Nishan Degnarain Hello, everybody, and welcome to our first episode of ESG in space. My name is Nishan Degnarain, and I'm one of your host today and delighted to be joined by. Miki Sode I'm Miki Sode. And we are happy to be here. And we're happy that you are here.
Nishan Degnarain Fantastic. And, today we're grateful to our partners, the Exponential Academy and Secure World Foundation. And the theme of this series of podcasts is to talk about whether we need some sort of sustainability or ESG framework for space. Was it fine for the space industry to continue developing as it is with the development of the private space industry?
Nishan Degnarain Special guest for this podcast is and I'm delighted to introduce is our friend Bruno Sanchez. Nuno. Bruno has got a long and storied, history. He's a scientist. He's an author. He's a young global leader of the World Economic Forum. He was the, founder of the planetary computer work at Microsoft. He's been a scientist.
Nishan Degnarain Work in satellite and AI technology at the world Bank. He was a chief scientist at Mapbox, looking at various mapping, projects that uses satellites and and the author of a book called Impact Science and How Science Have a Positive Impact on the World. And today, he is the executive director and founder of Clay, which is using artificial intelligence, artificial intelligence, open source and open access data for using for satellite imagery, for for good.
Nishan Degnarain With that, Bruno, welcome. Bruno Sanchez Thank you. Thank you very much, Miki and Nishan. I'm really excited to talk with you about this. Ideas. Miki Sode Yeah, the best thing. Wonderful. Yes. Yeah. Thank you for, coming on to our show. And, should we just dive right in? Yeah. Nishan Degnarain Let's do it.
Miki Sode Okay. All right. So I have, first question, and it could be a kind of intro into setting the context and, I know you have an extensive background working at Microsoft, but it's always around, like, mapping the Earth. Right? And now you started this, nonprofit organization, Clay, about a year ago. Can you tell us, what Clay does, like, how does it work? Miki Sode And what inspired you to start Clay?
Bruno Sanchez It's, it's a long time coming. Dream to work on these issues. It's the confluence of a lot of pain points and challenges I saw during my career. Like. Like Nissan was saying, I helped create a company like Mapbox. It was a startup. Or large corporations, like in large corporations like Microsoft, the planet that a computer, which is about putting the data together but also work at the world Bank for a while and how to maximize the social impact, of these technologies.
Bruno Sanchez And it it has always been and it is really difficult to to use that data like we have data over the entire world of the thing in the background. We have literally data of every place on Earth for the last decades, but it's still really hard to to use it. Like the answer of questions, like how many trees are there in the world or where there are fires like this?
Bruno Sanchez Until now in Portugal there are a lot. So lots of fires happen. You know, the floods also across Europe. The data of that, it's there and it's there either from public sources like Sentinel in Europe or NASA, Landsat or commercial companies. This is not our commercial or versus noncommercial use about them using that data. And what we realized a year ago is that this last wave of AI, it's gonna unlock that beauty.
Bruno Sanchez So one of the beauties of this revolution, in the case of text, is it's not so much that is something new, but it is that it reduces or removes the barrier of knowledge. Like if if I want to learn quantum physics in Swahili, I can do that with ChatGPT or other methods. So what is the equivalent of Earth?
Bruno Sanchez What is the equivalent of putting that knowledge or that awareness of Earth? And what happens if we do it as a nonprofit so that it's not only those who access to the, the, the, the budgets, like, for example, right now a lot of the models are done commercially, and that's perfect. That's great. It's a lot of generate a lot will generate a lot of profits and value for services that are important.
Bruno Sanchez But then what would happen if you do this as a nonprofit to unlock impact in the world, but also unlock investments of companies that build on that? And that's what I believe. That's basically right.
Nishan Degnarain So this is fascinating. You're innovating in two areas, right. So you innovated number one in terms of how do you use application of space or satellite technologies for good. Right. For purpose, for your innovating on the business model, right, that you have as a nonprofit, you know, philanthropically funded, not a government funded project, not a the project as a so that is that is fascinating.
