When AI Meets Atoms with Windscape AI

Join us for today's podcast where our host Jake Wombwell-Povey interviews the two co-founders of Windscape AI, Eric Thompson (CEO) and Jason Yosinski (CTO).

Windscape combines low-cost pressure sensors and artificial intelligence to provide cleaner and more accurate data to optimise wind turbine operations. Though selling this technology to utilities is challenging due to the long, costly, and bureaucratic process, they have found early adopters and are working towards permeating the industry. 

The co-founders also express their views on artificial intelligence, machine learning, and their potential uses in sustainability and climate change. They identify some overhyped trends and discuss their preferred policy reform of a price on carbon.

Contact Jake
Linkedin
Substack

Contact Windscape.ai
Linkedin
Website

00:01 Introduction to the Podcast
01:01 Guest Introduction: Co-founders of Scape AI
02:29 Origins of Windscape and the Founders' Journey
08:54 Understanding the Technology Behind Windscape
09:04 The Role of AI and Machine Learning in Windscape
09:13 The Impact of Technological Advancements on Windscape's Existence
11:12 How Winscape Improves Efficiency for Wind Turbine Operators
15:55 The Distinction Between AI and Machine Learning
21:56 Challenges and Lessons from Selling Novel Technologies in the Wind Industry
24:14 Understanding Wind Farm Ownership and Operations
24:52 Building Relationships with Asset Managers
27:19 Challenges and Barriers in Wind Farm Operations
29:42 Navigating Stakeholder Mapping and Incentivization
31:33 Finding Advocates within Organizations
33:28 Targeting Early Adopters in the Wind Industry
34:07 The Impact of Windscape on Wind Farm Output
37:46 Potential Applications of AI in Power Generation
45:33 Investing in Sustainability and Climate Crisis Solutions
48:04 Enacting Policy Reforms for Climate Crisis
48:54 Contacting and Learning More About Windscape

If you enjoyed this episode please subscribe to our channel and leave a review. 

Transcript

0:01

Hello, and welcome to the seed. Our intention with this podcast is to illuminate the climate tech venture space, sharing the stories, the insights of the founders, pioneers, innovators, and investors who are shaping this space at this critical juncture. Through the stories we'll share, we'll provide insights, knowledge, as well as solace, reassurance and hopefully community to those who are pushing the boundaries in the climate tech and sustainability space who are working so hard to make the world a better place by building or investing in the very companies that are going to make the change we need to see. So if you enjoy what you hear, please do share this episode with a friend or leave us a review on whichever platform you're on so that others can also find our work. I'd love to hear from you, so please drop me a message at Jake at 7gen. vc if you want to continue the conversation and with that, thank you for joining us and enjoy the show. Hello, and welcome to the seed. Today I'm joined by the two co-founders of Scape AI CEO, Eric Thompson and CTO, Jason Budzinski. So Eric has been the co-founder of C-Suite member of multiple Cleantech Ventures, including Ensemble Energy, the utility wind, utility scale, wind turbine manufacture, Nordic wind Power, where they rose from Coastal Ventures at Goldman Sachs, as well as Nettel. Energy. Dr. Jason Yasinski is the CTO at Winscape and this has been confirmed by other investors as well as one of the leaders in the machine learning and AI space where he holds four patents and has written over 50 papers and has co founded three companies beside Winscape AI. He led the deep learning research team at Uber, where he helped to grow the team from 14 to 125 members. And his work on AI has been featured in NPR, Fast Company, The Economist, Telex, and on the BBC. So Jason and Erica are co founders of Winscape. Winscape uses AI and machine learning to rapidly process localized weather data using Wi Fi mesh sensor networks to map actual and local wind changes. so that wind farm operators can improve the performance, the ROI and the turbine life of their turbines. So gentlemen, it's great to have you both here. Welcome to the show. Great to be here. Thanks. Thank you. So, gents, rather than, because we've got both of you on, rather than going through both of your stories individually, which obviously I've just summarized and I hope I did a pretty decent job of, it'd be great to learn where Winscape came from and how you guys met. In developing and co developing that, that vision. Okay. Eric, you want to start maybe? Sure. Well, Windscape was the the technology itself, was the brainchild of Dr. Richard Ely. Someone I've known in the wind space for 15 years or more. We got to know each other during the Nordic wind power startup days. He has been a... Innovator in a lot of areas in wind and also in hydro in the hydroelectric space. And he came up with this idea 10 years ago that pressure data out across the landscape could be used to create a prediction of wind that would reach a wind farm. And that concept. Was looked at quite a bit by him and by a small team. They created a patent sometime ago. And over those years, he and I consulted about the idea, discussed the. Viability. And it was clear that only until recently the technology was not there. The pressure sensors, the machine learning and so on was not there. But in the last couple of years those technologies have converged and made this possible. I left Ensembl Energy where I was doing AI related work, but not in the same, same vertical in the wind industry. And started talking with him about commercializing this, jumped in after a lot of due diligence with a lot of other people in the industry. And then began looking for a co founder that had the expertise in machine learning that was clear, was necessary to make this happen. I was very happy to, get to know Jason through that process. Yeah, I can do a bit of an intro on my side too. I guess I've been doing ML for a long time for a little over a decade since starting my PhD in 2010. So I spent 10, kind of 10 solid