The Future of Earth, AI + The Environment
9:02AM May 11, 2020
Good morning. Good afternoon. Good evening, and welcome to episode eight of the AI for Good Webinar Series. We hope that you, your family, your friends, and your colleagues are all keeping healthy and safe. My name is Fred Werner. I'm from the ITU, the International Telecommunication Union in Geneva. And it's a privilege for me to introduce today's webinar. Now, vi t u is the United Nations specialized agency for information and communication technologies. And we're also the organizers of the AI for Good Global Summit, alongside with XPrize Foundation, and then partnership with 36 UN agencies, and ACM and Switzerland. And the goal of the AI Summit is to identify practical applications of AI to advance for Sustainable Development Goals, and scalable solutions for global impact. And like most of the world, the AI Summit has gone digital with weekly programming between now and the end of the year, allowing us to reach even more people and today's
Webinar could be considered as part one of the AI and environment breakthrough track that would have been taking place in Geneva this week. We're not for the virus. Now before I introduce today's moderator, I'd like to go over a few housekeeping issues. First of all, your microphone has been disabled. So please use the chat and q&a function if you wish to communicate. THE MODERATOR is responsible for identifying and asking questions to the panelists. And we're counting on your active participation to create a very interactive session. So without further ado, I'd like to introduce today's moderator. His name is Marcus extrovert and he's the XPrize lead on environment and also running the XPrize on NRG by NRG cozia on carbon, so basically the carbon XPrize so Marcus, welcome and the show is all yours.
Thank you so much, Fred. It's great to be here. So as Fred mentioned, my name is Marcus x devore. And with X Prize, we're really thrilled to be part of
In partnership with ITU eu, supporting the AI for Good Global Summit,
part of me is sad that we can't do this in person in Geneva, but part of me is happy that we get to experiment with this new format, which will allow us I think, to take the conversation in different ways and actually include a broader audience. We're really thrilled to see that there's strong turnout here today, I see the number of participants clicking up. So
we're really thrilled to have you on board. Thank you for taking some time with us. I'm really excited about today's session, because I'll be really honest with you, I've spent a long time in physical science,
even dabbled in computing, I come from that that world. But I'm not an insider in the AI and machine learning world. But I do work a lot in environmental sustainability and climate oriented technologies and innovation. I've always seen AI and machine learning as another frontier, which is I think, an obvious thing to say but specifically as it's applied to Earth systems, whether it's satellite imaging, ground based imaging, tracking and monitoring in different data, environmental monitoring, but also actually developing new solutions that can take us
Think forward to that sustainable future. So I'm personally really excited to get to learn a lot about this topic and dig in with a couple of experts. I'd love to introduce our experts. For today, our panelists are really going to take us on this journey and guide the conversation. We are looking forward to a lot of questions and feedback from the audience. There'll be a lot of time for that we have a good amount of time for our session. So please prepare your your questions and comments and we've got a q&a function for you that we'll get into in a few minutes time. But with us today, our two fantastic panelists, Andrew Zoli, is the Vice President of global impact at planet planet is a space and AI driven organization that's known for having deployed the largest constellation of Earth Observing satellites in history.
Our other panelist is Sacha luciani, who is director of scientific projects for AI for humanity initiatives at the Mueller Institute, and also the director of scientific projects for AI for humanity. Without further ado, I'd love to invite Sasha and drew to join our session.
Hi. It's great to be with you. Likewise. Thanks for joining us here. So what we're going to do for your sake and for our audiences, I'll ask you both to give some short presentations and I know you've prepared a few remarks, to not just introduce yourselves and your work, but how your work fits into the context of our topic today, future of Earth AI environment. I think we're gonna go to Sasha first and then to Andrew. So Sasha, whenever you're ready, please take it away.
And give us a background of where you're coming from. From there. We'll go to Andrew and after that, we'll be able to get into a bit of a discussion. So over to you, Sasha.
Can you see my slides?
Yep, they look good.
Okay. So just to give you a little background, I am in Montreal, where it is actually snowing today.
Be happy wherever you are when it's warm. And to give you a background about about me, I have a background in
Machine Learning in computer science. And for the longest time, I felt really overwhelmed by the climate crisis just because it seems so much bigger than I was that I am. And back in my previous life when I was an applied machine learning researcher, I was reading about these all these solutions like capturing carbon in the air spraying aerosols in the atmosphere, going renewable, taxing carbon, and it all felt like such a, such a big endeavor. And so I decided to plunge right into dive right in. And two years ago, I joined the Mueller Institute to work on AI and climate change initiatives. And and so I work on a bunch of different projects. I'll present some of them today. And essentially, I'd like to present different levels of things that can be done to tackle climate change. So things that have to be done by governments, by companies, by individuals like you and me, and essentially what it would take to tackle this crisis.
So just to give you a brief background of what I mean when I talk about machine learning, so AI, which machine learning is a subtype. So it's a general way of using data or information to help computers learn. And so it can be things in the in the present tense. And in the present time, for example, putting things into categories, it can also be something in the future. So predicting where things are going to go. So there's no magic, it's not. It's not this black box and mystery. It's actually really concrete ways of helping computers just get a better grasp grasp of data. And now what does it mean for climate change? So me and 20 of colleagues, working in machine learning all over the world, we put our heads together and wrote a paper that's quite long, but essentially, that talks about the different applications of AI in tackling climate change. So in this graph, you'll see it's kind of hard to read, but at the top you'll see different applications of machine learning like vision.
Which is using images and LP, which is using text, etc, etc. So we're trying to cover all the different types of machine learning and on the left are essentially applications in which they can be used to tackle climate change things like electricity, transportation, industry and farms and forests. And essentially, the darker the blue, the more potential we found. And and so we wrote that we wrote this paper altogether. And then after that, we created an organization that we call climate change AI. And we meet regularly and we try to facilitate essentially, the vision of our organization is how to empower work that meaningfully addresses the climate crisis using AI. So we make connections happen we we work on our own projects, we will work on partnership projects. So and within this organization, I'm in charge of the of resources and datasets because, as I said, computers learn from data so they need a lot of data. So we're trying to put together resources and data that people can can use themselves in order to get a handle on this climate crisis. My my
day job, my main project is with regards to visualizing climate change. And so essentially, I find that the main obstacle for humanity to act on climate change is this problem we have of actually imagining it and visualizing climate change. And I think that AI can help with that. So we're working on on a project that takes climate models that exist that have been around for a long time and have are updated regularly by the IPCC by organizations, and essentially their numerical model, right? They're really hard to to actually see like, even if you download it, it's just a bunch of numbers. And someone has to do this conversion. And often, the conversion is in degrees of warming. It can be in rcps, which are projections of where the plant is going to go in the future. But we're creating images from these from these models, essentially taking their projections and transforming them to images. So this is an image of Old Montreal as it is now. And actually, since it's so close to the St. Lawrence River is going to experience a large amount of flooding
Based on these projections, and based on the warming scenarios, so here's an image of an Old Montreal flooded. And so our idea is to help people really realize what this means what 1.5 degrees means means what what an RCP means. And so here's some other images that we generate. Essentially, we're creating a website where someone can log in, or just visit the website and enter an address. And see based on the on the global climate models, how this address is going to change. So it can be flooding. So currently, flooding is our best transformation. But it can also be drought, it can also be small it can be
it can be lots of different things, depending on where people live. And so our idea is to make this kind of abstract concept as concrete as possible,
and to help people get a grasp of climate change. So I'm looking forward to this discussion and answering any questions you may have.
Okay, terrific. Thank you, Sasha. A lot to unpack there, and I'm looking forward to digging into a little bit more. Before we do that, let me throw it over.
To Andrew, and ask Andrew to share some intro remarks.
That's great. Hi, everybody. I'm just going to start my video. It's, it's terrific to be here with all of you. And I'm getting with, with Sasha and Marcus and Fred's permission, I'm going to give you a about a 10 minute quick presentation on on some of the work that we're doing using data and AI to help drive systems level change and how we think about stewarding the planet. And I. And I just want to say this is such a weird moment. And it's a particularly dispiriting moment when we realized that global pandemic that we're in the middle of
is, is really just a subset of this much larger set of nested crises and that we, you know, we're living in this really remarkable moment. We, we know from the scientific consensus that we have less than 10 years to avoid locking in the worst effects of climate change which Sasha's making visible to everyone.