Nishan Degnarain I mean, can you give us maybe a 62nd demonstration of what Clay is just a yeah. And so we can see it. Bruno Sanchez And by the way, it's also to use the ESG acronym, the G of governance. This is the one we are not innovating but we are struggling with. It is not clear what is the right governance for a nonprofit AI tool. That is something that we are very explicit to say. And so we want to innovate in the business model to unlock a lot of value commercially and not.
Bruno Sanchez But we also would like to be smarter about the governance of these things. And we just don't know. I think the nonprofit governance, how it was designed was not meant for technology in general, but for AI in particular. And that's something we have. So a shout out to all the audience here is that if you are experts on governance of AI tools, we are struggling with that because what we're building is clearly extremely powerful.
Bruno Sanchez But also we want to make sure that it benefits the most people. And to answer your question, Nissan, if we can, there is a demo that I will do in a second that I will, speak what we see while we demo that, so that those who are listening to this also get the idea. So if you go to the tool and we can put the screen, there you go.
Bruno Sanchez You can go to it to explore dot made with clay.org.org. That's the demonstration product and what it looks like a map. And you can select certain regions. And it basically allows you to assuming in a place. And the idea here is that this is a special map and it's a special map because it understands what is looking at.
Bruno Sanchez And what I mean by that is that let's I'm scrolling to a semi random place, because I've done this a few times and I say I click on I saw a lot of it. I know this is not a panel in the middle of the Mojave Desert, and I click on it and basically I'm telling the AI that thing that I'm clicking, I want more of those.
Bruno Sanchez And traditionally this would be computer vision. You would need to create a whole complex environment from pixels to solar panels to find more of those. But with these AI tools, because it understands the concepts, that is, that it's looking at through these examples immediately in like hundred milliseconds, it's only five other examples, 500 because the that's the limit.
Bruno Sanchez I put five. And the other example of solar panels and it's remarkably, remarkably good. I'm scrolling through the examples and all of them are correct. But imagine some were not correct at some point. This one down in the results is not a solar panel. So basically this is not the one I'm looking for. This is what I'm looking for.
Bruno Sanchez And I click again. Fine. What I'm doing is I'm telling the AI what is my intent? What is the thing I'm looking for? I need like a couple of iterations of a human in the loop. I'm finding that what is does not the the the value, the values that just like for one how two panels you can do exactly the same for swimming pools, for biomass of the forest, for plots, for fires, for all of those things we had in our images.
Bruno Sanchez Now we can index, then we can find them. And if we do this again as a nonprofit that allows anyone to use these, combined with the petabytes of open data we have, we basically revolutionize the expectations of what anyone can know about Earth.
Nishan Degnarain And you're rewarding it. So now let me come in and you're rewarding, you know, satellite image providers and other, you know, Earth observation data providers, in some way, by being part of the clay system, I'm assuming, you know, just so that they can showcase because I'm assuming a lot of this data otherwise by private providers would have been behind paywalls and not accessible for scientists or so.
Nishan Degnarain Yeah, those would the global South, those who can't afford to pay the fees.
Bruno Sanchez This image well, what we're seeing here and what we used to play to train with, we've trained play only with open data so that the result is also an open model, which means that there is no restrictions or licensing like we have problems with text, models of AI, train of text. There is a whole discussion about IP licensing, but because we only use open data for play, we can do whatever we want, including private commercial satellite company can take play and continue training with their own data to make it even better.
Bruno Sanchez There are things we cannot see in open data that we could see in commercial data. Not only that, you can use clay, trained or not, with data from your prior, commercial data to index so that not only I know solar panels are seen from open data, I can also search for solar panels as seen in commercial data in places I might not have, or times I may not have access with open data.