years just pushing forward the fundamental science of neural networks. So understanding how neural networks work, figuring out how to make them work better when they kind of fail to train, understanding how they represent information internally. We have some fun papers on figuring out how to like hack neural networks, how to make them say kind of arbitrarily bad, bad stuff. It was a fun decade. We made a ton of progress. The field as a whole made a ton of progress. The end of that decade, maybe 2020 or so to round round things off neural networks had become kind of a mainstay of industry, like a tool that we can all rely on, but they're now used inside of. you know, near the every company for many, many purposes. I had a personal realization that I wanted to work on something a little bit more applied. So kind of like transitioning a lot of my science to more working on product, more engineering and got really interested in climate change. So I kind of listened to a lot of podcasts podcasts like this one, read a lot of books, met a lot of people, a friend and I interviewed about 100, 150 people in different industries to kind of learn. What are people working on? What are people excited about? What changes do we need to make as a, you know, the human, the human species? to address climate change. And I sort of became slowly convinced that making more green energy was one of the biggest wins we had available to us right now. So I decided to try to find kind of an intersection between AI and machine learning and making more energy in particular from wind, solar, battery storage. The nuclear is also important. And so I started looking for the right co founder, which was someone with like deep experience in one of these verticals. And I met Eric, we got along really well and decided to work on this kind of crazy idea called Winscape together, where we're bringing together the latest machine learning and the latest and most pressing needs of wind operators. Well, thanks for, thanks for filling that in for us. And for me, Winscape. Seems to be the quintessential example of a company that is bringing together software so that the existing hardware that we've already got, and I, and I think a startup found the other day was there's about 500 trillion worth of heavy assets. Property, plant, equipment in the world. And if you think about that asset base, you could argue that we don't necessarily need more assets. And there's obviously some very strong exceptions to that, especially in clean energy development. But if you think about the power of software to make that more efficient, Windscape appears to be a really powerful example of how you guys can take. and develop innovative software to make existing infrastructure, in this case, wind turbines, more efficient, prove the process efficiency, the output, the ROI, the turbine life. Yeah, definitely. That's something that's really exciting about this. We don't have to go out and invent anything. New on the wind turbine. We don't have to install anything new on the wind farm. We just take the existing assets and some innovation that is in Jason's department as the machine learning brains here and create more renewable energy just through through application of that innovation and new technologies. The way I think of it, though, Jake is we, we People in the U. S., people around the world. We do need a lot of new assets, right? We do need a lot of new wind assets, a lot of new solar assets. If we're gonna hit eventually like net zero net, being net carbon neutral as a species, like some numbers I've seen are we need to roughly 3x the size of the electrical grid. And roughly 6x the number of, the amount of wind produced, also drastically increased solar and other means of producing clean energy. So like, it's, I think it's a both, right? It's, we need more physical infrastructure, literally more new steel transmission cables spanning the country and the world, a lot of new assets on that grid. And what we're working on, right, is trying to make the grid smarter, more efficient. That way we can bring down the cost of producing green energy so we can incentivize investors and wind developers to jump on board. If we can make it more profitable for them, that's how we're going to accelerate the rollout of this physical. Expensive, high complex physical infrastructure. Would you guys want to take a step back and just explain how Windscape works? How it makes those wind turbines more efficient? How it makes it more profitable for them to operate? And the key, what I'd love to understand a bit more about as well is, why is AI and machine learning so critical here? I want to phrase that in another way. Why are you able to exist today when, say, Jason, at the beginning of some of your academic work, this just wasn't even feasible? Should I take a stab, Eric? Or you wanna, you wanna try? Yeah, no, please, go ahead. Yeah, so you asked a few questions, Jake. Basically, what do we do to make... Life more profitable for wind turbine owners. So wind turbines right now work by adjusting constantly to current wind conditions to optimize production. So when they're pointed, let's say, straight north, and the wind swings 15 degrees to the west, then they slowly track it 15 degrees to the west, back to the east, back to the north, whatever. A lot of people know that turbines kind of shift where they're pointed to optimize production. What fewer people know maybe is that each individual blade as it turns around also adjusts its pitch to capture as much energy as possible. The pitch you can imagine like you're on a sailboat and you're kind of sailing and you're in trim so the sail is perfectly trimmed into the sailboat and you're making great progress. And then if you're racing, say, and the wind shifts, you need to retrim that sail and readjust it to be either more aggressive or less aggressive to capture maximum energy from the wind. So wind turbines do the same thing. This all works today. They're already generating energy in this way today. They're doing so using data, though. That's a little bit late. So they only measure the wind after it already