Thank you. Good morning. I would like to talk to you about all these wonderful AI tools that we've been building and that are available to everyone, but there's a precondition to be able to use AI for good, and it is to actually have access to good data. Data is a precondition for AI, and especially if you think in the area of mapping and geolocation, data is crucial to be able to apply our technology for good. But to talk about that, I actually want to start off with a more personal story. This is Jim Gray. And if you're not out of computer science, you may not know him. But Jim Gray received the Turing Award, basically computer science's Nobel Prize, for the invention of transactions. And Jim was a good friend of mine, as well as of many other luminaries in Silicon Valley. On a quiet Sunday morning, January 28, 2007, Jim took his boat out, he was a very experienced sailor, to go to the Farallon Islands. And shortly after leaving San Francisco, all contact was lost with him. And so at some moment, given that there was no sign of him, all his friends jumped into action. And it meant that, you know, actually getting satellite imagery, moving reposition satellites to be able to find him. Coast Guard went out, the airplanes flying over this area. And after all that work and after all the generation of this data, Jim was still not to be found. And so the question is, you know, with all this data available, would today's AI have actually found him? Maybe, I don't know, you know, but there are things that we can do now that we definitely could not do in 2007, and that was actually sort of move much faster. There's much more technologies now that can actually get this data in the hands in real time to be able to track things. Originally with Jim Search, we made use of an Amazon service called Mechanical Turk. Over 6,000 people signed up for Mechanical Turk to just be able to look at the hundreds of thousands of small tiles of satellite data to see if we could find any trace of Jim or his ship, but he remains completely lost at sea. But the only reason why we would be able to actually use AI to actually find Jim would be because there was access to privileged data. Government clearances were required. Private sector relationships were being used to actually reposition satellites to be able to find him. This shows actually a little bit of what I would call a data divide, because for one privileged individual, we could actually deploy this representation, but when disaster strikes communities, they're effectively invisible. There's no detailed maps, there's no real-time monitoring, there is no predictive capabilities to guide the response. And accurate mapping can really make the difference between life and death. And we've seen this in multiple crisis zones. You know, think about Haiti in 2010. On January 12, 2010, the catastrophic 7.0 earthquake struck Haiti. And as the international rescue teams arrived in the city, they couldn't do their work because literally the city was unmapped. Emergency responders had coordinates, but they couldn't find where that would be. They couldn't see the difference on the maps that they had between an alley or a major roadway or where is critical infrastructure. Another great example is actually Makoko in Lagos, which is a wonderful community of well over 300,000 individuals that live on houses on stilts in Lagos Lagoon. If you look at the map, it's a big blue spot. These 300,000 people are not mapped, not seen, and as such cannot access services. Just basic services remain unavailable to them because they're not mapped. And for AI, it is crucial to have access to accurate data. And it can't model all the things that we would like to do with AI for good, whether it's in crisis or whether it's in climate change or anything else. It requires access to accurate data. Why isn't this data there? Because I mean, after all, we all walk around the city these days with a phone in our hands with a map on it. But it turns out that most maps are made for commercial use. They're not made for disaster response. They're not mapped for humanity. They're mapped for economic progress. Detailed street views of shopping districts are fine, but areas like Makoko completely remain unmapped. And when we only map what's profitable, we perpetuate existing inequalities and we leave the most vulnerable communities exposed. This is not a technological problem. It's an equity problem. Now, if you think about maps, there's not just one map. There's not one map of the earth. The moment that you have a traditional map in your hand, it is out of date immediately. So real good maps require us to be active at multiple different time levels. Basically, I think of this as having four different layers. So there's the earth layer, the one that sort of changes in terms of decades. I mean, the Himalayas are not going anywhere. But then there's the infrastructure layer, roads and things like that that basically change on a yearly time scale. Then we have the seasonal one where things are changing over seasons, and think about rivers, think about floods, think about vegetation, water levels, things like that. And then there's the real-time layer, the one that's in constant flux with dynamic events like human activity, weather patterns, emergency situations. Modern mapping needs to accurately simultaneously combine all of these data sources to be able to actually support these vulnerable communities. Now maps are nothing