negative from a sustainability impact, whether that's because you know there's a potential that there's child labor in the factories in my supply chain, or my supply chain itself is using too much dirty fuel, or if because there's too much plastic or whatever else, then I could say, okay, that's what I'm going to tackle first and start to think about how AI can be used in that context.
• 38:28
  SPEAKER_01: Absolutely.
• 38:29
  SPEAKER_01: Yeah.
• 38:29
  SPEAKER_02: So if we kind of try to bring this all together again, thinking about a company that is not focused on AI in this way, what advice do you have?
• 38:39
  SPEAKER_01: I always start with the advice get started.
• 38:41
  SPEAKER_01: Waiting to see what happens is typically not a good strategy, especially now in the innovation uh that we're seeing, the pace we're seeing.
• 38:49
  SPEAKER_01: And you don't have to go all in or make big investments, but I'd say you personally should try these tools to see how they can help you be more efficient.
• 38:57
  SPEAKER_01: I do it all the time.
• 38:58
  SPEAKER_01: Just in my personal life and my professional life, how do I I become more efficient?
• 39:02
  SPEAKER_01: Which, if everyone does that, has a net impact to an organization or an enterprise.
• 39:06
  SPEAKER_01: So trying something, waiting to see, I think is a bad strategy.
• 39:10
  SPEAKER_01: Starting with responsible use and thinking personally, how do I want to do this responsibly?
• 39:15
  SPEAKER_01: And integrating that into an organization, I think is really key.
• 39:18
  SPEAKER_01: Trust in the design of how we use this is always the second start start.
• 39:22
  SPEAKER_01: So experiment yourself.
• 39:23
  SPEAKER_01: Then talk as an organization, how do we do this responsibly?
• 39:26
  SPEAKER_01: And then you can dip your toe in.
• 39:28
  SPEAKER_01: You don't have to make big investments.
• 39:30
  SPEAKER_01: But to your point, find those big areas, those hardest problems, and start to chip away at them with a really super intelligent partner, accelerator, superpower that you never had before.
• 39:40
  SPEAKER_01: You know, leverage it.
• 39:41
  SPEAKER_01: And you'd I think you'd be surprised at how creative the humans will be once they're enabled with those tools.
• 39:46
  SPEAKER_02: All right.
• 39:47
  SPEAKER_02: Well, that is definitely a good note to end on.
• 39:49
  SPEAKER_02: So, Scott, such a pleasure to uh to talk to you and thanks so much for joining me today.
• 39:53
  SPEAKER_01: Thank you so much for having me.
• 39:55
  SPEAKER_02: That's our show for today.
• 39:57
  SPEAKER_02: Tune in next week for more fresh episodes so that you never miss any of our audio content.
• 40:03
  SPEAKER_02: Follow the PWC Accounting Podcast wherever you listen to your podcasts.
• 40:08
  SPEAKER_02: And to stay up to date on all our latest accounting and reporting news, sign up for our newsletter at viewpoint.pwc.com.
• 40:17
  SPEAKER_02: From Thought Leadership at PwC, I'm Heather Horn.
• 40:22
  SPEAKER_02: Thanks for tuning in.
• 40:29
  SPEAKER_00: This podcast is brought to you by PwC, All Rights Reserved.
• 40:33
  SPEAKER_00: PWC refers to the U.S.
• 40:34
  SPEAKER_00: member firm or one of its subsidiaries or affiliates, and they sometimes refer to the PWC network.
• 40:40
  SPEAKER_00: Each member firm is a separate legal entity.
• 40:43
  SPEAKER_00: Please see www.pwc.com slash structure for further details.
• 40:49
  SPEAKER_00: This podcast is for general information purposes only and should not be used as a substitute for consultation with professional advisors.