Good morning, Benas. It's great to be in a place where if people introduce you, you have no idea what they're talking about. For some reason, I could tell you I'm the dog walker of Amazon. After, was it actually, if I look in the audience, I've been longer at Amazon than you guys been alive.
And I'm pretty sure that quite a few of you do not know a world without Amazon. But that's not what I'm going to be talking about, not at all actually.
Over the past decade or a bit more, I've become really, really interested in the fact that we have a lot of hard problems in our world. If you listen to the United Nations, they have a list of 19 sustainability goals. Mostly because in about 20 years from now, there will be two billion more people on this planet and most of them will be below the equator. How can we make sure we have food security for them, economic future, healthcare, all these kind of things? I think us as technologists have a massive, important role to play in all of that.
In all of that world of technology and trying to do your best in helping other people, there is a fundamental part: technology is basically a map. We have a map. If you didn't know where the venue was today, you probably walked around with your cell phone and used Google Maps, Microsoft Maps, or Apple Maps, and you were able to find it here. Yet there are billions of people on this planet that you would not be able to find because there is absolutely no map that leads to them.
Of course, we're all very excited about the capabilities of new AI and what it can do for us. One thing to remember: AI is about 70 years old. That stuff that just happened two years ago is just a new version of it. There's a whole body of work before that that works. The term AI was coined by McCarthy and Minsky in 1956. A number of years later, McCarthy said, as soon as it works, we don't call it AI anymore. Now, basic AI is data—without data, no AI.
Let me start off with telling you a personal story. This guy is Jim Gray. If you've never heard of him, I urge you to look him up on Wikipedia. Jim Gray is the inventor of transactions. He worked at IBM in the 60s with a number of other people and invented a lot of the technology we use today.
Some Sunday morning in 2007, Jim—he was my mentor—went sailing. He was an amazing sailor. He had this beautiful boat, Tenacious, and could handle it all by himself. He left San Francisco to the Farallon Islands on a special day because his mother had died. He had her ashes and was going to distribute them over the ocean. Halfway through that trip, he disappears. The boat disappears completely. Nothing to be seen.
He wasn't just my mentor. He was a mentor of the founders of Google. Everybody knew him. So what happens when we notice he's gone? All the technology world resources come to life. The Google guys repositioned satellites to take imagery of the complete space where the boat could have been. At Amazon, we had a service called Mechanical Turk, which Jeff ironically called artificial artificial intelligence. On the outside it looked like AI, but inside it was just humans doing tasks. In this task, they looked at little squares to try to find either a boat or remnants. But nothing was found. To this day, it's an absolute mystery what happened to him. The Coast Guard sent out airplanes, everything. There was a lot of data being generated, not very fast.
The question is: with all the new AI technologies we have today, would we have been able to find that boat? Maybe. If you look at technology today, there are many things we didn't have in 2007. We would be able to have a complete cloud with tens of thousands of machines processing data. Predictive modeling is much better today, and we can bring multiple data sources together. Who knows, maybe today's AI would have been able to find him. But there's one really important lesson: that was only possible because the executives of very large companies repositioned everything to get access to data for one person. It was privileged data accessible only to the biggest executives in the world.
If that were the only crisis we ever run into, it would be a one-off, but it's not. Today, we see continuous crises around the world. The most important thing is that if there is a crisis and you don't know where people are, they're not mapped, meaning they're not seen, and you can't serve them in any way. I don't mean serve from an Amazon package perspective; I mean in times of crisis, can you reach these people? Accurate maps can be the difference between life and death.
Let's go back a number of years. Haiti was hit by an earthquake in 2010. Half of Port-au-Prince was destroyed. Even though emergency workers knew where they had to go, they had no idea how to get there because there were no accurate maps of Port-au-Prince. That's crucial in an urgent emergency. But I recently visited Nigeria, and in Lagos there's a part called Makoko. In Makoko, 300,000 people live in houses on stilts in the lagoon. If you look at a map, it's blue. There are no houses there, no 300,000 people. It just looks like water. So how do these people get access to emergency services or talk to the government about where they live?
As much as we love AI, whether today's or from 30 years ago, if you don't have the data, AI can't model it. It can't model what it cannot see. What I've learned about mapping is that the maps we have are built for economic reasons. If there's no economic reason, there's no map. Billions of people are unseen because they're unmapped. We need to change this. We need to start mapping for humanity instead of just for economic reasons.
Let's think about maps themselves. Maps have multiple levels of dynamism. The basic layer is the earth layer, which doesn't change much—mountains don't move every month. That layer may change in decades. Then you have the infrastructure layer: roads. Roads don't change every day, but maybe yearly. Then you have a seasonal layer—in areas with a lot of foliage, some satellites can't see through it, but at another time of year you can. Then there's the real-time layer, important in crises, with things that change daily—a tree might fall on a highway, an obstacle to reaching people. These are different time scales. It's not just one map; having a single satellite image is not sufficient.
Maps are nothing else but data. We're used to looking at maps to go somewhere, but they are data you can push through technology to analyze or make decisions. They are data models with different data sources: satellites, drones, IoT, mobile devices. In 2008, there were 150 Earth observation satellites. Now there are over 10,000. Not counting Kuiper, Starlink, or spy satellites—just Earth observation. Some look for crops, some for agriculture. Here in Argentina, you make a significant contribution. There's a company called Satellogic with 50 operational satellites tracking things. They use something called a ground station, which we at Amazon provide for these satellites. Satellites aren't just about imagery; they have radar, LiDAR, multispectral sensors. There's a massive dataset out there.