else but data models. Combining these data layers of the different time scales is crucial to actually be able to help and do AI at a for good level. And data models of course need data sources. There's many ways of bringing sources that create this particular model. The barrier of entry in all of this has been significantly reduced. But let's talk about remote sensing first. You know, satellite imagery has really transformed the way that we see our planet. From tracking deforestation to monitoring crops, from planning cities to responding to disasters, the technology keeps improving. Better resolution, more frequent passes, smarter sensors. Thousands of satellites orbit the Earth at different layers, very crowded low altitudes and all the way up to high altitudes. But in 2008, there were only 150 satellites that did Earth observation. These days, it's well over 10,000 satellites that are orbiting the Earth and observing the Earth continuously. And it's not just with imagery. There's a whole range of new types of sensors available. The European Space Agency recently launched the EarthCARE satellite. It has a cloud profiling radar, an atmospheric lidar, a multispectral imager, a broadband radiometer. All of this information and data they can give us is incredible. However, in humanitarian context, simpler imagery is all that's needed to make a difference. Look at Haiti. Come back to that story of the earthquake in Haiti in 2010. Within 48 hours of the earthquake, 600 volunteers started creating the first reliable crisis map using satellite imagery and OpenStreetMap. It became the first example of real-time crisis mapping. Here it is before the earthquake, January 10. Basically the map is empty. Two days later after the earthquake on January 12, after hours of volunteer mapping, you can see how quickly the map of Port-au-Prince is actually starting to take shape. This will become the default map for organizations responding to the crisis. Including urban search and rescue teams, NGO partners, the United Nations, the World Bank, the US Marine Corps, all used that map. Out of this, OpenStreetMap really emerged as the volunteer-driven platform for creating free open map data with recognition of the human potential. It actually came out of the 2010 Haiti earthquake mapping came the Humanitarian OpenStreetMap Team with a focus on crisis response and community development. Missing Maps is co-founded by HOT and takes this a step further, proactively mapping those communities using volunteer mappers. Then all three share the same foundation, the OpenStreetMap platform, to ensure that map data remains open, free, and accessible to all. I once filmed a Now Go Build episode with the OpenStreetMap team in the Philippines. Meeting for example with the Red Cross, they indicated that the only way they could do their work is actually using the OpenStreetMap and all the data in there from the Humanitarian OpenStreetMap Team. The Philippines sit in what's called Typhoon Alley. Earthquakes, typhoons, disasters, weather all the time. And for people to be able to reach fundamental infrastructure, it is crucial to have this open data available. HOT also works on AI-assisted tools, and FAIRE is one of their tools. It's an AI-assisted mapping service that combines machine learning with human expertise to accelerate humanitarian open mapping, and it enables continuous improvement through real human feedback. When the war in Sudan broke out two years ago, the question was, are people going to plant their crops or are we going to see a famine in Sudan? Comparing and contrasting the current situation with previous situations, they were able to indicate whether or not this famine may happen, and as such proactive action should be taken. Now not all satellite imagery is created equal. Free low-res is often the barrier to entry. The low-res data is available freely to many, but the high-quality stuff remains privileged, and often in disaster situations you need high-quality detailed imagery to be able to do your work. Now think about those four dimensions. The real-time data may be cost prohibitive for smaller organizations, and in such cases other tools are required. Drones these days fill one of the gaps for those tools. They can capture detailed imagery on demand, fly under cloud cover, and reach areas that are difficult to access. And for communities that have never been properly mapped, drones are actually a game changer. A local team with a basic drone can now create a high-resolution map of the area in days, not in months or years. Go back to Makoko. If you look at that unmapped floating community of Makoko, drone mapping has become a way to turn around the fortune of those 300,000 people that are living there. In 2019, Code for Africa started a project called Mapping Makoko, and it was funded by the Humanitarian OpenStreetMap Team, and they trained local residents, including women from the community, to pilot drones and to use OpenStreetMap to actually map their community. And this grassroots effort didn't just create a map, it provided them with a powerful tool for advocacy, because now suddenly they were seen and their infrastructure was seen, and it became a data source not just for humanitarian rescue efforts, but it became a political tool to be able to show how their community looks and where crucial infrastructure was still missing. Now the most accurate maps of course combine these multiple perspectives, from space to