Bruno Sanchez So it surfaces basically it it makes the analogy that I sometimes use is that we have a library of books in a language we cannot read, so that it's really hard to find the right page. Now we have an AI tool that has read all the books. I can tell you exactly what book to go to find what you're looking for.
Nishan Degnarain So so so one question is an approach, Miki. So so so with this, this assumes a couple of things, right? It assumes that you have some sort of interoperable standards. Right. Because you have found a way, you know, with your base map, to ensure that there is some sort of interface or intermediary, you have imagery that could be geolocated to set the file types up into operable.
Nishan Degnarain I'm assuming if you take so somehow, you know, the industry has created it, and I'm assuming there are some areas which are proprietary data that's not interoperable. And so they can't be included. And there'll be others that will be developed maybe from other nations, for example, that have a different set of standards that will not be interoperable here.
Nishan Degnarain I mean, how do you create? Because that's one of the big challenges when we talk about ESG and even standardization. We see that recently with the astronauts stack, right now in the space of the ES, the space station, because they were on a Boeing craft and space is a different. So so how are you thinking like how do standards get created in this industry? Nishan Degnarain And is there a need for some sort of standard setting body for this sort of.
Bruno Sanchez This is only possible, or the reason this is much easier to do than text is because there are there are already standards of data formats that are cloud friendly and very eye friendly. So that already is possible because of existing standards, because this GR I think is so new. There are also lack of standards for many use cases, many applications like model definitions.
Bruno Sanchez So the output definitions of this embedding which is this I summarize. So we don't have those standards. So again being a nonprofit puts you in a special position to help set the standard. So if now anyone can use clay either with or without fine you know, commercial, then we can start saying as a de facto that clay is the standard or could be the standard.
Bruno Sanchez And we are happy to shape the outputs and talk with the partners and say, hey, if we do this this way, then it would be easier, more interoperable. So, yes, the standards are very important. In some cases it's possible because there are standards, but we are also discovering that this is also new, that there's no standards, because this is terra incognita.
Bruno Sanchez This is something extremely new we don't really understand. Well, it's it also connects to the governance who should be on the table talking about these issues. Who should be like is the places like when we work in the, you know, communities? It's good that they could use the model because it's an open model, but do they know how to use the model?
Bruno Sanchez We get their input, so we will make sure that they benefit not only from the model landed on them, but also from the work and the knowledge they have locally to make the models better. Those are important.
Miki Sode So let's talk about the community right. And then who should be in this conversation? We did touch upon, like commercial satellite imagery providers. But how about the user side. Right. Can you tell us more about, who are the early, users, for the clay platform? Because like, I, if I remember correctly, Google Earth Engine, I think it was like forestry people who kind of latched on to it.
Miki Sode Because the reason why I bring that up is like, you know, there would be first wave of people trying to use it, and then that would be a bigger voice, I guess. And then as the other people will come in, that will kind of form a community. But I kind of wanted to get the sense of like, who, like see, immediate value in what Clay has to offer that's different from other platform. Miki Sode And that could be kind of get a glimpse of what might, you know, transpire. Yeah. From now.
Bruno Sanchez On. Yeah. It's also a very good question. And not only that, but it is a cleaner slate in Texas models or vision models or audio models, that is already the customers and expectations on the story is good and not so good to to shape the narrative. But with AI for Earth it's less common. So there is more opportunity of what users do.
Bruno Sanchez We talk about here to say the expectations out of this defense and intelligence, customers, which, for example, dominated the whole study of remote sensing from the beginning because they were how the industry basically got started. Or is it more on environmental issues or social issues? Right. So the real customers of the model that we started with, the first one was a biomass estimate.
Bruno Sanchez So one of the things that is extremely important is to have really good measures of how many tonnes of CO2, so how many tonnes of carbon are captured, for example, with forests. So we can create carbon credits so we can sell those carbon credits. The quality of those can reduce all of those things. It takes a lot of effort to do that properly, to literally go to the forest and measure the diameter of the tree.
