 Live from Las Vegas, it's theCUBE! Covering AWS re-invent 2019. Brought to you by Amazon Web Services and Intel, along with its ecosystem partners. Hey, welcome back, everyone. It's theCUBE's live coverage. I'm John Furrier with theCUBE. We're here at re-invent day two as it winds down. Wall-to-wall interviews, two sets here. And we want to thank Intel, who's a big sponsor of this set. Without Intel, we wouldn't have this great content. They support our mission at theCUBE. We really appreciate it. We're here, extracting the signal and noise on our seventh re-invent of the eight years that they've been here. We've been documenting history, and we've got a great panel lined up here. We've got Sebastian Dehauer, who's the CO, sail drone, Henry Stahlstuhl, EVP of science and technology at Bowery Farming, great use case around the food supply, and Janet Pazera, space weather scientist at NASA, the Helio Physics Division. We've got a great lineup here, a great panel. Welcome to theCUBE. Thanks for coming. Okay, we'll start with you, Janet. You're doing some super cool space exploration. You're looking at super storms in space. What's your story? Yeah, I work at NASA, and NASA has, in its mandate, to understand how to protect life on earth and in space from events like space weather and other things. And I'm working with Amazon right now to understand how storms in space get amplified into super storms in space, which now people understand can have major impacts on infrastructures at earth like power grids. So there's impact. There's impact, I believe. And you guys are measuring that. Not like a supernova critical thing like that, but more of like practical space stuff. Yeah, yeah, actually the idea that the perception of the world of the risks of space weather changed dramatically in 1989 when a super storm actually caused the collapse of a power grid in Canada. And the currents flowing in the ground from the storm entered the power grid and it collapsed in 90 seconds. It couldn't even intervene. Wow, some serious issues. We want to get into the machine learning and how you guys are applying, but let's get through here. Henry, you're doing some pretty cool stuff that's really important. Mission, food supply and global food supply, something that you're doing. Take a minute to explain your mission. Yeah, so at Bowery, we're growing food for a better future by revolutionizing agriculture. And to do that, we're building a network of large warehouse scale indoor farms where we go all sorts of produce indoors 365 days a year using zero pesticides, using hydroponic systems and LED technology. So it's really exciting. And at the core of it is some technology we call the Bowery operating system, which is how we leverage software, hardware and AI to operate and learn from our farms. I'm looking forward to digging into that, Sebastian. Sail drone, you're doing some stuff, you're sailing around the world, you got a nice tan, is that you? No, tell your story. Suddenly, no, we use wind-powered robots to study the 70% of the planet that's currently really data scarce and that's the oceans. And so we measure things like biomass, which is how many fish there in the ocean. We measure the input of energy which impacts weather and climate. We map the seabed and we do all kinds of different tasks which are very, very expensive to do if you use ships. And super important now that climate change is on everyone's agenda, understanding potentially blind spots. Super important, right? That's what I'm trying to understand. There's some question of if it's a question of when and what and how much. And so the ice is melting, the Gulf Stream is changing and Nino is breaking havoc, but we just do not understand this because we just don't have the data in situ. We use satellites where they have very low resolution. They cannot see through the water. We use ships. NOAA has 16 ships here in the US. So we have to do better. We have to transfer this into a big data problem. So that's what we're doing. We have 1,000 cell drones on our planet. We have 100 in the water right now. And so we're trying to instrument all oceans all the time. You know, and data scales your friend because you even want more data. Yes. Talk about what you're working on. What kind of AI and machine learning are you doing? You're just gathering data and you're pumping it up to the cloud via satellites or what's going on there? So one of the use cases is trying to understand who's out there, what are they doing and are they doing anything illegal? So to do this, you need to use cameras and look at the horizon and detect whether you have vessels. And if those vessels are not transmitting the position, it means that they're trying to stay hidden on the ocean. And so we use machine learning and AI that we train on AWS to try to understand where those things are. And it's hard enough online. I'd say it's very hard because every pixel is moving. You have waves, the horizon is moving, the sky is moving, the ship is moving. And so trying to solve this problem is a completely new thing. That's called maritime domain awareness and it's something that has never been done before. And what's the current status of the project? So we've been live for about four years now. We have about a hundred sail drones. We're building one a day towards the goal of having a thousand which will cover all the planet in a six by six degree squares. And we are operationally active in the Arctic, in the tropical Pacific, in the Atlantic. We just circumnavigated Antarctica. So it's a thing that's real. It's out there, but it's very far from land. So the spirit of the cloud and agility, the static buoy, goes away. You want to put these sail drones out there to gather and move around and capture. That's right. A buoy is a massive steel thing which has a four mile long cable and it's- It's a silo. In a fixed station, it's one point and the ocean goes by you. Having a robot means that you can go where something interesting is happening, where you have a hurricane, where you might have an atmospheric river, where you might have a natural catastrophe or man-made catastrophe. So this intelligence in the platform is really important and the navigation of that platform requires intelligence. And on the data side, getting a thousand times more data allows you to understand things better just like my colleagues here are doing. And is it a non-profit? Is it a for-profit venture? It's a for-profit company. So we sell raw data at a fraction of the cost of existing solution to try to create this kind of transformative impact on understanding what's happening. Super exciting for all the maritime folks out there. Of course I love the ocean myself. Henry, you're tackling a real big mission. How are you using technology? I can almost imagine the instrumentation must be off the charts. What's your opportunity to look like from a tech perspective? Yeah, so the level of control we have in our farms is really unparalleled. We can tune just about every parameter that goes into growing our plants, from temperature to humidity, CO2, light intensity, day-night cycles. The list keeps going on. And so to do more with fewer resources, to grow more in our farms, we're doing something called science at scale where we can pull