air and ground, a complete picture of a community, and the ground truth in all of that is of course mobile and IoT. Now, if you look at all of the devices around, over 8 billion mobile devices, many different IoT devices, all are good data sources for real-time information on top of your maps. IoT sensors like temperature, humidity, pressure, motion, light, gas sensors. It's important to build capacity not just to solve immediate problems, but to actually really build out the community owning their neighborhoods and really track changes in real time and contribute to a more complete picture of what's happening on the ground. This is true democratization of data and technology, and mobile and IoT we've seen that actually really empowers everyone to become a citizen cartographer. Look at Grab, if you know Grab. Grab is Southeast Asia's super app, whether it's food or transport. The problem is that large parts of the countries where they need to operate are not mapped, and so Grab took it upon themselves to actually map the countries and the cities where they're operating in. This is Jakarta. The empty map is what normally is being shown. The detailed map is available for Grab because this is where the tuk-tuk drivers and the car drivers and the motorcyclists go every day to deliver either people or goods. And the same is for example in India, where YamaMoto is an app that actually connects tuk-tuk drivers with their customers. Again, they're creating maps by themselves over the roads that they are running on. And so these IoT as well as mobile devices create these kinds of maps. Many of the technologies I've just looked at are being pioneered by very forward-looking individuals and social impact organizations. I'm very passionate about this, and I've set up the Now Go Build CTO Fellowship that last year kicked off with two cohorts in climate change and disaster management. We've also produced a short video series to highlight the work of these fellows and their organizations. I'd like to show you a small clip from that. There are many places around the world where there are no maps available. Food waste is a huge problem. The population on the earth is growing every year. The digital divide, resource constraints, these are all real. Join me as we meet the inaugural cohort of the Now Go Build CTO Fellowship, which brings together technology leaders tackling humanity's greatest challenges. Tools like AI and ML can forecast agricultural production. Use of technology in climate adaptation is extremely important. Data is key. If we leverage the technology well, it has immense potential to transform the entire rural sector. Drones map the area after the disaster to provide data to assess the situation, and that saves lives. Solving global challenges isn't just about new technology. It is about who we empower to use it. Now Go Build. You can watch the first episodes on my blog, All Things Distributed. Follow the QR code there. Many of the other episodes are being delivered in the coming weeks, and some of the fellows actually will be on this stage this week as well in the conference talking about the work that they're doing. Now you can't solve these big problems in this world with just one perspective. You need to have multiple layers and multiple levels. But accurate data can make a significant change. And one of the places where I was completely blown away recently about their use of mapping and data is in Rwanda. I visited Rwanda a few weeks ago, and I had the fortune to visit what's called the Rwanda Health Intelligence Center. And what I saw was actually amazing. Just a decade ago, Rwanda was one of the last countries to be comprehensively mapped. But today they use very sophisticated mapping technology to transform, for example, healthcare delivery. This is their operation center, which has huge screens in real time exactly tracking who is using which health services under which conditions. And this is a truly impressive system. And one of the things, for example, that they can do with this very accurate mapping data, combining that with health data, is actually figuring out how far women that are pregnant actually have to walk to get to a health center. And so this is a map of health centers in Rwanda. And they continuously track that a woman should not have to walk more than an hour to get to a health center. And they use this mapping data as well as this walking data as input to where to place their next health centers. It is immensely satisfying to see how accurate data can actually drive significant change at the healthcare level. Another great example is that of the Ocean Cleanup project. It's tackling one of the biggest problems in this world, namely removing plastic from the oceans and our rivers. And they aim to clean up 90% of all plastic that is floating around by 2040. They have already removed tens of millions of kilograms of plastic from, for example, the Great Pacific Garbage Patch and from rivers and elsewhere. Recently they started a new program. They announced their 30 Cities Program, which targets the rivers that are responsible for one-third of the ocean's plastic pollution. And the program relies on a new river model which they've created using drones, AI analysis, and even GPS-tagged dummy plastics. And this computational model predicts how and where plastic will travel, helping them position their cleanup system to have maximum impact. And part of this river model actually identifies