kind of passes the turbine, hits the back of the turbine. And it's also noisier than it has to be. So the anemometers in the back of turbines are right behind the blades. So they're in the wake of the blade. So every time the blade goes around, it generates a little vortex, which hits that anemometer and kind of makes the data noisy and, and even biased in certain ways because it's rotating in a specific direction. Yeah. This is all a long explanation. What we're doing is producing a cleaner data source. to help turbines adjust their operation. And also in addition to being cleaner, we can deploy hundreds of meters in some cases upwind of turbines. So we, we feel the air as it comes in before it gets there. Right. Pause there. We also asked another question, but does that make sense? Absolutely. It makes sense. And then the follow up question to that was, how does that help make life better for wind turbine operators, utilities? Yeah. So first off, filling, filling in something that all three of us know, cause we've talked about this, but some of the listeners may not know. And when, when Jason's saying we can deploy upwind, what we're deploying is a whole network of very low cost. Pressure sensors that are mounted on fences or fence posts about that height off the ground, dozens of these very low cost communicating through a wireless mesh network back to our, our system, and we're picking up these signals and some other signals from from the surrounding environment and, and communicating that back to this System that Jason is now going to elaborate on as the is the machine learning part. Yes. Jake, you also asked why is it, why is it sort of possible today? And Eric alluded to the history of the project, how it was, it was an idea a long time ago, but it was an idea that wouldn't have worked using the level of technology we had 10 years ago. There's two, there's two reasons it's possible today. One, which is something everyone sort of knows about is machine learning didn't used to work as an approach. And now it. It does work well. It's reliable. There's a lot of cloud infrastructure available to do ML. There's a huge pool of talent to hire from to, to work on, to scale at machine learning companies. We need machine learning in our, in our case, because mapping between kind of the feel that a bunch of pressure sensors are feeling across the landscape to eventually predict a single point of wind, which is not at any of those sensors, but is removed is a hundred meters above the landscape. That's a complicated physical system. If you try to model it directly using like a simulator, for example, it would just be intractable. Especially if you don't even know the shape of the hills exactly, right? Where your sensors are in specific locations behind a specific tree that buffets it in a certain way, right? This is a complicated function that we need to use machine learning to, to learn. The second reason is the sensors are just way, way better these days. So here's, here's one of our sensors. It's available on Adafruit for under 10. It's a Bosch presser sensor. The actual sensor itself is like not even this whole board. It's this little tiny silver thing in the middle, like a millimeter square. Roughly what a lot of people don't know is three or so years ago, these were not that accurate. So a plus or minus 12 Pascals, which is one meter of elevation change. So you think if you go up in an elevator and your ears pop after a few floors, that's how sensitive your ears are. These would be sensitive to within one meter then. About a year and a half ago, they came out with one that was plus or minus three Pascals. So that's like this much elevation change. And more recent ones for other companies are even like sub Pascal accurate. So like you can detect an elevation change this small just by the difference in air pressure. So we're taking advantage of all this recent improvement in the sort of MEMS hardware, which is all being driven, not by wind, not by us, but by people's. You know, manufacturers putting smarter sensors in mobile phones and smartwatches. We're just taking advantage of that kind of tide of improvement to the physical sensors. Yeah. Yep. Okay. Understood. So there's a combination of hardware and software that wasn't available before that is available now. Yeah, to make this happen. And even when, even when we started looking at this a couple of years ago, just all of the machine learning off the shelf tools and companies providing databasing services and you know, open source. It was just in this this very nation stage of people figuring it out. A lot of things breaking up, not necessarily working. And these days, I'm just amazed at the pace of development that Jason and his team can make using these now reliable tools. And Jason, you and your team, you're focused here on the application of AI here. And the words you've used to me before are kind of, you know, software meeting atoms, which I think is a really nice kind of sound bite to exemplify the work that you're doing here. But AI is a word that's used a hell of a lot. If you're an expert, you obviously know what you're talking about, but I would say the vast majority of us probably don't. And whereas before, let's say two years ago, you had chatbots that came back to you with predetermined answers. Nowadays, all of that, all of that work, which wasn't previously called AI is now called AI. So I think there's, I suppose what I'm saying is there's a lot of hype. The word is used pretty loosely. And so. You have been written, you know, a number of papers in this space. How do you think about the application of AI here? Why is it so important? And, you know, specifically, what is it here? Where's the intelligence here that makes Winscape so powerful? Yeah lots of, lots of unpack there. You're certainly right that there's a lot of AI hype these days. AI and machine learning are definitely very much kind of like suddenly in the press, whereas they, they didn't used to be. A lot of that press is, I think, well deserved, and I think a lot of the hype and the conversation around how machine learning and AI will change