And, and we are living through a really a series of connected crises in, in climate, in the loss of nature and in the the effects that they have on within the human community. So these three, sort of huge challenges that that are all happening to us concurrently. And we're in this sort of weird vertigo moment where we are facing civilization scale and planetary scale challenges. And at the same time, so you know, if you just look at the scale of those things, that's terrifying. But we're also living through, you know, at the same moment that we're living through the sixth extinction, and we're living through the climate crisis. We're also living through the second Renaissance, we're living through a period where we have the most powerful tools that we've ever had to bring to bear on those challenges to help illuminate them and to help guide our collective action on them. And so you know, if you're we're in between
These very powerful downsides and very powerful upsides and reasons for optimism and reasons for pessimism. And so if you're optimistic, there's good data to suggest why you might be optimistic about our ability to solve these problems. If you're pessimistic, there's good evidence for why you should be pessimistic and if you're confused and in the middle, there's lots of reasons for feeling vertiginous. I want to share with you a little bit about what we do here at planet. As Marcus mentioned, we're a space and AI organization. And the focus on how we create more inclusive planetary stewardship is a significant focus for us. So I'm going to share a few slides with you this morning or this evening or this afternoon, wherever this finds you.
Okay, so I'm going to just click in here and talk a little bit just just by quick way of introduction, because many of you on the phone today or on the chat are not going to be from an aerospace background. Let me just tell you a little bit about planet what we do. Our mission is to use space to help life on Earth. We were
founded with that mission, and we're structured and governed in a way to sort of drive that we, our practical mission is to image the whole earth every day and to make global change, visible, accessible and actionable. So what Sasha is doing is terrific. It's she's actually also making change visible, accessible and actionable. By helping people understand what the potential consequences are, we use the same tools to try to understand both what is and what might happen in the future. And and the biggest reason for that has to do with salience as Sasha mentioned, which is that these changes are absolutely critical, but they're kind of cognitively removed from us on a day to day basis. They're not the sorts of things that have a high degree of salience. So making change, visceral is a really critical part. And making it visible, you know, is a really critical part because seeing change is the first step to making change, at least making intentional change. So let me walk you through a little
bit about what we do that the organization that I worked for, and I oversee what's called our global impact portfolio. So I touch much of our work on sustainable development and on climate and on Humanitarian Affairs and and how we use these tools to achieve their highest and best purposes. The anchor is the development of these small scale satellites, which we've deployed in a constellation around the Earth, essentially orbits from the North Pole down over the equator between the Sun and the Earth, and then over the South, under the South Pole mount up the dark side of the earth. And as the Earth turns sideways underneath them, they collectively image the entire surface of the planet every day, that roughly three meters per pixel. And then we have a second group of satellites that can zoom in selectively anywhere on the earth, and image multiple times a day at 70 to 80 centimeters per pixel. So we use the daily monitoring of the earth to understand what gross level changes occurring, and then we use the higher resolution assets to go in and understand
But taking pictures from space is really just the beginning. Because, you know, the use of AI and machine learning is extraordinarily to an extraordinary degree, it's gated by data by our ability to pull new kinds of information. And so we're producing, we start by producing large amounts of information, then what we do is we use the tools of machine learning and computer vision, the same tools that Google uses to tell you whether that's a picture of a puppy or kitten to classify what we're seeing in the imagery. So for instance, this is a picture from from downtown San Francisco near where our offices are, all the buildings here are in blue, all the roads in red, and we continuously extract this kind of structural insight and not just for roads and buildings, but for planes and boats and deforestation and a number of other critical categories of things. And that allows us to get a sense for the kind of next order of change. In fact, actually, we're not doing it just here in San Francisco or I have By the way, I happen to be in vinyasa.
St. Paul today where it's not snowing Sasha, but it suddenly turned very chilly. So it's a we're on our slow path to spring. But what we can do here at planet is not just look at all the roads and buildings in downtown San Francisco, but we can capture them at any arbitrary time point, continuously for anywhere in the world, or indeed for everywhere in the world, which is what you see here. This is at one time, all the roads and buildings that the machine learning system can can extract. Now, what can you do with that kind of data? I'm going to give you an example in a minute. But before I do, I want to give you just a basic framework for how we think about this. So we're all living through this massive information revolution, there are sensors in the sky, in your pocket, in space, on the ground in the oceans that increasingly are giving us a sense for what's happening in these complex systems. It's actually way too much information to pay attention to. So there's what we think of as the insight revolution which is really about machine learning deep
Learning computer vision and other forms of statistical analysis that allow us to pull out meaningful insights. And if you have those kinds of meaningful insights, I can see where all the roads and buildings are. Eventually, you can build new kinds of indicators. And we really think of these as big indicators, in the same way that we think of big data, that is to say, real time indicators that tell us about the health and wellness of the world's most vital systems in real time in a way that can drive action. And how do we get that action? It's really the last layer here. It's building new kinds of instruments by which we mean Technology Policy, and especially finance instruments that can guide capital and resources to the problems and to the issues where, where they can make the most difference. There's a lot of information at the bottom and there's lots of value at the top. So to go back to that question about what what do we do when we can see say all the roads and buildings let me give you an example of how we can use that to produce a real time climate risk indicator for for vulnerable populations. might not have otherwise benefited from these technologies. Because the satellite system doesn't discriminate between Bangladesh and Beverly Hills we're seeing everywhere with the same resolution and frequency, what we can do is take that daily satellite imagery. And here I'm going to give you an example in monitoring climate, flood risk, climate related flood risk, we can take that daily satellite imagery, we can extract using the tools of supervised machine learning, those critical insights that give us a sense for where all the buildings are in this case and how they're changing month on month. And then when we overlay on top of that a climate risk flood flooding model related to climate risk, suddenly, we can see an aggregate picture of risk for populations and communities that haven't normally benefited from these kinds of tools. So for example, here's a city in Central African public called banking. Here it is in October 2017. This is zoomed way out. We can zoom all the way down to a few meters in the actual imagery. Here. It is. About a year later, here, according to the machine learning algorithm is where all the growth was in terms of new buildings. And here is the flood risk map. And if you look right there, you can see a whole population of people, places where building and an urban growth is occurring in places where people are going to be vulnerable to new forms of risk. So it's essential that we then use these tools to help a whole host of actors. This is the foundational kind of insight that can be used for new forms of climate risk insurance or new realms of climate risk financing, or even for communities to advocate for their own climate rights and for climate justice, which is just as important. Again, it's it's, we think of data as not just a tool for taking action, but also as a tool for building social equity, which is a theme I'll return to in a second. The second example I want to share with you is is using this tool for biodiversity conservation and an ecologic protection. This is an example of a huge project we're doing with our colleagues at Vulcan Paul Allen's group and,
and a group of renowned institutions, including the University of Queensland and Arizona State University. And what we're doing is collectively mapping and monitoring all of the world's coral reefs in unprecedented detail. And we're making all of that data available as a public good for conservation purposes and for scientific research purposes. Just to give you a sense of how we do that, this over here on the left is the current 2010 map that the folks at the UN use to map where in the world all the coral is, we use the satellite imagery, again, this is zoomed way out to take crystal clear images of where all of the world's reefs are. And then we use the tools of machine learning to characterize every single piece of those reefs. And then because we can do that every day, and every week, we can build monitoring systems that show us the very earliest signs of coral bleaching. So that we can de intensify human impacts when they when they really need to be de intensified. The thing that's really important about this is that this kind of work can only happen when you have data that comes from the planetary scale, which in this case happens to be satellites, or it could come from other kinds of sensor networks, but needs to capture the whole system and needs to do it regularly, with data that's also collected on the ground. And here we have some of our colleagues in one of the communities in Papua New Guinea, who are helping us actually do the earth, the grounded observation to understand where that changes, and we're in a dialogue with these communities around the world where they're helping us collect the information. And we're using the tools of machine learning to combine targeted crowd sourced information or targeted field collected ground truth information with the satellite imagery. And the thing that's exciting about that that I want to mention is that these tools can be used when when the right social architecture really is in place really powerful things can happen. In fact, actually, just recently in eastern Sri Lanka, we just had some of the first new national parks named, that were named in part because we had actually helped map them using this methodology. So if we can connect communities on the ground to these tools, we can do really transformational things. The last couple of examples I'm just gonna give you them will will turn the conversation with Marcus and Sasha is some really extraordinary work we're doing with colleagues on climate in particular, I want to mention just these these couple first is an organization called Carbon Tracker. Carbon Tracker uses exactly the same tools to estimate the utilization of coal fired power plants around the world. And what they do is they measure not just the plumes, so to see when they're actually active, but they measure the volume of the coal piles that are next to the plants so they can see those going up and down because coal fired power plants aren't used every day to really carefully calculate emissions and utilization to get a sense of how much carbon how much the plants are being used and how much carbon is being put into the atmosphere as a result. Similarly, we have seen great results using these kinds of technologies to estimate things like pollution in the atmosphere, in particular, emissions that are directly directly intersect with COVID. Right now, dumping Pm 2.5 into the atmosphere, these small particulate matters has a really profound impact on how many people get sick, and being able to ensure that we know where emissions are happening where air pollution is occurring is is an essential part of our global COVID response. This is just one example that actually allows us to see from these plumes exactly which sectors in which plants are actually operating, even though they might have been required to shut down or not, this happens to be happening during the COVID crisis in Beijing. And then finally, on mentioned that in particular in relation to COVID. In general, we always, you know, we're going to have to de intensify agriculture. No, no doubt to achieve our climate and sustainability ambitions, we're going to have to learn how to be more ecologically efficient, and and how to grow more with less. The next billion people that will arrive in this decade aren't even here yet. And we've got to figure out how to feed not just a hungry world today, but a larger world tomorrow. Our colleagues at Atlas AI are using the satellite imagery to produce highly accurate, not just analytics of what's currently happening with agriculture, but predictive analytics. In this case, maize yield predictions that allow people to get some sense of how much food will be available. And in an era right now, in a moment, right now, when, especially in East Africa, we're seeing the intersection of COVID plus the locusts, infestation, creating real potential risk and vulnerability these kinds of tools we use
In a very actionable way, in fact, they already are to help determine how to keep people safe and make sure that they get fed. Anyway, this is just a few examples to give you a sense that, you know, there we can give you 100 more but, but there are, there's a new regime, I think emerging where institutions have the ability to be guided by real time, and even predictive information to make more effective decision making, and, and to take more inclusive stewardship of the earth. And that's a really that's a systems change to the way that we, we typically think about governing the planet and the way we think about stewarding and living on the planet. And, and it's a really exciting time to be thinking about what this means for the future of all the institutions that we work for. Anyway, thank you very much. And maybe Marcus back over to you. All.
