 Welcome everyone. It's great to see so many people here for the Ideas Festival. I'm delighted to be here today. My name is Karen Plout and I'm the Dean of the College of Agriculture and I have the privilege of introducing our special guest. This is a unique event. It's part of the 150th anniversary of Purdue celebrating giant leaps resulting from daring intellectual risk taking of the past, present and future here at Purdue and throughout the world. Today we delve into the sustainability of our planet and economy as we explore the question, what if food was digital? With distinguished guest, Miles O'Brien, award-winning science correspondence for PBS NewsHour and Caleb Harper, principal investigator and director of the Open Ag Initiative at the Massachusetts Institute of Technology Media Lab. In his role at MIT, Harper leads a group of engineers, scientists and educators in research that aims to shape the development of future food systems. He and his colleagues are developing an open source agricultural hardware, software and data common that Harper hopes will foster a more collaborative food system. Harper's main areas of research focus on controlled environment design, actuated sensing, controlled automation, energy and biological optimization. Caleb Harper's work dovetails very well with some of our efforts here at Purdue in the research arena and so we're really excited to welcome Miles O'Brien and Caleb Harper. Thank you, Karen. Appreciate it. It's great to be with you all again. A little warmer here in Indiana from the last time. I remember when it was like, I think 30 below zero on the vortex. This is like summer, right? Is this summer in Indiana? This is fantastic. It's really great to be talking about a fascinating subject, the intersection of science and technology and agriculture and what that all portends with my friend Caleb Harper, who I've known for a few years and have been following him all around the country doing some filming for a project forthcoming. Maybe you'll see it someday soon on PBS if we can find the right interest level there. And perhaps maybe you'll have some ideas on how we can sell it after you've seen this elongated pitch tonight for what will be the Caleb Harper nerd farmer show. In any case, when I saw him speak a couple of years ago, what really struck me was this. The apple. We're going to sort of begin our adventure here tonight as Newtonian physics began with an apple. It's not going to hit my head, but the apple itself was a part of his talk. You might have seen it online. It's a big part of his TED talk, wildly popular TED talk. I bought this apple just a short time ago. There I am down in the Union. It's not a great apple, frankly. It's a little mealy, but that's this time of year. But the real question is, Caleb, what's your best guess on average? How old do you think this apple might be? So first of all, I don't pay miles as a pitch guy for me, but thank you very much for all those amazing things. So let's take that story back to beginning. So I started working with National Grocer in my work to see, you know, at the very beginning, we're really a bunch of engineers. We have now, thankfully, a bunch of plant scientists that joined the team, but at the time we were thinking, if we develop these tools, what could we affect? And so we did a survey of eat fresh, pick fresh produce, and the thing that jumped out to us was this apple. So the average age of an apple in a U.S. grocery store is about 11 months. That's 11 months from the day it was picked until the day you put it in your mouth. And I know we're all a bunch of ag folks, so probably more than most places we understand why that is. But they're picked, they're waxed, they're put in cold storage, and they're hedged along with the market as a commodity. And I think that was a brilliant thing at the time when we thought, how could we get everyone an apple a day and the whole doctor away thing. But when we did the chemical analysis, that apple had lost all of its antioxidants by the point it was available at the store. It was basically a little ball of fiber and sugar. So a sugar ball a day? Yeah, you know, diabetes apples. No, don't quote me. But the revelation for us was how big our system has gotten and how it was really and continues to be mostly optimized for yield and optimized for aesthetics. And could we do better? And enabled by the miracle of transportation, fossil fuels of the industrial evolution. The average apple travels more than 1500 miles before it gets into your mouth. That's an extraordinary thing. Walk us through what we're seeing here in this map and why that's relevant to your work. Yeah, so you got a toddler aged apple walking a really long way before it gets to you. If you look at this map, anything that has a color, okay, there's grays, and then there's red is the first base color. And red is under every other color, the purple colors. I know it's a little confusing, but just stick with me. Anything that has a color is considered food insecure by the World Health Organization, meaning they don't produce enough food in their domestic boundary to feed their domestic need. Any country that's both red and purple means that it's a country that's both food insecure and producing food for other countries. Of course, in the United States, we're looking into Mexico as our climate is moving. China is now the largest landholder in Brazil. The Middle East gets a lot of its food from Western Central Africa. So this global farm is the way we feed ourselves. It's an amazing thing that so many people are alive and overlaid on top of that are the products of the Industrial Revolution. The ability to be as far away as possible from where our food is produced. So let's flash forward to a future vision and see what that would look like. Well, so we look at this and we see all these lines and we wonder, you know, where are those lines connecting? And those lines connect primarily to large cities. Let's go to the next slide. There we go. There we go. Q. So the idea that we had was, you know, could we actually start sending more information about food to the places that it's being grown to optimize it, or to the places where it's actually needed? Could we start to transform an industrial or an analog system into much more of a digital one? So we could imagine then a world just like Willy Wonka? Yeah, so, you know, in the best Willy Wonka of all time, no offense to that. Gene