our society is, is justified. I think it's, it's clear that these technologies will change our society like a ton in the next couple decades. Nobody really knows how, even the people that work directly on machine learning, the people that work directly in the intersection of ML and policy. No one really knows how it's going to change things. Maybe it'll replace some jobs, but that's kind of like a vague statement. It doesn't convey necessarily the concrete way this will meet the road for us. Like I said, our core product, which you can think of as this new sensor that combines. Cheap pressure sensors and machine learning to aggregate information. Our product is possible thanks to machine learning because we need to learn that kind of physical system, learn to invert that physical system which wouldn't have been possible a few years ago. I'm not sure if that's where you wanted to go with, with that or if you're kind of getting at something else. I think you're muted, Jake. I think people use it as this blanket statement for what was previously kind of automation. And I think they often, they often convolute and automate a piece of software. You know, simply a digitalized workflow with the actual intelligence or even the, the ability for that system to learn itself by that machine learning that intelligence that grows over time and interaction and activity rather than something that's just kind of dynamic. So it's, that's why I just wanted to drill in there as to what is it specifically that makes this. Artificially intelligent. I see. I see. Yeah. So people don't often or don't always make a distinction between AI and machine learning. Those of us in the field do think of them as slightly different. So we would say artificially artificial intelligence is a system that's designed to be or to appear to be intelligent in some way. So for example, you mentioned automation. So let's say you have an automated set of scripts that you wrote and it reads your email and it filters it into different little boxes. And to an external observer, it seems very smart. But you wrote all the rules yourself. So we would call that, that could be an example of AI, but it wouldn't be machine learning unless that system itself is like learning and improving over time. So machine learning is a smaller set of algorithms that through experiencing the world, through experiencing data, change their behavior over time to get better and better. So our system would be an example of a machine learning system because as we gather more data on pressure readings and wind measurements, our system is constantly training and improving. So we would say that's a machine learning system. These days, all the impressive progress in AI is actually progress in machine learning. So in particular, like. large language models, large models that generate images. They're all learning from existing images or existing text, in many cases, script from the Internet. So that's all machine learning and AI, although there's a subtle distinction. Yeah, understood. Okay, thank you for setting that out there. So, and, and, Jake, I think there's another element of this that I've been learning about more as I've been exposed to different AI experts. And of course, most intensively Jason over the last few years. And that's that there are definitely different approaches in machine learning to to an individual problem. And then within that also, there's also an approach of Engineered factors that are included in, in the machine learning where the machine learning starts to get constraints. Or aids given to it by the modeling team that reflect the real world physical, physical systems. And I think that's a, from, as an engineer, I think that's a really fascinating merger of physics engineering and the statistical and programming methods that are at the core of machine learning. And maybe Jason could elaborate for a second on how that's potentially helping Winscape in what we're doing. Yeah, for sure. I think that's, Eric, you bring up a good, a good general point, which is most Most machine learning systems in the middle somewhere of the stack is a core often. It's a model It's a often times a neural network model or in some cases other models and this model takes inputs produces outputs and internally changes its structure when it experiences data when you train it sometimes we think of this as the whole system, but in practice, there's a lot of A lot of the stack that's before that model and a lot of the stack that's after the model before the model, you have to think about where are you getting your data from? Is it clean? Are your features, are your data points available? Where you need them when you need them, like is the whole system streaming in the data in time to get to the model in time to make a prediction when it goes to the model, then you take your prediction and you have to do something with it, right? You can't just produce right answers and somehow have that be valuable to the world, right? It's you have to get it to the customer. You have to put it in a form that they can. Use in the case of communicating with turbans the turbans need this data in a particular format They need particular guarantees like this output is not going to swing so wildly that it makes our control system unstable, right? That's a very important feature of this output which has little to do with machine learning and a lot to do with engineering the entire system appropriately. So just to say there's a big stack around the machine learning core model that a lot of companies end up spending, you know, far more than half of their time on it, probably in that, in that book too. Okay. Thanks for illuminating that. That's really helpful. The other huge part of what you guys do, obviously you have this, you have the software element. They're machine learning, but you guys are clearly experts in the wind space. That's a huge and growing industry. Wind turbine projects themselves are huge undertakings, but selling to the utilities, selling to wind turbine operators, selling into those projects, they are vast undertakings. They take years, and it's not