Right. Thanks so much, Andrew.
And Sasha. So let's bring let's bring everybody back. Right Everybody ready? I'll find the right buttons. So here we are. Listen, I appreciate both of your presentations. I'm coming to you from Los Angeles, where it is we do have good weather, and we're starting to slowly make some changes to our lockdown, but it's great to hear from you both. Let's get into a bit of a discussion. I see a lot of questions coming in through the chat function and the QA. Remember, use the q&a function off the chat. All these great questions in chat will not be answered. Just kidding. But not really kidding. Use the q&a function. Okay, so I think one thing that jumps out to me immediately from both of your introductory talks is that you're really both focused on visualization of Earth systems in some way. That's a clear link between the two of you. Interesting, broad question, I just want to throw up to you if I can ask you to get a bit philosophical. So we celebrated 50 years of Earth Day this year. And part of the sort of original environmental movement that led to the celebration of Earth Day was one of The first photos of planet Earth from space, which I think was the late 60s photo
70s book. Absolutely. Okay.
And that was that led to, for instance, the little blue dot poem by Carl Sagan and a lot of sort of reflection by some folks on, you know, what does what does Earth mean, in the context of broader society and our universe and our place in the universe? So the question is, now that we have new and accelerating tools for visualizing Earth systems, do you think that there might be a similar cultural impact with being able to visualize either Earth systems, or insights or even some of the projections, like what Sasha is working on about visions of what could be in our changing climate? Do you think that will create a cultural shift? Or what are your thoughts on that? Maybe? Sasha and, Andrew, I'm interested in both your perspectives.
Sure. I think that cognitive biases are a big part of why we're having trouble acting on climate change as a society. I did part of my Studies in cognitive science and it's true that we're wired in a certain way to respond to threats that we see and that are imminent, imminent, sorry. And so this is obviously not helping us act on climate change. So it's not really like a tiger pouncing from the bushes. It's something that's very abstract, far away in time and space. So I think that we're finally at the point where we can use remote sensing and AI to make it more concrete. And I think that there's precedent like you mentioned the blue dog but also I think images of the ozone layer were a big part of signing Montreal Protocol like people finally saw what it meant that the ozone layer had a massive hole in it and it started people you know, there's this action intention gap in a lot of things like even stopping smoking like climate change like you want to do you want you want to stop smoking, you want to you want to you know, stop eating red meat, all these things, but then the the actual action is harder. To to attain. So I think that imagery is really something that can bridge that gap.
It's a great question, Marcus, you know, what you're referring to is, I think, especially is what's commonly referred to as the overview effect. You know, it was a wet when I think it was the Apollo 11 astronauts who, who maybe actually made I think the original pebble was Apollo 17. In the early 1970s. The all the astronauts who went to the moon described this overwhelming feeling of spiritual awe and reverence, not as we commonly imagine, just because they could see the disk of the earth. But because they could so easily put their thumb over that and actually blot out all of existence as they understood it. So there's a sense of not just the beauty of of the earth, but also its fragility and and how easy it would be to sort of cognitively cut how easy it was for them from that vantage point to to cognitively remove themselves from from existence. And that's, that's such a powerful and transformational experience that that they, astronauts who had that experience said, we just want to drag every world leader up here to have that experience. They're all of politics would change. The wonderful book, by the way, I just turned my thing off. I don't know if any of you've seen this. It's a book by by this guy named Benjamin grant. And it's just filled with pictures that are designed to encourage the same kind of reverence, reverential moment. It's a it's a psychological and spiritual effect. And I think in some ways, there is a unique opportunity to bring that to the ground now to bring that to to all communities. And I think such as exactly right we are problem is one of cognitive bias that, you know, human salience is a terrible proxy. For whether or not we should conserve species, or change our emissions, or grow in a particular way, or engage in a certain set of practices, because so much of what's so damaging by almost because it's not salient just falls out, it falls outside of our perspective. So a critical way of, as I say, like seeing change as a way of encouraging stewardship, right, it's it. And so I think we're gonna see a lot more of that with these technologies.
Okay, let me just pick up on that exact theme. Something we're seeing a lot of in the culture in the press, is that people aren't people are noticing changes. Not everyone's had the opportunity to go to space. It's something I'd love to do. Let me digress for a second. Our CEO of XPrize, a new shot Sorry, I remember we were at a staff meeting recently and somebody asked her to reflect we were celebrating the anniversary of her trip to space, she spent time on the International Space Station. And someone asked to reflect on that experience and she brought up that effect that overview effective sort of not just a shift in literal perception of the world, but also sort of a shift in philosophical perception. As a result of that vantage point on Earth, many people are noticing things like rewilding bird songs are louder, the air is cleaner, traffic's down, less noise in the in the ocean VISTAs that we can see for the first time. There's a lot of discussion through April as we get into Earth Day about, wow, people are really noticing what the future could be or getting a glimpse into a different way of living. Curious, do you think this is a galvanizing moment? Do you think it's just a blip that will pass in particular, do you see an opportunity for visualization and learning based on let's say, machine learning or computer vision to move forward in this moment?
Sasha, we'll take that.
Um, I was skeptical about this because we still think it's an external effect. So I think most people, you know, we didn't, most people didn't choose to be on lockdown. So we see it as an externality like people, someone pulling stuff to stay home, right. And then this happened, which is obviously positive. But I don't think people see it as their their impact. So it's not it's not like as a direct consequence of your actions. It's indirect because someone impose this. And then so we have, for example, cleaner air or birdsong. So I think that I mean, people still have this disconnect between our impacts on the environment. And we often underestimate how much power as individuals, we have to change our environment. And so maybe this is a first step, but I'm not, I'm now skeptical that it's going to really help people be like, Oh, yeah, I'm gonna stay home all the time, and then see the Himalayas, right. It's more like, this is a temporal, a temporary blip on the horizon. And then when we go back to real life, it's gonna be like, Oh, well, that only happened because someone forced us to do it. So I think that people should start being more aware of what their options are as individuals on the environment. And that's not necessarily visual, sadly.