different levers and make changes to recipes in real time. And we're using AI to understand the impact that those changes have and to guide us going from millions of different permutations, trillions of permutations really, to the perfect out on for converging. And you iterate, look at the product outcome, and you circle that data back, is it all on Amazon? We do operate on Amazon, yeah. And we're using deep learning technology to analyze pictures that come from cameras all over our farms. So we actually have eyes on every single crop that grows in our facilities. And so we process those, learn from that data, and funnel that back into the system. Yeah, like maybe put more light on this or do that and kind of adjust the conditions, is that that thinking? That's exactly it. There's lots of different types of plants. We grow butterhead lettuce, romaine, kale, spinach, arugula, basil, cilantro. So there's a lot of different things we grow and each of them require different little tweaks here and there to produce the best tasting and most nutritious product. That's cool. Janet, space is obviously on one end of the spectrum. We're going to live on Mars someday. So you might be a weather forecaster for what route to take to Mars. But right now, the practical matter is this real correlation between these storms. What kind of data problem are you looking at? What is the machine learning? What are some of the cool things you're working on? Right, we have a big data problem because storms of that magnitude are very rare. So it's hard for us to find enough data to train AI. We can't actually train AI. We have to use learning that doesn't require us to train it. But we've decided to take the approach that these super storms are like anomalies on the normal weather pattern. So we're trying to use the kind of AI that you use to detect anomalies like people who are trying to break into, to do bank fraud or do web server tax. We use that same kind of software to try to identify anomalies that are the space weather and look at the patterns between sort of a normal, more of a normal storm and a huge space weather event to see how the patterns compare and how you're amplifying the regular storm into this big super storm activity. So it sounds like you have to be prepared for identifying the anomaly. So you're looking at anomalies to figure out where the anomaly might be ready to be ready to get the anomaly. Yeah. If that's not wrong. You look at the background and then what sticks out of the background that doesn't look like the background is identified as the anomaly and that's the storms that are happening which are quite rare. Well, all three of you guys are doing some real cutting-edge cool projects. I guess my question would be for the folks that are putting their toe in the water for machine learning, they tend to be new use cases like what you guys are doing. Whether it's just a company trying to re-bacter themselves or become reborn in the cloud, ran legacy stuff. When you're here at Amazon re-invent, this is the big question for these folks that are here. You guys are on the front end of some really cool projects. What's your advice to the people who are trying to get in that mindset? Yeah, so I think the way to think about this is if you're good at something and if you think you have the solution for something, how can you make that a million times more efficient? And so the problem is there's just not enough capacity in the world usually to treat datasets that are a million times larger. And this is where machine learning should be thought about as an extension of what a human is really good at using a pair of eyes, ears, you know, or whatever other sense. And so in our case, for example, counting fish, acousticians, train acousticians, look at sonar data and understand schools of fish and can recognize them. And by using this knowledge base, we can train machines to do this on a much grander scale. And when you do it on a much grander scale, you derive a whole entire team. And your point is that humans are critical in the process. So scaling the human capabilities and maybe filling in either scale issues or... That's what AI machine learning is. It's the greatest enabler of our time. It enables us to do things which were impossible to do before because we just didn't have enough people to do them at scale. A key is being able to ask questions, right? And so if you have the questions to ask, you can apply this technology in a way that's never really been before possible. Janet, your take. Yeah, I'm actually someone who didn't know anything about AI or ML when I started. I'm a research scientist that does space weather. So coming into this, I'm working with the ML solutions lab here and putting AI experts with experts in space weather, we're getting, we're doing things that are going to give us new advances. I mean, we're already seeing things we didn't know before. So I think that if you partner with people who really have strong AI knowledge, you can use your knowledge of science to really get to the really important issues. Okay, I have to ask the final lightning round question. What is the coolest thing that you've done with your project that you've either observed, implemented? That is super cool. Super cool. What's the coolest thing? Well, in terms of us, we're using anomaly detection to identify storms and in the first round through, it actually identified every single super storm which was not the major super storms, but it did, but it also started identifying other anomalous events. And when you went and looked at them, they were anomalous events. So we're seeing things, but it's picking out the weird things that are happening in space weather. It's kind of exciting and interesting. I work for a day with you. I would love to just look at these anomalies. Henry, what's the coolest thing that you've seen or done with your project? I think the fact that we've built our own custom hardware, our own camera systems, and that we feed those through algorithms that tell us something about what's happening minute by minute with plants as they grow to see pictures of plants minute by minute, they dance. And it's truly, it's remarkable. Wow, fascinating. Sebastian. We've counted every single fish on the West Coast in the United States, every single year from Canada to Mexico. I thought that was pretty cool. I didn't think it was possible. Very cool. What's the number? It's huge. Yeah, if I could tell you, I would, but I'm not allowed to tell you. Next year. You can get the salmon, you can know where the salmon are when they're running, all that good stuff. Awesome. Well, congratulations. You guys are doing some amazing work. This is pioneering. A great example of just what's coming. And I love this angle of making larger human impact using technology, where you guys are shaping technology for good things. Really, really exciting. Thanks for coming on. Thank you, John. Thanks for having me. I'm John Furrier. We're here live in Vegas for Reinvent 2019. Stay with us for more coverage. Day three coming tomorrow. We'll be back with more after this break. I want to thank Intel for making it all happen, presented by Intel. Without their sponsorship, we wouldn't be able to bring this great content. Thanks for watching.