known as being the most agile of industry. And that's not just picking on wind that's picking on any utility scale power generation project. So I'm curious how you guys are finding not only selling into this. This industry, but selling something which is so new and novel and therefore potentially less understood, you know, with more and more reticence and hesitation around adopting it, how are you finding that what are the lessons that you are learning and what are some of the lessons that other founders might be able to pay heed to as they try and sell novel technologies into. Large established capital intensive industries, you know, yeah, that's that's a really good question. Of course, very relevant for us on a, on a, on a daily basis. Windscape is primarily or in our, in our early efforts selling more to the owner operators of wind farms as opposed to, and luckily, as opposed to the electric utilities. And you know, as a, as the briefest overview, you know, the, the, the, a simple view of the wind industry is that there are large and small organizations that go out and they find sites, they develop wind farms, and they either operate them over years, or they develop them and sell them to someone who's an owner operator, but in either case, they become an owner operator, and they're, they're selling that electricity To the electric utilities and sometimes to private off takers. And so the people that care at the first level about improving, increasing the the ROI of the wind farm are those owner operators. And those are some of those are very large organizations like Nextera um, EDF Renewables, Avon Grid et cetera Engie, Pattern Energy and some of them are small, small organizations that only own a few wind farms. But, you know, there's a 1000 wind farms across the U. S. and so there's a great. Variability in the ownership and operation structures. Those are very known entities. So, Windscape in the BizDev activity can, can easily database and catalog those entities. And then it's a, and it's a matter of building relationships over time with the people who are responsible for improving the operations. Asset asset managers in general. So our experience to date has been very positive. We we through the network we have, we connect with these asset managers and we described to them. An opportunity to increase operations with a very low investment and. Because we're in the startup stage, what we are doing is creating a, a pilot program with them deploying sensors around their, their wind farm, getting the data from that, getting some data from them and starting to demonstrate the value proposition and working with them to integrate with their operations. And Most of these organizations especially the ones that we get good traction with have a mandate to apply some of their efforts to looking at new innovative technologies. The wind industry knows that there's a lot of innovation that's possible. They have a mandate. There's a group probably that is their innovation group or some similar name. And so they, they often become our point of contact. And we are. We're then in the process of, of creating a proof of concept, proof of value in that, in that organization, I can fill in a lot more detail, but that's that's a little bit. So I imagine like most innovative companies, most startups, you can demonstrate there's a clear use case. For your technology and your innovation. And I imagine like most startups, you can also give us ample stories and anecdotes of where, no matter how viable, how useful the solution is, it just isn't getting adopted. You can see the kind of efficacy of this, even, dare I say it, the customer can see the efficacy, but they just can't press the button or they're just not pressing it quickly enough because I'm sure for you guys, you would just like to see this rolled out. You'd like to see the benefit, the impact rolled out to as many wind farms as possible across the United States, across the world. So what, what are the, some of the barriers that you guys are encountering and how do you overcome some of those? Yeah, the biggest barriers are around the. The different wind farm operations and maintenance and ownership structures that are out there. So a wind farm starts as a, as a developer, they'll get financing for the project from a number of different entities and then they'll build the project and buy the turbines. So what we want to do is we want to help those turbines improve their operations, but as soon as that project's purchased, those turbines are under warranty with a, a very big and cautious and slow moving OEM, you know, Festus, GE, Wynn, Siemens, Kamesa. And so it's very, but so they're, they're under warranty. Went from owner operator is saying, Oh, I want to increase operations. The, the OEM is saying, Oh, well, we, we promised a particular level of performance we're meeting that it ain't broke. Don't fix it. And we're saying, well, we can help improve that. So Winscape has the strategy of completely. Circumventing that and initially working with older wind farms where at least that warranty period is passed. And so then we're working with, with one less entity, but then there's still the the, the wind farm owner and maybe they contracted out for operations to independent operations and maintenance organization, which is a very, economically efficient way for them to operate. But then there's still sort of two entities and how do we make sure that What we're offering aligns with all of their incentives. So that's, then it's important to get to know what the the motivations are and the incentives are and the leeway is that say that operations and maintenance organization has to make investments, what their budgets are what, what sort of changes they have made, what sort of technologies they've adopted to improve the operation of that, that wind farm in the past. And. Align what we're presenting with what is easy for them to talk about and work with in the organization. So you're painting a picture of a typical enterprise kind of stakeholder mapping and navigation challenge. And also alluding to the fact that often the incentives aren't