Hmm. I'd love to share a quick observation with you, you know, I'm not gonna I'm going to You're a slide guy. I didn't know I was going to do this Marcus, but I'm going to pull something up just to share with everyone. You know, we talked about how COVID and I'll just say it at the top. I'm a little more optimistic than Sasha is about this for a couple of reasons. But before I get a show that just want to show you one thing, which is that we talked about how COVID this moment in which there's been all of this fundamental changes, a kind of site seismic moment, we, you know, we'd love those kinds of metaphors. In this case, it literally is a seismic moment. I don't know if you can see this. But the earth has literally quieted as human activity upon its surface has diminished all over the world, people who are seismologists have noted this signal of the reduced human footprint. So and I think there's a there's a number of reasons to suggest why there might be new political movements and new political opportunities. First of all around the world, people have seen with shock and awe what it means like for there to be clear air. And we know that the effect of air pollution on COVID is really significant. And and the question is on the other side of this will we make will we force people to make choices between, say their health and their work between between health and in between their lungs and their industry. And I don't mean this. They're their traditional industries. I mean, just the industriousness of working on the earth. I think the other thing that we we've come to an appreciation of is that COVID in particular is not a natural disaster. It's an ecological disaster and as an ecological disaster, it's rooted in the relationship between human beings and the natural world. And, and, in particular, most of these novel zoa notic illnesses, the ones that jumped from wildlife into livestock and humans and, and go back and forth. The flashpoints are where he Human beings are putting pressure on the landscape where where we're actually changing the land. And there will I am certain just as a matter of global public safety comm pressures on governments around the world and on communities around the world to de intensify their their impacts on these wildlife habitats, in part because it's just not safe to do so. So the combination of kind of emergent political forces, the real world experience of actually being able to breathe deeply in downtown Delhi, or in downtown Beijing for the first time in decades, and an understanding of the intersection of the way these complex systems works, I think they create a new opportunity, at least for me, the fact that we've stopped now it's it was the last piece we stopped, and we're going to have to make intentional choices about how to restart all of those things come together in a way that I think if we're quite careful, we can we can align to bouncing forward because opposed to just bouncing backward to where we were
Okay, great. I appreciate that. A nuanced answer for us all to think about. Just a throwaway comment. I'm interested if the, the unwinding of these effects in the short term will be reported in the same way that sort of the economy reopening has been reported or the the revelation of cleaner air and quite a bit song and less vibration and things would be reported, something we should watch out for the next few months. Okay, I'd like to shift into a slightly different theme,
you've guys have presented or you both presented really compelling and interesting visions of applications of AI and ml to not just observing Earth systems, but actually putting that insight into Practical Action. And they're rooted in things like the SDGs environmental sustainability climate. I think that's pretty clear. I want to point out though, that I think for a lot of folks, AI and machine learning are maybe new technologies that may be a little bit intimidating. And they certainly aren't part of the traditional environmental movements toolset. So it's it's almost like the tech community is coming at this from a different perspective, even though we're all marching toward the same goal. So I'm asking if you think that there truly is a cultural divide between quote unquote, traditional environmental movement, and let's call it the AI and machine learning environmental movement, or do you think that's just a misnomer? And there's a different way we should think about it? What are your thoughts on that?
No, go ahead.
It's a great question, Marcus. Let me let me start by acknowledging a fundamental problem, which is that and there's a set of connected problems here that you're pointing to. The first one is that the tech community writ large, has sometimes sort of the experience of people who are working in in social and ecological stewardship. On the front lines, you know, it can be a little bit like the UFO lands in the front yard and people walk out and like we have the tool, right, you know, like we come from the future and here it is right and and it's designed for and not with the people who are intended to use it. There's no sense of social pole, that it's like we've developed this in relative vacuum. There. It happens in the context of, of long standing issues of colonialism and neocolonialism in which the many of the centers of excellence for these technologies come from the north and the West. And so there's a whole set of questions around around how the social relationships between the people who are the makers and the people who are the users and beneficiaries and stakeholders of these technologies. Many of the, one of the critical challenges we have is a problem of, of what what I think of is sort of Death by projects. And what often happens and there's like so many tiny little examples of the of these projects that that end up being extractive That is to say, you show up and you eat up a community, you're trying to serve attention, and then you extract data and you return leave, frankly, very little value back in the communities, and then often the funding dries up and the attention moves on. So I think it's not a function of whether these technologies can be helpful. There's no question they can be because we we have institutions as a consequence of all this and communities that are trying to solve 21st century problems with 19th century tools when they could be using more effective tools in some regards. But it's about the power and social relationships between the people who come together around these problems. It's about the institutional framework that that privileges big multilateral institutions over people who are living In these communities, I'd much i'd as much like to work with the, the local indigenous community as who's actually stewarding the land, as I would the kind of big nonprofit in the north or the West, in the global north, you know, and so so we have to figure out new architectures of participation and social equity. But I think the gap is closable. So I'm optimistic that is closable. But we have to go at the problem, which isn't tech, it's people.
I totally agree with Andrew. And actually, the more I learned about the different kinds of projects that exists and that exist in ML and climate change, in general, I feel that the most successful projects, the ones that make the most impact concretely are those that are on the forefront between ml and domain experts. So I mean, I have seen a lot of these like pilot projects that are super cool, and then you know, bring together different datasets and then sometimes also invent a problem that they themselves solve, which, you know, sometimes gives us a bad rap. Like, I remember meeting with some stakeholders who climate scientists who came to Mila, and then I showed them some of the work that people are working on. And I mean, they kind of laughed, they were like, yeah, this is cute, but how about like real problems? And then I'm like, Well, what are real problems? Like, let us know because of course, everyone's kind of, to the extent that they know a field you're going to try to make or mean to do your best and to deploy your your knowledge but until you have someone in house who's telling you know that that's, that's silly, don't spend time working on this like, like, what's priority? What's the priority is this and we don't have data for that. So how about you help us gather data instead? Like what Andrews doing right? And sometimes it's like, what you think is a problem is not even a problem is just that you're projecting your own kind of either colonialist or just, you know, just naive kind of notions of what you think are problems just based on your own experience. So there's a lot of that going on. But that's why we advocate on climate change. Like work with the experts connect with the experts reach out and and you know, the projects that are gonna go farthest are the ones that have the most diverse and complimentary teams. Of course, if you stay in your ivory tower of ml, it's hard to make a real world impact. If I
could pick up on that just for a second, because I think Sasha center have a couple of really important things there. The first one that I just want to point out is, is we there? First of all, there are amazing organizations that are working with communities and with people around the world. And I, I notice a funny cognitive bias among people who develop these tools, which is the when we describe the engineers and the scientists and the technical folks are working in this we refer to them as people. Like Yeah, you work with the people who know what they're talking about. But then we use this whole set of other distancing language when we talk about the people that are the they're also in meshed in a web of mutuality with us. We Are the people that we want to use these tools, they become stakeholders or beneficiaries, these words that sound like they belong in insurance policies. And if we just flipped them around and said, you know, we refer to everybody as people, if we refer to everybody, if we think in community terms, and not just institutional terms. If we get like, I notice in my own work, Sasha said, this one little thing, and Sasha, this is not I'm not picking on you at all. It was great that you said this. And I believe everything you said, I agree with everything you said, but use this word, you know, in your house, right? It was just it was just a little throwaway phrase. One of the things we should think about is just this question of like, social distancing, we have problems of social distancing, under normally good times, which is where to socially distant from the people in communities, like, one of the things that we're working on is like, let's get into the field. Let's go live with the people you want to serve for a bit. Let's actually live in, let's build some social solidarity. And then think about What happens and that's a different way of of saying what Sasha said, which is, you know, you like the, we're solving a lot of problems. This is one of the one of the biggest problems is we have a lot of people who made a lot of money building AI can openers or dating applications, or you name it, and suddenly say, Well, I can apply this to this problem. Because because they're naive, and I think it's mostly naivete, their naive view is that this is the way the world works. In fact, actually, it's not even remotely how it works in most most instances, so we have to close our own social gaps before we can begin. It's a little bit like close close the social distance before you start building. I guess this would be my advice.