Wilder. We are Gene Wilder, definitely. OG Willy Wonka. So he zaps the candy bar, right? Turns it into particles, sends it into the air and reconstitutes it on the other side. I mean, we live in this world. Our digital physical world spans dating, spans food, spans all of our social connections, spans our information, spans our cars that speak to each other in the planes above. So is it so crazy to think that we might take a physical object, a plant, digitize it or encode it, be able to send it digitally somewhere else in the world and reuse that information? Sounds fanciful, but maybe it won't be just a sugar ball if you do it that way, right? Yeah, we'll see. All right. So the apple, the gulf of space and time that separates the food production from the people, it's kind of symbolic of a larger thing we call, you know, the food crisis. But what's interesting is, when you start thinking about the food crisis, it's really multiple crises and the crises seem to be somewhat contradictory. Walk us through this myriad of crises and how you view it. Okay, so if everyone would close their eyes and hold hands with their nape, I'm just kidding. But if you close your eyes and I said food crisis, it's likely that the person next to you and the person on the other side would be thinking completely different things. This is the last 10 years, 15 years of the illumination of all of the myriad of crises. We're talking about places that don't have enough food, places that have too much food, algal blooms that are creating toxic oceans, not enough water, GMO being the savior of the world, GMO is a curse and a blight on the world, land taken by farmers that that land was originally taken from a rainforest in the first place, pollution, nutrition, attenuation, urbanization. It's this kind of long litany of things that really is disempowering. You know, it takes the individual out of the equation and it's also not really seen, you can't see a derivative, something that you could put your effort and life towards as an individual that could attack a lot of these things. Yeah, it seems like too, when you address one problem, you might exacerbate another and that makes it more difficult. And couple that with the fact that there's a tremendous amount of misinformation out there about this. Sorry, this story is great. So previous administration, I was hanging out with the Secretary of Agriculture and he showed me this survey of what Americans would support or oppose in government policy related to food. And it was all the things that you'd expect, all the things we argue about, tax on sugared sodas, some people say yes, some people say no, we don't like to be told what to do. But down the list was a control question and it said, mandatory labeling of food containing DNA. Eighty percent of the surveyed audience would ask for mandatory labeling food that contains DNA. And so I think we're in this age and we need to just be self-reflective and let's just put it out there. Most people don't know anything about what they're eating. They don't know necessarily that there's DNA in their food. To have the conversation about GMO will take you down a wormhole that's going to be crazy. You know, so being able to bring people outside of our world into this new era as science is exploding and create some literacy in that area so that we can have intelligent conversation is one of the biggest challenges. So time for a revolution. And it's probably, I'm a history major, I will confess to that. I just play a scientist on TV. And so let's talk a little bit of history. Why don't you give us a little thumbnail of the previous agricultural revolutions and how that leads us to today. I think the most amazing thing, and again we all know this in this room but most people don't when I talk to them is that everything about our society comes from food. You know, the fact that the first revolution was hunters and gatherers now become, you know, we domesticate crops, we can stay in the same place, we can hang out with each other more, we don't have to run after food so much, we can run after each other and the population explodes. That's the birth of civilization. The next one, the industrial revolution, obviously the ability to get very far away from our food which was a great thing, you know, the ability to build up cities and then continuing on in the third revolution with the full vertical integration of farming. And I think that this also still includes biotechnology which is hot on a lot of our lips right now it's very squarely in that kind of yield at all costs, you know, support the huge populations that everyone said weren't going to make it in the 60s and 70s and obviously have made it today. I think the next one is about something that is seemingly boring but actually exciting which is accounting. It's a full accounting of what did it take to grow this, what does it give me in terms of human nutrition, what was the environmental cost and how good is it for me when I get it on the other side. So it's really about a full accounting that takes in, you know, the environmental, the human nutrition and everything else and that requires tons of information. It gives new meaning to the term bean counting, doesn't it? It's little bean counting. So anyway, there is widespread agreement, there's lots of evidence that this fourth agricultural revolution is already happening and some of Caleb's colleagues here, some of them in this room are spending a lot of time right now thinking and envisioning the future of food. You say you want a revolution? Well, it is happening right now, down on the farm and you can count agriculture in to digital technology. But why? Because humanity is facing a huge existential question. Can we continue to produce enough food to feed the populations that we have to do it in an environmentally sensitive, sustainable sort of way? The last agricultural revolution in the 60s and 70s brought chemistry to farming that are pesticides and fertilizers and genetically modified crops. Yields far eclipsed malthusium predictions of too many mouths to feed. But farmers are victims of their own success. The glut makes it hard for them to turn a profit. All those chemicals have huge environmental consequences and of course the climate is changing. I think we are at a point where we cannot go backwards. We can't unroll the changes in climate that have happened. They have happened. It means cotton now grows in Kansas