quite as clear or direct. for the various stakeholders within that decision making process as they might first appear to be at first at first pass i. e his windscape we make stuff go faster and make you more money great but who cares because i don't and that guy doesn't even though someone might so i mean how do you guys think about That incentivization challenge, because whether it's wind farms or any other software application, that's trying to target the climate impact the positively impact the climate crisis in any capital intensive industry. I imagine there's exactly the same version of this problem, which they are encountering. Yeah, certainly jump in at any point here, Jason. I, I, I look at it as those organizations are, are made up of individuals. So I, I think there's a lot of business school speak that, that often tries to, to categorize these organizations as big monolithic entities that only operate a certain way. But I found throughout, throughout my career you know, focusing on Bringing new technologies to the renewables industry that what's important is finding those individuals that get excited about something and those individuals, they work within the organization they've spent decade or decades of their careers, learning that organization and how to work within it. So you find an advocate. Within that organization. And sometimes that's a C level person and they say to someone that works for them, it's your job to work with Winscape to explore how to make this happen. And, and sometimes it's someone at mid level who knows how to navigate, navigate the organization and get things done. One, one organization in particular we're working with. Probably shouldn't give, give names, but we we connected because it's someone I've networked with over the years with the CEO, and he got kind of excited about it. And he said to the person who is their head of meteorology and, you know, windscape is in essence. Predictive meteorology. So this person would really understand what we're doing. He said to his head of meteorology, Let's work with Windscape and, and explore some opportunities. And that person then, you know, had the mandate, knew the organization very well because he'd been in there for a long time, and was also able to Do the on the ground detail work of saying, okay, this wind farm is the one that we should work on because it has these characteristics. Let's explore these others. They aren't right because we have to, it would take longer to do this pilot or something. And So he could help us navigate that level. He could help us navigate getting the the approvals from their legal team to put this out there. Luckily, that's fairly straightforward in our case. And then you can also navigate down to the nitty gritty of this is how we plug into the Campbell Scientific Data Logger at the base of the Met Tower to get the data that we need. So Thank you. So just backing up a little bit through building relationships, find the right people who can can help you. Help us navigate through that, that model with, I guess I have fairly little to add here. I would just mention also that it's a, it's a diverse array of organizations that are out there that are operating these wind turbines and some of them will end up being early adopters. Some will be middle of the pack, some will be late adopters. And so we're definitely targeting kind of the early adopters. Our thought is once we target the early adopters and they start making. A lot more money because of deploying our system, then it will move throughout the industry and whether that takes. Two years or six years to kind of permeate the industry. Hard to say for now, but we think there will be early adopters and they will. A demonstration of the technology will drive the change throughout the industry. Yeah. And just so we can put some specifics on here, can you guys give us a figure as to, for example, the output improvement an operator might experience when they're using. Windscape. Sure. Sure. We, we have a, a benchmark that we use a lot in our conversations with wind farm owner operators with potential investors and so on. And that's that sort of the, the median size wind farm in the U. S. is 100 megawatts and that wind farm is going to be, if you take some of the, you know, Average factors for capacity factor and power purchase agreement prices and so on. They'll be selling around 16 million dollars in electricity a year, which is, you know, is a very good cash flow, but it's also just as a footnote here. Wind farms are often operating at very tight margins. So they watch that. Cashflow very carefully because they have a very high expenses. Also, another statistic is that sort of the average expenditure for O and M for wind farm. Just just in the O and M cost is 45. 1, 000 and then the wind farm has a debt service and then they have lots of other expenses that they have to have to deal with 45, 000 per turbine per year. Yeah, per term. I'm sorry. So that, that exactly. So that, that 100. Megawatt wind farm would have 100 turbines time 45, 000. Yeah. And so, Windscape targets an improvement on the order of one to six percent additional energy output, which directly translates into additional revenue. And that's based on studies that have been done of what is possible. If a wind farm can know if the wind is Just before it, it gets to the wind farm and make these adjustments to pitch and yaw and so on that Jason was talking about earlier. So in our example that we give to the potential customers, Say, say 2 percent additional energy. So that's going to be on the order of 300, 000 additional revenue to the wind farm every year for you know, SAS service fees to wind scape, which would be only a small fraction of that. In addition. There would be no actual investment necessary on the part of the wind farm, so, and, and no disruption to their operations. We don't have to, like, bolt something to the top of the turbine. We don't have to insert something into their gearbox or into their nacelle, you know, the box up on top of the tower, so no, no interference with the wind farm and a potential boost to Thank you. Revenue