Yes. Okay. I hear where you're coming from. I want to probe on this topic just a little bit. I'm gonna throw in one of the questions we got in the q&a. Are we affecting anything with all these sensors or just building a gigantic cage? To me, I think this is part of a broader theme of are these just shiny new tools that The people to us kind of that language are interested in or is there something fundamentally new and different and better outcomes that we can expect from these things? And I think also I'll just add, I think it speaks to, you know, sometimes the history the track record of technology, increasing our broader environmental sustainability is mixed. Some people see it as really, really helpful. Other people see, technology, technological advancement is something that actually takes us backward in that depth. So do you think machine learning and AI fall into that trap? Curious, your thoughts?
I'll speak from my self first and I'd love to show one something that came to mind as you were as you were asking the question. First of all, there's no question that we build a lot of things because they're buildable and not because they're necessary and because they're useful
and and we live in bubbles of our own Ignorance was was discussing a minute ago.
So there's a lot of wasted effort. And there's a lot of things that were built because they were what people knew how to build. And we need a bigger social imaginary, to have wider variety of people who are participating in the acts of imagination that drive, what could be. That said, there is no question. I mean, we see it every single day that these tools can, can but are having really profound actual consequences on the ground and in communities and I want to just say, I think they're like, less than one 100th of 1% fully unfurled, so we are just seeing that the beginnings of the 10 roll shooting through the ground, we're not seeing a fully grown tree, or anything like that. But But you know, the reality is that we know that if you put information in the hands of indigenous communities They can advocate in the context of, of systems that are often aligned against them for their land rights, for instance. And we know that one of the most effective ways to ensure conservation of biodiversity and intact forests is to ensure indigenous ownership and stewardship of those communities. So these tools are powerful in their evidentiary role in empowering communities. And, and we say information is power, big information, big data is big power. If we just give that power to big institutions, even if we made it free today, if we didn't take the extra effort to make sure that it was in the hands of everybody, then what you'll do is if you have in a society, you have two groups, you have the, you know, this group level of social powers here, this group's level of social powers here, and you you throw a bunch of free data at them, this group becomes this much more powerful and this group becomes this much more powerful and you've actually increased the neck and equality between them. So we have to make sure that we take the action extra effort to ensure that we create more balanced forms of social power. And, and but we see this actually happening every day we see it happening with real communities and NGOs. And and one of the most important things about these tools is the last little bit of this answer is that they represent a, an independent source of truth. Just think about how embattled the truth has become, in the last, you know, five to 10 years in particular, how suspicious we are of global media, how suspicious we are of storytelling and of evidence in general. This is a new form of evidence. It's difficult to tamper with, and it enables people to say, okay, we're having a political argument, but this is what's actually going on. And so, I could give you countless examples of where that's happening today, with COVID. With, with the conflict between nations with large institutions engaging in malevolent environmental practices on and on and on. So you know, We have to turn AI turns data into social power, if we think about it, in those terms, and if we if we do it right.
And just to build on what Andrew said, we did a hackathon at some point that was like AI for Good. And then we went out and we were like, are not going to be one of those. One of those people were going to go and reach out to NGO. So we went around Montreal, meeting NGOs, and asking them, like, give us your data, how can we help you? And then most of the answers we got was what data? And can you give us lessons on you know, how to use Excel, or how to set up a like a, you know, a management system for our inventory and things like that. And then we realized that, you know, we wanted to use AI to do cool stuff. And then, honestly, what would be a bit like that, maybe five, five years down the road, we'll be ready for that. But now we should start doing instead like, just just tutorials about what data is or how it can help right and we ended up doing some workshops in Things like that. And it made me realize To what extent like a lot of these especially the on the ground deployable solutions that have people who are supposed to, you know, take them and do something with them not not like not necessarily a sensor or something that can be fairly independent, but solutions that essentially hinge upon people using them. That that we're not we're not there yet. We're not there there yet. So that the, you know, the vast majority of people are okay with taking a neural network and being like, Hey, I'm gonna use this in my daily life. Right. So I think that there's a lot to be done. We have to prove ourselves first. Okay.
Okay. Appreciate that. I think there's so much more we could say in that topic. There are a lot of great questions coming in on the q&a function. Just a reminder to people that you can submit your own question or you can upvote or thumbs up, click thumbs up for an existing question, if you love it. I'd love to hit one more theme. Before we get into some of the audience q&a. And this is we're having this conversation in the midst of a global pandemic. Most of us have never seen anything like this. First question. What's the conversation inside these communities about application of these tools specifically to COVID-19? Whether it's the health aspect, or the economics aspect? So I realized we're going a little bit outside of the theme of environment. But also, I know there are a lot of close links. So what are your thoughts on that? Is this movement? Is it already happening?
It's definitely happening. We wrote a chord a paper about AI and COVID. I mean, how can be used to help COVID and we identified actually three scales, which I found are really interesting and kind of defined it on a on a good, you know, succinct level, there's the molecular scale. So AI is being used for vaccine discovery. So molecular discovery, and essentially, you can explore a bigger space of molecules using AI. There's like the clinical and patient scale where you can, for example, analyze CT scans, you can predict patient outcome using more data and things like that. So that's been used already in hospitals. I know that hospitals for example in Montreal and across the US are using these kinds of AI systems in order to essentially flag patients that might need ICU capacity, for example, and then there's a societal scale. So there's the actual epidemiological side of it. And then there's actually another, they call it the infodemic scale. So there's actually so much information and news that people are overwhelmed. So the who is having a lot of difficulty in keeping abreast of this kind of wave of misinformation. So on these on these three scales, identified all these different applications, and I think that, you know, the, it all hinges upon adoption. And so for example, how do you get a hospital to use a tool, given that it's a new, it's a new, like, we've never seen this kind of virus before? I mean, in our lifetime, so how do you expect them to trust you with your tool, right, and where's the data coming from and things like that? So I think it in terms of opera operationalization we're not there yet. But conceptually speaking, there's a lot of things to be done.
I couldn't agree with you more Sasha. I really think that points about infodemic. In particular, I really wish we should talk more about that in this session.
Just a quick thing on on COVID
geospatial data, and AI and machine learning models together have relevance for four categories of use cases for four kinds of things in which they can be helpful. So the first one is is about modeling, epidemiological modeling of the spread potential of spread epidemiological modeling of the risk of importation into communities. So, those are all proxies. And in fact, actually geospatial data is used enormously in spatial epidemiology today already to help to help model that and we are helping and numerous organizations that are working in academia logical modeling where we're, we're both working with an exploring new ways of helping them with with us and in real time. The second one is around monitoring for risk and then There are risks within the system that is the spread of Coronavirus. But it's important to note that how many one of the signature effects of COVID and any pandemic is the ways in which it amplifies risks that are not related. They're ancillary. So we give you three examples. What What is that in California, we have these enormous wildfire risks, right. And we're huge. You can have enormous, you know, not quite Australia scale burns, but really, for our for the region, they're really significant well, because of the shelter in place effort. The in California, a bunch of prescribed activities that had to happen in advance of the fire season didn't occur, because there was nobody to do them. And so the risk of wildfire goes up during COVID. In Sub Saharan Africa that no one has quite figured out the relationship between say malaria, and TB and COVID risk we assume that there are negative but there are complicated interactions. In the public health space that we don't know, and then it for in general for Africa more broadly and many other places in the world, they're going to be these issues of food security, which you know, our colleagues who who study this look, they use words that are biblical in terms of scale. So they're monitoring emerging forms of risks, both within the particular epidemiological path and other forms of risk that get amplified as a consequence. The third one is actually looking at where the recovery where resources are, and connectivity are so you can plan response. And the last one, which we hope comes soon, you know, as soon as possible is going to be about how do we measure the recovery? And how do we begin to see what it looks like for the system to come back? data and AI are going to play a role in all four of those places without question in really significant ways.
Okay, appreciate that. I'm just going to note in the chat
participant called Kishore put up the collected a resource list for fighting against COVID-19 as well as a review paper landscape of AI applications against COVID-19. So everyone can check this out. Okay, great. It looks like looks like that's a preprint. It's on archive.
And there there are awesome questions in the chat. I just say these are great. Yes. Yeah. If there are any. Wow, guys,
I'm scanning through them. I can't say that out loud, because we're supposed to use the q&a function. But yes, make sure the chat
looks okay. Good, good. Go.
Really quickly to the both of you, do you perceive that COVID-19 is going to dramatically change the field of AI machine learning, or do you think it'll have a you know, just the tail winds will sort of continue? Is this a pivotal moment? Or is it really just something different?