Bruno Sanchez So there are many ways using satellite images and using other methods to speed up. And now with models like clay, you can take the example of a pupil at a few places that we've done that manually or with other methods, and then train an AI model to use those semantics. It has learned to speed up the process and what we are seeing is that we need far fewer examples than traditional to make the same, or 90% accuracy of the same outputs.
Bruno Sanchez Hundreds or a thousand times faster. And because it's an open model that third parties can kind of sense, I can test, then we can said there are standards, which is an independent from the model provider or the the the creator of the credit. So that to me is a fantastic story. And by the way, similar stories exist for weather forecasts.
Bruno Sanchez So for other cases where we see AI models being a cost play, basically doing things cheaper or doing things even better, faster, that's that's basically the studies we want to highlight there. That's another story of a predecessor of clay, which is a product called Earth Index, which is a similar technique which was used to detect illegal mining and even the front page of The New York Times.
Bruno Sanchez It's a fantastic story to follow, with a few examples of illegal mining in the Amazon, because we have contracts in the Amazon to say, hey, I know that identical mine is in here, can you use your AI to find other places that might be illegal mining or industry types of to land claims and what they did, and they discover more.
Bruno Sanchez And that's again another good story that highlights the social value, not just the commercial or corporate or mining or defense, which has its own uses and its benefits and challenges. But that's not how we want to shape this story.
Nishan Degnarain If it's, because interesting. I love the premise of how this podcast came about has been, you know, what will it take for the space economy to grow right up until, you know, the yeah, the the 20th century, you know, the 20th century, the space industry was dominated by nation states, right? The US, Soviet Union, Russia, the emerging country, or Japan, European Space Agency, now China and Israel and India and other other nations.
Nishan Degnarain But then since you know the space X Prize, we now have the launch of the private space industry. And in this conversation we had before this podcast, you had mentioned that, you know, a lot of the space industry is either dominated by the defense industry, by science, or and now we have the struggle for kind of what is a viable commercial sector look like an even like viable commercial sector.
Nishan Degnarain This is a clear case of where you're trying to use, you know, an economy for good, right? We've we've been very extractive in some of the other private sector practices. I mean, do you see companies being rewarded for, you know, putting up technologies into space that kind of a beneficial impact? You know, can we create business models, for example, that can track, as you said, whether it's methane emissions or whether it's.
Nishan Degnarain Yeah, yeah, illegal fishing activities, like how do we incentivize that sort of activities. Miki Sode In that way? It's like, it's like a carbon credit. Right? You need to be able to track to see the, your carbon footprint and, you know, impacts to the environment and stuff like, can we extend to think like, are we as we are going to outer space, are we keeping the space environment in as good of environment as we found it in a way.
Miki Sode Right. Like so that so the longer term, we can all have an access and benefit from what unique value that space offers. You know, what the value is a is a big question. But you know, yeah, yeah.
Bruno Sanchez There are this is definitely a complex issue. So it's good that this kind of podcast, exist because you can think about the sustainability of a space in the sense of keeping the commons of a space in good shape so that there's an equal opportunity. So that maximizes the impact for the most amount of people, things like, debris in orbit and things like that.
Bruno Sanchez Right. There's a whole set of issues, but also sustainability on Earth using the space assets. It's not only the it's not only the liabilities of being in the space, it's also the opportunities of being in a space or even the cost of not doing the space. Like if we not in a space and we're not able to do the things we can do commercially or not, we would be in a full race to understand what's happening on Earth.
Bruno Sanchez So sustainability on Earth depends more and more on space assets. So we should also incentivize those things. And then of course the question of my who of course, right now the largest operator of communications, images of of space are private companies like planet and space X with them Starlink. And is that a good thing? Is that a bad thing?
Bruno Sanchez What is the governance around those things? Like if it were native stats like we used to be the US, basically the country would win these constellations. It might seem that it's easier, better governance, but all such types of values that also dominate in the space, it's a whole other conversation. How do you ensure that the governance of a space when it's dominated with commercial players is defective?