of 300, 000 and just to fill this out when we're talking to wind farm owner operators and saying, does that 2 percent does that 300, 000 mean something to you? And they say, yeah, we, we struggle and we put a lot of effort into something. If we think it can get us half a percent or 1%. More energy out. So answering maybe ahead of time, the question of does that 2 percent matter to people in the wind farm operations world? Yes, it matters very much. Yeah. So where else can you guys see there being real potential here for the application of either software in general, or specifically AI to utility scale? Power generation, the clean energy transition or sustainability, you know, in its broadest sense. So this is, you're asking AI in general, not necessarily, which is great, not necessarily the Windscape product. Yeah, because for example, Jason, you mentioned you did a lot of research and interviewed a lot of people before joining and co founding Windscape. And the climate crisis is part of a bigger sustainability crisis, which obviously Impacts every ecosystem on this planet, every natural and human system on this planet. So you guys, as experts in the power generation industry and in AI and machine learning, you must see from your vantage point other areas where you think we could really make an impact if we could get the right software to interact with these assets. I think it's a good question and it's a good, very broad question as well. I spent a lot of time thinking about this question as I was sort of trying to transition my career from just the science of machine learning to machine learning meets climate change. And I found it was more or less it initially seemed very promising. And then after a little bit of investigation, it seemed Like the intersections were fewer than I might imagine. So there's this paper by someone named David Rolnick called Tackling Climate Change with Machine Learning from a couple years ago. That's 111 page paper where there's different sections and each section is like, we could use machine learning for this, for this, for this, for this. There's almost like eight ideas per page. So a total of, you know, six, 600 ideas. Not counting references, tons of ideas for what we can do with machine learning with data to address climate change. Great, great place to start. The further I got into it though, and the more I started talking to people in different industries and actually interviewing people and asking, You know, what do you actually think about? What do you worry about in your job? Why is your company making or losing money? The more I realized that climate change per se is about It's about atoms. It's about carbon atoms. They used to be in the ground, now they're in the air. That's the main problem with climate change, right? fiXing climate change is also about atoms. It's about atoms and electrons, right? To make the grid greener, we need to build a ton of new transmission lines. Which is steel, it's concrete, we need to change how concrete is made to make it less carbon producing, right? We need to fund all this using, using public funds. We need to get the entire public governments behind this. This is, this is politics. This is moving people's hearts and minds. So far, nothing I mentioned is machine learning. Nothing I mentioned is data. So you have this world of atoms and electrons, which is the cause and will be the source of climate change. And you have a separately this world of data and machine learning and software, which is kind of what we're good at in Silicon Valley. These two worlds intersect sometimes like a little Venn diagram, and there's definitely an intersection there. And that intersection is where some fun companies are getting started, including Landscape. And hopefully some fun companies will succeed, including Landscape. But this intersection is not everything, right? Building new solar farms. Has a little bit to do with data, finding the right parcels, right. Assessing projecting the electricity costs for the next 20 years, but it's also just about corralling capital and pouring concrete and building with it. So, so there are intersections for sure, but I think there, uh, it takes a little while to look for them carefully and to find a reliable intersections with also where there's a business model that would make sense, the perennial startup challenge. Exactly. Exactly. Gentlemen, thank you so much for taking us through Winscape, your journey selling to utilities in the space that's been extremely valuable and hopefully insightful for the entrepreneurs and investors who are listening to this. So to wrap up the show, we always do our rapid fire questions. And I'm going to take one of them out because we've got both of you as guests and we're going to leave four of them left. So without further ado, I'm just going to dive into these and it'd be great to get an answer from both of you on all of these four questions. Limited to like four words or what are we, what are we, just, just shoot from the hip, whatever comes to mind, don't pontificate on it too much. There's no wrong answers. So let's just. throw some stuff on the wall and see what sticks. So who is someone who served as an inspiration for you both on your own sustainability journey or your journey with Windscape? For me, working at Environmental Defense Fund early on, Michael Oppenheimer was very early pushing the world to recognize that climate change was an issue. He wasn't the only one, but he was a big one and a big inspiration. On my end, I've been inspired by many people I've met along the way. One whose work I really like and respect is Saul Griffith. You might have seen some of his books, including Rewiring America. He kind of lays out climate change from the top down and bottom up in terms of why it's happening and what we should do. If you just approach that as an engineering challenge. How would we fix it? And it's not too, too, too complicated. And I really like his kind of engineering practical approach to things. Two good varied answers there. So thank you. So next question then is what is the most overhyped tech or trends right