I think that it's going to make us realize To what extent we're not we're not quite there yet to make to, you know, to reply to react fast. So I think that if we had a bit more time, a bit more, I want to say a bit more pandemic, but like I mean, when the pandemic kit, we weren't ready to say, okay, we Yeah, we've got these like AI tools we can deploy in hospitals. We've got, you know, all the levels set up. Yeah, we're gonna help the WHO filter all through all the noise. And so we're ramping up, but but I'm hoping that will make people realize To what extent we need more integration, we need more foresight essentially.
Hmm. No, no question. We definitely need to yoke strategic foresight to the application of these tools. And so this is a moment when I think you've got two things happening. One is we've got these very powerful tools, and we've got high degrees of uncertainty. And so how we think about using them is something for, you know, all of my friends who work in strategic foresight and long term thinking need to be engaged now, even as these tools get applied. Now, in real time, we'll just say there's no question in my mind that that this period is going to drive an enormous explosion in The use of AI and machine learning for a couple of reasons that are ancillary to the ones that we're talking about here, the main one being that we're going to see an enormous push for automation, automation grows, the investments, automation, sort of counter intuitively grow during periods of recession. In fact, actually, in part because we're going to have this very uneven effort at recovery, which is going to be very spotty, and entrepreneur, you know, improvisational, but but over the last 30 years, for instance, if you just look in the developed West, every time we have a major recession, that's when we put in spurts of of innovation around automation and labor replacement, which is itself an enormous set of questions beyond the scope of this discussion today. But I think we're going to see those activities accelerate both automation, machine learning and an AI applications. Especially because we've got a problem just to think about it very basically we've got a problem that's bigger than us. And we need and we've limited, we've limited the number of people can work on it, because we're all sheltering in place. So we're going to have to use tools to fill in the gap. And so I would expect it will be a tipping point.
Okay, terrific. As if on cue somebody wrote in the chat, when are we going to get to the excellent questions from the audience? The answer is right now. Everyone go to the q&a. If you haven't submitted one already up, vote your favorite. I'm going to start at the top, according to upvotes. And then I propose we just blast through as many of these as we can. I'm noting at the top of the hour, it's possible some folks have to drop but we do have another 30 minutes to chat. Okay, this one's for Andrew, and Andrew Sasha, I'm going to ask you to be succinct in your answers so we can get through as many questions as we can. Andrew, what is the business model who pays for this who receives the information?
planet is a private enterprise but we're permission led private enterprise invested in by triple bottom line investors in the World Bank, among others in Silicon Valley, And we we sell data to organizations and analytics to organizations, we also provide it under certain circumstances that really dramatic discount. So I think I would say is our business models a little bit like an airplane. So there are some customers who are like sipping orange juice and drink a champion in the front. And there are some people who are, you know, backpacking across Europe who are sitting in the back, but we try to get everybody across the line. Some people are, you know, our frequent flyer, flyer, so the goal is, is always to to serve everybody as to the greatest extent possible, irrespective of their their cost, but to find a solution for everyone.
Okay. For Sasha, is the making climate change visible website where you enter an address, is it live, or is it still under development? And if it is live, can you share the URL?
It's not live yet? We are testing. Right now we have two pain issues. Essentially we're working on the front end because None of us are, I have absolutely no sense of taste and design. So we're trying to figure out what's the best way of presenting it because the machine learning part is done. The climate modeling part, like essentially translating climate models into something that we can use for generating images is mostly done to that. So now we're focusing on actually like making the message coherent, you know, once you see an image, what are the things you can do? So we're trying to make that really impactful. And and we're aiming for essentially the fall, ideally. So I will share that when it's available.
Great. Okay, stay tuned.
The next couple of questions are similar. So I think we can take them together.
I think it's for Andrew, but really, for both of you, is the kind of data you have showed and shared publicly available. And then there's another question which says, Can you open source the data, because this will empower everyone to proceed and test great ideas?
I think let me take the latter half of that question. You know, the The let me just say a couple things. First one is, we try to the greatest extent possible. One of one of our kind of signature hypotheses at planet is is to try and ensure a very broad access and democratic access to information we were founded to democratize access to these tools for everybody. So in many instances, like I'll give you an A good example is that the Ellen coral Atlas that I mentioned before, we're going to be producing the highest resolution, highest quality map of all the world's reefs by the end of this year, and that data is open source for government conservation and scientific purposes. It's limited in terms of its rights for commercial application, because we hope that commercial actors will help. We can commercially licensed that data in a way that allows us to continue to service the conservation aims. We're about to learn Something in the middle of this year called the California forest Observatory in which we're going to be producing the highest resolution map of where all the fuel loads are in California because of the forest fire issue, an effort, we also have to scale also under the same terms, this is an open source from for most of the, for all the scientific and non commercial purposes. But we create an economic model that allows us to put forward funds from from commercial activities. Okay.
Do you have anything to add to that? Sasha?
No, we don't have much data that we produce as such. But everything we do is up on GitHub, and we use existing climate models which are open and out there so anyone can access them.
Okay. Thank you. Another question about making this data actionable? So this is a great question. We've talked about data. We've talked about an insight layer on top of the data, all extract sort of abstract question to say how do we make the kind of data and intelligence actionable for presumably policymakers, local leaders, Other people? Okay, can I start? Go for it?
Um, so we've been thinking about this a lot of climate change. Yeah. Because so I'm a content chair. And essentially, we have all these great datasets and resources and primers and tutorials and what have you. And then we've been thinking a lot about, but like how to get people to use them. And essentially, we've tried to figure out kind of two use cases. And essentially people who are, who know exactly what they want to do. And that's fairly easy, like we're gonna have a relatively structured resource for that, but also want to bring, bring, like ask people to come forward, like people from actual domain experts, like for example, energy, and to propose a data set and a challenge kind of like a kaggle way but without the competition part, but just like I'm working in an energy utilities company, and now casting for example, predicting solar energy out of production in real time or in a few minutes intervals is a huge is a huge problem. For me for Example. And then here's the data and, you know, do whatever you want with it. And and we think that in this way that if the problems are actually formulated by people who know exactly what's going on, then we have, that we can get people involved without having this kind of pilot project issue of you invent your own problem and you solve your own problem.
Yeah, that's great. Sasha, you know, a couple of a couple of thoughts. Let me let me break it down into into three ways in which we make it actionable, what's the path to actionability? Right. So the first one is about access, and it was asked earlier about licensing rights and things like that. I mentioned that planets data is relevant to measuring 13 of this, of the indicators and targets under 13 of the SDGs. We've started a program where we take philanthropic funding that covers 5% of the cost, just the transmission costs for a planet to take the data out of our system, pump it through The UN statistics agency so through the un un Statistics Division that aggregates all the national statistical offices and makes that data available about the 47 least developed countries accessible to anyone who wants to use it for SDG measurement. Okay, and and that's the kind of mechanisms that we want to see happen more and more. And we're trying to set examples there for others to follow. The second thing is about the shape of the data. And one of the things I find interesting about this is that I think a lot of this question about how do we make it actionable is the wrong question. Because actually, people don't want to use this data to make decisions. They actually what they really want is to be able to ask a question and get an answer. How many ships are in the harbor? How many hectares of deforestation have we seen? That's a question you can answer you can ask in plain English and the answer is a number. It's not a map or visualization. We spend a lot of time talking about visualization because we're trying to prompt people to urgency. But the reality is for utility, what you really want to do is get the answer down to like a one or a zero. Is there a thing there? Is that? Is that deforestation or not? Is that habitat loss or not? And those are yes or no questions. That's the ultimate thing. And that really relates to the third category, which is about the data shape. So we need to take these voluminous terabytes of data, we planets generating terabytes of data a day, every day, and we're storing it forever. But no one wants a dump truck backed up to them. And no one thinks about this when they go to Google, right? No one, no one says, How do I utilize the data with Google because they they index the Internet, and they put it behind the world's easiest use search. user interface and now everybody uses it literally thoughtlessly. And that's a that's an interface question. So we can reduce the shape down to like simple yes or no questions, and we can put it behind a world you know, world beating simple user interface. These questions about utility, I think will solve themselves.