Bruno Sanchez One and one of the mechanisms of doing that, I do not have the answers, but at least we are in a position where we are starting to really ask those tough questions and, pay the price of not having given at least an answer right now. Technically or practically speaking, yes, the US can do whatever they want, and US companies can do whatever they want to spend, but then at the cost of the opportunities that are much harder for others.
Bruno Sanchez Is that what we mean? So when we do play as a nonprofit, it's also a way of saying maybe there is a different way we can complement is nothing against commercial players. They provide us really provide a value, otherwise there wouldn't be a commercial value on there. So how can we maximize that benefit for everyone? Miki Sode
Nishan Degnarain And so that raised an interesting question. An interesting question. You know, Clay is a very early stage, you know, nonprofit or project right now. But you could imagine, let's say, and hopefully a clay scales off becomes fairly well used. Yeah. We talked about or week talked about the risk of space debris. And we know that there are different, you know, satellite providers, for example, some of which create more debris because they don't have navic ability, for example, of their satellites or they don't have the right information communication to other.
Nishan Degnarain Yeah. What could you see? A world where you would also talk about the ethical sourcing or ethical use of data, right, where you actually source and have some sort of rating scale. So that Clay itself said we would only source data from ethical satellite providers that comply with a certain code of practice, and you would refuse data from, from outside, not not life.
Bruno Sanchez And what is it cost like literally like, if I'm not mistaken, the the only way is to get daily images of anywhere on earth is planet and in the constellation of dogs they have do not have the capacity to avoid in short term the collisions. But that also gives us daily images of anywhere on Earth. If we force them to have these collision avoidance systems, they might not be possible to put them.
Bruno Sanchez So is this something we are okay with? Really? They are. Are we? Is the war okay with those things? Maybe the answer is yes, because those daily images maybe turn out to be extremely useful in kinds of disasters. Or maybe the answer is no, because if there are collisions that might happen, or when they happen, then the cost of datasets every one or many years to come.
Bruno Sanchez But again, this is this. This is not hypothetical. This is not an academic question or theoretical questions. This is a real choice. We've done not asking the question. And the challenge we've done is that ethics okay. Is that okay?
Miki Sode In a way kind of similar conversation we've been hearing lately in the news about AI. Right. I, I just boomed in the last, you know, a year or two, but now we realize that it has a big, environmental impact because the computational power is so large and yes, it has potential. And, you know, lots of may solve many issues.
Miki Sode And, will give us answers to hard questions, and all so has a benefit, but at the same time, at what cost? Right. And we also, you know, it's AI is still powerful, but many people are using it for, I don't know. Nishan Degnarain I think other things that they should be using it for as well.
Miki Sode Right, right, right, right. So so you know like you know, what's the cost versus the benefit. And I think that's kind of like I have a two there. Seems like a lot of that like trade offs I guess. Yeah.
Bruno Sanchez And by the way, the answer is not the yes or no. The answer is the how we build it. AI uses a lot of resources, and that's a problem with not using AI also has a lot of cost. When we see AI being able to do better forecasts like a thousand times faster, like Microsoft Aurora, that means we have potentially 50 times more time.
Bruno Sanchez When is a hurricane come in to put people out of harm's way? So the cost of AI is clear for the missions to train that model, but the cost of not doing it goes, I think so something that is going to affect the community. So then the question is, okay, there is a cost. How do we minimize the cost?