now? And conversely, what's the most under hyped and Eric, we'll keep going with you first. Okay. Well, I'm going to, I'm going to really stick my neck out here and I, I believe there's. Just a little bit too much hype around hydrogen and what all the problems that that hydrogen will solve. I believe that there are going to be some technical challenges and some cost challenges and and and solar and wind are just going to keep coming down and batteries are going to keep coming down and cost so fast. I think the difficulties of hydrogen for it definitely have its niches, but it won't solve all the problems. So and uh, under hyped. I, I think that solar and wind don't get the recognition for the cost curves that they're still on. And we'll keep, keep coming down. I could give a pithy answer, which is that. I think large language models are both over hyped and under hyped. Clearly they are changing the world and they will change the world in huge ways. But there's also a lot of stories being spun that aren't really going to play out, I think. I think a more real answer for under hype though would be just electric heat pumps. So, I don't know who I'm going to offend by saying this, but electric heat pumps are unfortunately like a little bit boring. You install them in your house, they heat your house slightly differently, they make the world greener, and they save you money over the long term, but they're a little bit expensive to install. They're not very new technologies, but I think they'll be part of the solution that's required to address climate change. Yeah. And especially in Europe, you know, where a lot of the wealth is, is in Northern Europe, which is a lot cooler. You know, there's electric heat pumps, the scene is this huge opportunity. But yeah, a lot of people want to stick with gas. I think they see the cost. I think they see all this upheaval, et cetera. So, so yeah, I can really empathize with that. And well done gents. Cause usually founds in this juncture say under hype technology is usually. I don't know why mesh sensors being applied to wind farms, for example. So you both. I don't know if you've heard, but that's usually the self promotion. You guys are swerve there. So if you guys had a billion dollars to put towards the sustainability and climate crisis, where would you invest it? I would think it's not a little vague, but I would invest it in an organization that was focused on the equity side of the climate crisis. And if, and that's just, we've got a lot of cool technologies and a lot of money going to cool technologies, but there are literally billions of people who are going to be first and most adversely affected by the climate crisis. And a moral. Level and also on a just sort of a keep the world sane level. If we don't first invest in that we'll be, we'll really regret it. Jason. Yeah. I like, I like, I like Eric's answer. Actually. I think that's it's a really important factor. I don't exactly know how it would literally invest a billion dollars in doing that, but I think, yeah, you know, you can play someone smart to help you help you figure that out. Yeah. I also think if we could institute a price on carbon, I think it would go a long way toward. Addressing these externalities that are affecting everyone in a fair way, and I do believe in fundamental economic and competition and capitalism. And I think if we could implement a price on carbon, it would really help drive a lot of these other changes, including helping the people in New Northern Europe realize that maybe gas is actually expensive when they consider the externalities on everyone. So, so I think your answer is that firstly, Climate equity, it's probably the first time anyone's ever talked about adaptation rather than mitigation of climate. So just want to commend you on that and, and that's, I think that's a, an accurate observation, but we'll have to check the rest of the episodes and, and a price on carbon is always mooted. And I think it seems to be that obvious truth that staring everyone in the face, but for a huge number of reasons, the world just can't get his head around implementing it, or at least some parts of the world have been more effective than others. But We haven't got the unified carbon price that we really need to make a difference here. I think what we required, and I'm not the expert on this, but a price on carbon. So we, we tax the use of gasoline, right? Which brings in a ton of money, but then we need to distribute that money to help people that can't right now quite afford it to buy electric cars and everything. Right. So it's, it's not just the one directional flow of money. It's the bi directional flow. Yeah. Yeah. Well said. And so if you could magically enact one policy reform. With no financial considerations to it, we're just talking about the policy on the cash. What would that be? I think I gave my answer. You designated me as going first, but I'll echo what Jason said in the last one. Yeah, that bi directional price on carbon. Yeah. If you can make sure that the proceeds are distributed. To address the, the regressive nature of tax on carbon. But, you know, climate, the climate crisis is extremely regressive. And if you were to implement a program that would address that and carbon emissions less attractive, that would be really. Powerful. Well, great answers there, gentlemen. So thank you very much for sharing those with us. Thank you so much for coming on the podcast today. If any investors, founders, people in the industry want to find out more about Windscape and what you guys are doing, where can they find you and how can they contact you? Just our website, I guess, windscape. ai. And we have a Contact info there. If anyone wants to reach out. Well, we're happy to talk to people on LinkedIn as well. Well, we'll put all that in the show notes. So thank you very much for coming on, sharing with us about how you guys are trying to optimize the wind industry, the approach you guys are taking to AI and ML. Sounds fascinating. I wish you all the best of luck and thank you very much again for coming on the show. Thank you, Jake. Thanks, Jake. See ya.