Okay, really interesting. I'm going to try to synthesize a couple of other questions that are coming up. So hang on, I'm gonna throw two at you. One is based on the previous discussion. Do either of you have specific examples where insights generated in this way have influenced money? Specifically venture capital or government budgets? It's one question. Second question is about data sharing. So where there's a lot of conversation about global data sharing and sharing of data sets, but there also seems to be a trade off or even a contradiction between monetizing data and open sourcing and sharing maybe data or insights. So open sourcing, monetizing, sharing of data. And then the first question was, is can you think of a specific example, where these kind of insights have resulted in changes to the way venture or private or public money is spent?
o'clock. Maybe I'll take that one first. Just very quickly via that there are countless examples we were aware of where, where this kind of information, the vote where both the velocity and accuracy and relevance of information drives a change in behavior. And sometimes that behavior is related to finance. I mean, I'll give you a really specific example and it comes from Cameroon where there was a very there's a very large multinational company. That company operates through a subsidiary, it operates a very large rubber plant. The plant is literally so it's a it's a plantation where they're growing rubber trees and producing rubber. There they were engaged in in aggressive deforestation in ways that displaced both local communities and put pressure on habitats for a couple of days. animals including the lowland gorilla, and which is a critically endangered species. And as a consequence of this, they were divested from when the evidence was collected through this kind of analytics. They were divested from from the Norwegian pension fund. And instantaneously, they announced within a matter of weeks, that they were to change the leadership, they said, we're putting a sustainability plan in place, and we will commit to complete cessation of all deforestation activities. And that is continuously monitored by the very same satellites. So we know now that story is not complete, and it's not perfect, it's not over. But the ability to see capital change flows like that very rapidly. And to create consuming behavioral change, which can then be monitored transparently in the supply chain is happening all over the world and it's going to become, it's going to be everywhere because not only does the organization get to see what's going on On, but everyone else does Greenpeace, all the social actors, the UN, all those other entities, journalists, all have access to the same tools of transparency.
Hey, you want anything? Sasha?
I think Andrew covered the the first part of the question or the first question really well, in terms of open sourcing, I think that what what would really, really make a huge difference in terms of remote sensing, especially is having a single data lake that brings together different types of imagery, kind of like what Andrew mentioned, but also what if we could get LIDAR, which would if we could get, I don't know, infrared? What if we could get like everything and then superimpose them actually, using AI. There's machine learning techniques that are good at matching up, for example, you have a satellite image of, I don't know, for example, a factory and then you've got their emissions, and then you've got, you know, a heat map, and then you've got all these other things. And if you superimpose them, you can do really, really great things, but it's actually really hard because, you know, it's not it's not always like satellites can have different resolutions, satellites. have, you know, viewpoints and things like that. But if we had to use ml to superimpose them, and then to enable people to use that data, that would be huge.
Hmm. Sasha, I totally agree in it, Lucas. So Marcus, before we go to the next question, can I just, I just want to pick up on those one thing that in this thread of questions about open source, sir, I don't want people to look at me and think like, some Silicon Valley type. I'm trying to liberate as much data as possible in my daily work. But I do think that sometimes we get stuck in a trap between thinking that there are only two choices, that there's there's public open source data, and then there's private data that isn't shared. And that's not true. And one of the biggest challenges we've got to solve is not actually how to liberate all the data and make it public because the reality is that there are lots of datasets that can never be made public because they're, they contain sensitive information. Because they have commercial value. And I'm not talking about planet I'm talking about, there are countless other organizations that fit into this category. Facebook is a good example. But what would be really powerful is not instead of focusing on on, we should continue to focus on open data where we can. But in addition to that, we should make it a rule that every major corporation has an a, an open API, where if you have a question like Like, for instance, Facebook can answer all kinds of questions that that are socially where they don't want the data to escape, but they might be able to give you the answer, give you an indicator. And so like the ability to ping them on what's the level of gender inequality in a particular place where there's a lot of Facebook users, they could potentially calculate that kind of indicator inside of Facebook and make it available on demand so that you could programmatically plug it into other things. It's a sort of API for the earth. And I and to me, a more nuanced answer. To get a conversation is how do we drive the behavior that respects the need for some data to remain either private, proprietary or private, but still pull out the extract, extract the insights that we all need to steward the planet in a better way to do the very kinds of integrations that Sasha was, was just describing.
Okay, thank you for that. I feel like that is a topic we could also continue to go deep on.
But not right now. Sasha, this one is for you.
Let me find it. Have you been approached? Or would you be open to being approached by organizations that want to use your later stages of climate change images for commercial purposes, for instance, negotiating real estate approval of architectural design or building permits? And before this, will it be used in educational efforts, for instance, in New York City's climate Museum,
family are definitely trying to target educational efforts like we see this as an educational tool. have been approached by by insurance companies. Um, but I feel like I feel like it would be I mean, of course, all the the website is gonna be freely accessible. And and I mean, there's gonna be an open source license to it, that's not the issue. But I'd rather this be a tool that that anyone can use and then anyone can get access to any address any projection because that's really the idea. The idea is like, okay, you search for your own house, but you also search for your friend's house, your school, your work, and whatnot. And during this process, you learn so we want to have an overlay of the map to see different you know, risk zones because they change and it's really a more of a tool. So I think taking the image and then using this as an illustration, for example for for an insurance claim. I think that's a bit like it's reducing the potential of this tool. So for me, it's all about all the rest of it, not just like the AI generated image. Yes.
Another question. There are two that are similar This is a good one isn't the immense computing power needed for deep learning computer counterproductive for the environment? Despite traveling meet computing missions are one of the most co2 heavy emissions?
Please go for it.
We actually wrote a paper about this. So it started with this article that came out last year that said, neural network emits as much as five cars in their lifetime and meet a lot of my colleagues were like, That's impossible. And so we went through the article and they have a point in the sense that now there's bigger and bigger models, especially in natural language processing, that will take like, you know, the whole of Wikipedia or I don't know all of the European Parliament transcriptions and then and then run a model on it. And that kind of model Yeah, can can definitely consume, I mean, produce a lot of co2. But the real I mean, the day to day kind of more small scale. Ai efforts definitely don't produce that much. But and essentially, in order to dig deeper, we created an MLM missions calculator, which was also online and based on so much People run AI on the cloud. So I mean, some people have their own private infrastructure. But that's relatively rare. Most people use some kind of cloud infrastructure. So we based on the the place where your server is based on how what kind of hardware you're using, what kind of GPU for example you're using, and how long you run it for, we give an estimate of how much co2 is emitted. And essentially, we're trying to push that people will, will add this to, for example, research articles or code that they publish, saying that, you know, a run of this neural network on this architecture on this infrastructure in this place, and it's this much, and essentially, it's kind of like being aware of something is the first step to reducing it. And we hope that by sharing this and then by asking people to, for example, open source their models, so not everyone has to train from scratch, especially for these NLP models. You train once and then if you give that model to someone, they can reuse it or fine tune it a little bit, but it's definitely not as as as co2 emitting so so we're trying to do To generate this whole, we're trying to actually we're working on creating a Python package that someone can just load when they're running their AI, for example, neural network code, and it's going to track their energy consumption and then translate it to co2 equivalents at the end and say, like you ran 12 experiments, and at the end, given your location, this is how much you've generated. So we're raising awareness.
That's great. I, you know, I know that there's real, what we're seeing in this sort of the subtext of this question is that, on the one hand, the use of these Machine Learning Tools is exploding. And so we're going to see more of it, and they just consume energy, any amount of energy, right, either in the training and sort of the manufacturer of the models or the running of the models, which are obviously very different levels of energy intensity. I'm a little bit more bullish, not much. I mean, I think we have to watch this really carefully. But you know, we have organizations that are now Running wholly off grid data centers, think about the folks at DeepMind, who helped who were unit of Google, who applied the tools of AI to the energy efficiency of these data centers and compute centers and figure out how to reduce their energy utilization by 40%. That is to say these tools will be used in their improvements of their own efficiency. So I think you know, one of the things to see is that it kind of at an industrial scale, lowering the cost lowering improving the energy efficiency of AI, because both renewable solar energy and the computing architectures themselves are becoming broadly speaking more energy efficient and more powerful over time. I'm, I just think that this is a conversation we sometimes it feels like a trap conversation like things are going to get better and they're not perfect, but but I think the net value You, especially in the overall context is quite good for these tools in terms of their application.