different types of plastic. They have these AI-powered cameras on bridges and on ships that analyze video streams and identify plastic waste. The investment in data collection is what led them to accurate insights that underpin all their inventions. Now for many of you, you say, yeah, but this geospatial data is so big. All that data from satellites, drones, live, mobile devices, they generate enormous amounts of data. We're talking about hundreds of petabytes of data. You know what cloud is bigger? Cloud can handle this scale. Amazon S3 processes 100 million storage requests per second and stores more than 350 trillion objects. This means organizations can focus on the data to solve the problem, not worry about the infrastructure for storing and processing it. Having infrastructure to handle this big is important, but data access is crucial. Through the AWS Open Data Sponsorship Program, we help make these high-value public datasets available to everyone. And this includes essential mapping like OpenStreetMap, like Sentinel-2 satellite imagery, or USGS Landsat data. And by removing the barriers to access to these data, we help scientists and humanitarian organizations and local communities to innovate, not just the big players. A great example is Digital Earth Africa, where actually this data from NASA as well as ESA and JAXA combined shows the changes in Africa over the years. For example, Mangrove Watch is one of those organizations that looks at the changes in mangroves over the years, which is only possible because of available public data. So the Open Data Sponsorship Program actually covers the cost for these public datasets. They make them free to use for everyone. There's over 300 petabytes of data available for everyone. It's maps, it's imagery, it's sensor data, everything can be used freely. And a great project that recently came to my attention is called Clay. It really walks the talk when it comes to the importance of open source data. Clay is an Earth observation foundation model trained on data freely available through the Registry of Open Data on AWS. And their mission is to make Earth observation as usual and as useful and as accessible as, let's say, web search is today. They believe in open, responsible AI to make Earth observation a force for positive change. And they're being used by journalists, by nonprofit organizations, by NGOs to understand the changes of the Earth's surface, whether that is mapping seagrass or tracking plastic or finding illegal mining. For example, here, I know a small organization which is a small mining remediation organization that was able to find five times as many methane leaks as before using data provided by Clay. Now Clay used this Registry of Open Data on AWS. We also gave them $3.5 million in credits to be able to compute these models. And the data sponsorship program is available through source group. Clay has also made their model as well as weights available to the public through source group and through the Registry of Open Data on AWS. Now going back to the story in the beginning, when we searched for Jim Gray, we had the data: satellite imagery, weather patterns, ocean currents, all that particular data. We just didn't have the technology to actually make sense of it quickly enough. And today we're building systems that can process vast amounts of data in real time, from tracking health outcomes in Rwanda to predicting plastic flows in our oceans. And there's something else that we actually learned from Jim's story. He spent his career trying to help people make sense of vast amounts of data. And that's exactly what these mapping technologies are doing, helping us understand the world so that we can make it better. And with AI, we can extract insights and make predictions. This is a formula for change: take open data, use powerful AI models, and run it on a cloud infrastructure. Now you've got a planetary problem-solving machine. This trio is already turning what used to be intractable problems into practical ones. And crucially, open data and open AI means anyone can verify the results. Global challenges become a shared endeavor with facts accessible to all. And given that you can replay any of the investigations, it means that nothing can be swept under the rug because all of this is open accessible data. Now, if we have data that could save lives and protect the planet, keeping it private is morally indefensible. In this age, choosing not to share is choosing not to help. Sitting on data and capabilities while the world burns, you're complicit in the problem. Think about the 17 Sustainable Development Goals. Each and every one requires geospatial data. No mapping, no SDGs. Why? Because you can't fix what you can't see. If we can't map where poverty is, if we can't map where schools or hospitals are, how do we end poverty? How do we ensure health? From zero hunger to climate action, every goal is location-based. Progress on crop yields, disaster resilience, clean water access, all require maps and spatial data to measure and drive change. If we do not capture, create, and share data, we will not make progress. No data, no AI, no progress. This is a challenge for all of you. What data do you need? And more importantly, what data can you share? Remember, maps aren't just tools for navigation. They are tools for justice, healthcare, environmental protection, tools for making the invisible visible. And with that, thank you. And now go build.