Sometimes humans are important too. Let's go back to Haiti in 2010. Two days after the earthquake, we repositioned satellites to see the roads of Port-au-Prince in fine detail. That map was used by the Red Cross, the UN, and the US Marine Corps. Not just from satellites—there were also humans on the ground walking around. There were 600 volunteers mapping alleyways, indicating which ones no longer existed. They collected data and put it in OpenStreetMap. If you're working on projects, use OpenStreetMap—it's one of the most detailed, highly accessible data sources.
These volunteers triggered an organization that later became the Humanitarian OpenStreetMap Team. They organize mapping for unmapped places. There's a project called Missing Maps where you can contribute. I visited the Philippines with this team. On Saturday mornings, they get people from the community to walk around with a map on their phone, mapping their neighborhood and adding context: this house is made of cardboard, that church is stone. That's important because the Philippines is in typhoon alley—six typhoons a year, multiple earthquakes, volcanoes. Yet 90% of the country is unmapped. Without maps, the Red Cross can't go there.
The Humanitarian OpenStreetMap Team uses that data for a service called F.A.I.R. In a crisis, they take data from OpenStreetMap and bring it together in an AI model you can ask questions to. It's accessible for everyone to use, no payment required. Another team that uses this data is, for example, goes on the ground in areas with major issues. South Sudan is involved in a nasty war. The question is whether crops will be planted, because if not, there will be famine the next year. So it's crucial to have this information.
Satellites are great, but these are the cheap ones. You get a good meta view, but you want the detailed information that only the NSA or spy agencies have. That's what you need in crisis mode. But even satellites don't get that level of detail. So drones are the next level—you can fly drones and get good insight into maps. The Humanitarian OpenStreetMap Team provides drones to communities and teaches people how to create maps of their environment.
Coming back to Makoko: the 300,000 people who live in what looks like a lagoon on a map started using drones. Now there's a map that looks quite different. It means they can get access to government services because they're seen—there's an address. There are a lot of interesting things spinning off the fact that there is just a map of these people.
Space, air, ground: ground is equally important. Being on the ground is where the truth is. Whether you can see it from the top doesn't mean you can walk there. It provides real-time insight. We walk around with billions of mobile devices, and even more in the IoT world. The Watershed Organization Trust builds information centers that track crop growth in parts of the world with no insight. Mobile and IoT really enable everyone to become a cartographer.
Think about a city like Jakarta. A significant portion is not finely mapped. There's a company called Grab that does food delivery and motorbike rides. Their problem is that customers want to go to places that don't exist on the map. So they created Grab Maps. The official map of Jakarta is one thing; what Grab knows is different because their drivers go everywhere. Similarly, in India, a business intermediary between tuk-tuks and riders goes places cars can't, delivering people to unknown places. Now we have a map of India that didn't exist before because this company kept the data of where their drivers go.
Without maps, it's very hard to solve the big problems we face as humanity. Maps are the fundamental dataset if you want to improve people's lives. I was fortunate to visit Rwanda earlier this year. The Ministry of Health has integrated technology into almost everything. They took me to their central healthcare center. They track everything happening with respect to health and government policies. Many well-developed countries don't have this insight. Rwanda does it from scratch.
For example, a pregnant woman in Rwanda should not walk more than an hour to get to a health center. That's a massive improvement from walking two to three hours or not at all. By using maps and overlaying data, they figure out where the red spots are—places where women have to walk more than an hour—and place new health centers there. By having an accurate map with relevant data, it's crucial.
Let me give you a completely different example. The Ocean Cleanup Project: do you know how much microplastics are in your body? Tons, especially if you eat fish. They've pulled tons of plastics out of the Pacific garbage patch. They have the 30 Rivers Project, tracking the 30 rivers that are major contributors to plastic pollution in the ocean. Two weeks ago, I was in the Amazon for the first time after 21 years at Amazon. I was surprised to see plastic floating around. The Ocean Cleanup Project uses maps to track major polluters with cameras on bridges, AI processing at the edge, and even fake plastic with GPS trackers to see where it ends up. It helps figure out where major polluters are.
All this data together—satellites, drones, IoT, mobile—some say it's too big to handle. But my cloud is bigger. Whatever data you have, or your organization has, there are projects at AWS that allow you to store data for free. Upload is free, download is free, storage is free. The Registry of Open Data contains satellite data, drone data, OpenStreetMap data. If you need map data for a project, like Digital Earth Africa, that's 30 years of satellite data of Africa from NASA, ESA, JAXA. It allows projects to run, like detecting illegal mining by spotting new roads, or tracking mangrove deterioration over decades.
Your local company, Satellogic, has put data in there too. If you want accurate open-source data for free to think about how crops are grown in parts of Argentina, this dataset is available. It's open for everyone. There are about 300 petabytes of data. If you have data or want to access it, now is the time to get that QR code. We have a sponsorship program to collect data and support projects so you can compute, analyze, or use it in AI models.
Coming back to the beginning: the day Jim Gray disappeared. We had the data, but we didn't have the AI to immediately take action. Open data, AI (which is just a tool), plus cloud ability for global impact. I am convinced that data sharing is not just an economic thing—it's a moral imperative. We need to share data to solve the world's biggest problems. There are 17 sustainability goals of the United Nations for when we add two billion more people. If you don't have the data, AI is useless. If AI is useless, we're not making progress toward these goals. Data must be of high quality. Garbage in, garbage out—but now garbage in creates very convincing garbage out. Data, data quality: no data, no AI, no progress.
I urge you all to think about spending a Saturday morning on one of these projects. Does your organization have data that is not really IP but would help others solve hard problems? Go think about that, go do that, and then go build something.