And just to add to that, if people are interested in this, there's a concept called Jevons paradox that is interesting to read upon. So essentially, if you make something more efficient using technology, people will use more of it. And so you'll kind of reduce the efficiency and and that's, that's as part of any technology, including AI, and it's a trap. Like Andrew said, it's kind of a circular thing. You make it more efficient, so you do it. So that is that we're definitely
Okay. Here's another one that's directed at Sasha, is there any scope to also paint a picture of hope? I'm not sure that any of us know what good looks like when it comes to visualizing what our planet has the potential to look like if we do restore it back to full health. I wonder whether the opposite visualization would also add value and create a different narrative around climate change. What do you think?
So one of the things that we're trying to do on the front end is create like a slider so that we can represent different essentially rcps like our projections, so essentially, those are businesses usual, there's, you know, some, some progress is made. There's the Paris Agreement. So you can actually quantify this and tons of co2. And so essentially, we're going to start with the business as usual scenario. So like, if nothing changes, if no one does anything, and we continue as is, this is what it's going to look like. But we want to really let people see, okay, well, if everyone does what they promised to do in the Paris Agreement, what if right now we stopped all emissions? And actually, you know, because there's, there's a certain level, I mean, even if we stopped now, there's certain changes that are already happening, right? And so we're trying to make it a more of an empowerment tool, and especially focused on individual action as well, things that that people can do and then multiplied if everyone did this, the impact that it would have and the accompanying image. So we're really focusing on that. It's true that if you just give people one image and no context, it's just gonna be super depressing.
Yeah. So so you know, one of the things that I'm reminded of in terms of kind of this is just my response to your amazing work is that this year, the winner International Energy Agency has has forecasts that global carbon emissions are going to drop by 8% globally. And one interesting thing about that is it shows how little relative individual personal contribution matters to the overall system, we actually need this wholesale systemic change, you know, all of the pain that we have encountered, both economically and socially and obviously, medically, amounts to about 8% of the total. The UN Environment Program and and the related entities have told us that we need to reduce carbon emissions by about 7.6% every year for the next 10 years. So we need to do this every year and not just keep it at this level, but actually bring it down even further from here. And so the ability to imagine what that looks like to see what that really means i think is going to be essential for the galvanizing action toward a goal where we don't burn the future to the ground before it gets here. And so like the moral imperative of your work, it really resonates with me in this in this realizing that all this pain, you know, COVID did more for the climate than all of our arguments in a single year, probably for the last five years it you know, at a planetary scale. So we can't go this way. We have to go some other way. Right? We have to get there through positive action and giving people an alluring vision of the future. And I I would love to see how you guys would do that. That's just incredible.
Here's an interesting one.
Oops, sorry. I lost my train of thought here. The question my question queue.
Bear with me.
Okay, here we go. Sorry for Andrew. This is a quick one. Have you published some measure of climate vulnerability or land use changes which compare, for example, each nation in Europe or each nation within Africa or the states in United States?
Thank you for the question. Whoever asked it, it's, it's absolutely an aspiration of our organization to help the production of those indicators of land use change. And, and and what have you. I'm particularly impressed with the work that an organization called map Jonas is doing with planet and Google. There's a network of about 40 really extraordinary on the ground organizations working in Brazil to produce the highest quality, land use change land classification indicators that certainly I think they represent a a step change in terms of fidelity, accuracy and completeness. But but they represent, what we're seeing is is, you know, these tools have only been around for a few years at the scale that we can actually use them to solve practical problems. We've only had daily Earth imaging of the whole earth at this resolution for a few years, being we've only had industrial scale cloud computing for a few years, we've only had the ability to run these kinds of machine learning tools. And then you have to build the social relationships between the institutions on top of that, but what so what I would say to you is, I don't think we're there yet. But I believe that we're making unbelievably rapid progress to the point where we will have this kind of daily continuous classification of land use, for instance, at a meter or a few meters for the whole planet, you know, every day every week. The rate limiting factor will not be how often we can observe, but it will be how, how frequently that phenomena changes, how rapidly do the forests actually change, we'll be able to catch all that
Okay, we I'm just looking at the time we have about four minutes left. This might be one of the last questions, but it's a nice note to end on what is on your wish list for data inputs? And what new sensors or technologies, either ground based or above ground? are you most excited about?
I think that anything that can help the adoption of renewable energy is definitely high on my list, because essentially, the one of the main reasons why, you know, utility companies or governments or whoever has trouble buying into new renewables is because they're harder to predict, because you can't just flip a switch and then turn on the sun. And so for me, anyone who's working in the energy space, who's doing forecasting, now casting predictive maintenance, all of that is like is the biggest help that we can get here to push the needle on renewable energy use and any data I mean, actually, that's some there are some datasets that exists. They're mostly regional. Google has done some interesting work about detecting For example, solar panels and things like that, or wind turbines, but we don't have like a federated data set of, for example, renewable energy grids or any energy grids. Really, and so things that in that space, I think can just can can make the hugest impact. I think it's like 30% of greenhouse gases are from from electricity, something like our energy use, rather, something like that. And so if we could reduce that to you know, 5% that would be massive, right? Hmm.
I want to give you a what might sound like a slightly paradoxical answer, and then address a question that was asked by Sarah in the chat actually, Sarah Ramos, just a just a minute ago because I, I'd love to address it as well. So the first one is that, you know, there's so much entrepreneurial innovation that's happening right now, in terms of the creation of new kinds of sensors and datasets everywhere. You know, I Love this organization called sail drone that is deploying fleets, they envision and deploying fleets of thousands of autonomous sailing vessels, with instruments stacks in the water column that collect real time information about the oceans. There's immense innovation happening, you know, on the ground in terms of, you know, the Internet of Things, there's a tremendous amount of innovations happening in space. I don't think we're going to want for data, I think we're going to have a, frankly, a huge glut of data. So what's happened right now is that those kinds of new sensors are producing information, that with enough reliability, and enough consistency, and enough like frequency and resolution that we're going to see lots of new entrepreneurial organizations get built to translate the raw material into signals that people can can use to guide their behavior. I've no doubt of that. We're, we're in the earliest days of of that. The thing that if I could wave a magic wand and make a data set available. It wouldn't be any it wouldn't come from one of those sensors, it would come from what what is now the most critical missing information, which is administrative data, its asset level data, who owns what on the earth is actually really hard to know. When there's no central place, you go and go like, Okay, this little place on the earth, like, who show me the deed and the title and in some cases, it's because in many places in the world, there's no such structure at all, like the World Bank noted a few years ago that something like 90% of this of sub Saharan Africa was untitled land that people were living on. So so if we want to create structures of accountability, we have to know who owns what and that was, that would be the main thing I'd love to see produce. And I just want to say to you, Sarah, you know, I love the examples I showed today definitely are down payments on the future. So they're funded by governments, NGOs, philanthropies, but we're seeing more and more and more and more entrepreneurs and real organizations that that are being predicated on the availability of the data, not just exploring what it could do.
Okay, we are at time I'm going to jump in. Thank you so much, Sasha. Thank you so much, Andrew. Thank you to the XPrize family and everyone who joined our webinars, they fantastic questions. We didn't get to all of them, but they will all be saved and recorded, and I believe all of the chat and q&a will be shareable offline. And you'll know where to find us, folks. I want to point out that Marcus, thank you. You did a great. Well, my pleasure. My pleasure. I learned a lot and thank you for letting me needle needle you folks a little bit and bring in some questions in the audience. As I close I just want to mention that this is part of the environment breakthrough track at the AI for Good Summit. And you can check out the AI for Good website to follow along either past or future events. From the XPrize perspective, I want to point out that we have two active prizes, the carbon XPrize, about co2 conversion and the rain forest prize, which is what's reinforced preservation and conservation. And I'm actually going to share my screen just for a second here.
To give you a flavor for the just flashing images of some of the other projects that XPrize is working on, I share this to point out two things. One, we'd love your help in getting these projects, either off the ground are designed, the top row are things that are running, the middle row are things that we're trying to get launched. And the bottom things are things we're starting to design. These are all also problems in areas where we think AI and machine learning can really help drive innovation. So Sasha and Andrew will be calling on you to help support us, but we also look forward to participants in these prizes, bringing these kind of data driven solutions forward. So with that, I'm just want to say Thank you once again, thank you for taking an extra two minutes. Sasha, Andrew and everyone enjoy the rest of your day. And we'll throw it back to you. Thanks very much, everyone. Take care