Bruno Sanchez Are we okay with a world that is like five. Amazing really, really large models when OpenAI and Microsoft, Google and others? Or should we favor the creation of a longer tail or a smaller and smaller model? This is specific for specific cases that maybe could be combined together or could be minimized. Maybe is an example of obviously the second there is an effort to say, hey, we have $5 million instead of making ten, instead of having ten small models of $1 million, why don't we make a nonprofit that makes one model for $5 million and the missions that that carries, and then everyone can use those so that that we then need to do ten
Bruno Sanchez times the missions for a model that is ten times worse. Nishan Degnarain So this is interesting because it takes you and Clay. It is three ethical discussions, right. Number one can have an ethical discussion about AI. Like like the compute load, the energy, should you have large or small, should only have one monopoly provider versus kind of competition. So there's one that I the second is about the source of your satellite data.
Nishan Degnarain Like is it ethical provisioning. Who has actually been deploying the satellite. How you know, what was the fuel consumption or the carbon footprint of those satellites? It's kind of cool. Yeah. And then the third is kind of the ethical use of that, you know, with those satellites intended to serve, you know, certain industry, for example, the oil and gas industry or defense industries, others.
Nishan Degnarain So we may disagree with, you know, tobacco industry or was it, you know, designed to help with environmental challenges that so so the three kind of pose and I guess it could be overwhelming operationally to then start to navigate, how do we see the self so that who is best place to kind of host or regulate or facilitate these discussions.
Nishan Degnarain Because we now are no longer into this yes no legislative regulatory framework. It's about business operating practices that will shift and change evolving tradeoffs. Do you see a body in that place facilitating these discussions, or is that a the need for a new neutral body body to come in and help govern these conversations?
Bruno Sanchez The practical answer is that the market is one of those regulators. I'm not saying it's the right regulator, but clearly the market itself is regulated to because the these commercial companies are investing so much money in doing that. They do that because they believe that the customers, we reward them buying the products and services derive from those. So that's the de facto answer.
Bruno Sanchez Of course, there's also the de facto answer that we have bodies like, you know, so from the UN which which also aims to regulate that or software power like the World Economic Forum and others will convene and try to maximize the value of that. The reality is that, and then frameworks like ESG that came from the from the UN report, who to try to do those things, probably the right answer is a combination of both with carrots and sticks.
Bruno Sanchez And of course, the carrot of the commercial, the market, the, the consumers high paying for something or not is one of them. But there are also sticks to do that and in may have comes down in AI or even a space. Should it be an IP play where there are products and services and reward research and development for doing these things?
Bruno Sanchez So it could be a commodity where there are some set of expectations of standards and we can play, we can change back and forth because it's not the thing itself is using the thing. So our commodity all to be, public utility, what is much more regulated. There's this very strict expectation. Understand that if it's a public utility, I can see mechanisms where the governance play.
Bruno Sanchez And if it's an IP play, then it's basically the commercial sector. The right answer right now is that it's 99% a market play. And I don't think that's the right balance.
Nishan Degnarain Excellent. Well, now we've come to the top of the the time that we have over here. And you've given us a very heavy question. I think that's said that right now is a primarily a market play. We need a balance. We don't know what that balance looks like. But this is certainly gives us food for thought. Micki, do you have any reflections on, on, on on the discussion we've had with Bruno today?
Miki Sode Yeah. And I, have my thought is that, like, we really need to have this conversation, right? I mean, here in the podcast, too. But then, as the community of Clay involves. Oh, sorry. Expands, right. I think we will probably start hearing more voices from different sectors, different players. And now it is a market play. But then the would the if the voice gets amplified then that will start having more I guess, you know, authorities to come in.
Miki Sode But then another point that I really want to, take home, you know, personally is that there is a design ability to it. How can we, you know, make it so that it's, you know, there is opportunity, but also there is certain rule that people can abide by so that we could all benefit from, it, from, from that. Miki Sode And there's a design ability and the conversation is where we can start and, that's my takeaway and I'm grateful for this conversation.
Nishan Degnarain Excellent. Having intention with how we approach space. Bruno, thank you very much for your time and sharing your insights with us. And we were delighted to have you on today's podcast and look forward to our next discussion. All the best with everyone. Bruno Sanchez Thank you. Miki Sode Thank you so much.