and the corn belt is migrating toward Saskatchewan. So what to plant and where and how best to keep the food growing and flowing to market. It's about understanding your field environment so you can plant the right population at the right place at the right time to maximize the likelihood that you'll have the best possible productivity. But how? Tap into some other revolutions. In sensors, big data, artificial intelligence and machine learning. Suddenly farmers are focused on another kind of cloud and just like Netflix and Amazon know what we like there might be tailored ideas in the cloud for farmers. We're looking at parsing things out to get closer to a more individualized answer than we have in the past. There will be 10 billion humans living on earth by 2050. Feeding them all will not happen without this revolution. The crops of the future are probably going to look very different than the crops that we have today. They might look similar but they're going to have all sorts of technological enhancements to make them more efficient. So when Mitch Teistre says that the crops of the future will be different than they are today give us an idea of some of the technologies that are at play right now that are going to enable that in some fashion? I think first the recognition that we are at this moment of kind of biodigital convergence. Biology by design in a way that the world has never seen with the access to data, the access to tools, the access to the cloud the access to sequencing, it's going to produce amazing things. And of course we're seeing that with CRISPR but CRISPR is almost an old debate. It's already litigated between a couple of places on the east and west coast. Not to name names. But we're seeing the CRISPR but then the scientists are already way ahead. We have CRISPR but now we have gene drive and gene drive is the ability to take an edit and force it across into inheritance which has huge moral hazard. And I mean identifying that in the scientific community relatively fast the development of daisy chain came after that, the ability to edit something to force it into inheritance but then to spread it across chromosomes so that we know in successive breeding pairs exactly how long it will take to get the edit out of the population. And what does that mean? It means the potential end of malaria, the potential end of Lyme disease crops that are climbers just in drought tolerant more effective use of soil microbiomes there is so much exploding kind of in that space but it all revolves around this idea of kind of bio by design or the bio maker. Well as we've recently seen CRISPR also can provide designer humans and there's some moral hazard there to say the least. When we look at designer food coupled with designer consumers of that food that can be pretty exciting too, can it? I think absolutely. We have a project together with Harvard and Northeastern that we're calling designer self at the moment and what we're looking at is we're taking food that people eat which is these little pictures of icons of food doing gas chromatography, mass spec, metabolomics breaking it down into all of its chemical constituents just to take a side step on that nutritionists typically follow about 150 chemical compounds and something like broccoli there's over 20,000 the knowledge and the ability to gain 20,000 chemical constituents per sample hasn't been around very long. So we're working on building this library building up all of these barcodes of chemicals and then matching them against our own individual genomes for this idea of you know if you're genetically predisposed to Alzheimer's, diabetes, heart disease that you could be eating for that proactively and not just that but that we would actually be designing food systems that had a quantified amount of benefit. So we talked earlier about how there's a lot of misinformation out there but there is growing awareness among people about the beauty of this and I wonder how much... Well it's just incredibly complicated right now because every book that you read, every blog that you see every new thing that comes out tells you to do something different so I think it's time to get past that. It is and so how in part of this whole mix are the retailers, the groceries are they responding to this at all? Are they aware of it? Sure I mean you know as an aside my family's background in retail grocery and retail grocery is going through something that we've never seen in our lifetimes and I think it's fascinating and awesome. When you have Amazon buy Whole Foods and then the next day Walmart kicks all of its vendors off of Amazon's cloud the cloud now becomes political the cloud now becomes a point of contention of data of customers, their habits, their preferences and they have to get separate and how they want to receive food is also part of that so then of course we have the rise of the meal delivery services of which I am a partaker note to the wise start them when they're... you start using their services when they're young companies when they're trying to do customer acquisition when they switch into profitability it's not so much of a fun thing either way these kits are coming about and then you have most recently Walmart forced all of its vendors onto blockchain which I'm not sure that they even know what that means yet and they know exactly how it's going to work but this is all driven by three simple questions and I mean a lot of our funding comes from corporations in the food industry and we can talk about that but it's three questions that are rocking the world right now where did my food come from how good is it for me and what did it cost the environment they have to build an entire information ecosystem to answer those questions and that's what you're seeing So what about the poor farmer and all this how is he or she supposed to anticipate this kind of demand and do they have the tools in the toolbox are they available to them to really dial in what they produce I think we're getting a described field that we've never seen before an agricultural field this particular example is from a startup that now in the last five years has more infield data digitized than the USDA has in its entire history you know you have an 80 acre field here that's overlaid, satellite data, soil tillage maps yield data, whatever kind of data it's coming off of John Deere and Kubota and all those together with a lot of GIS and image based data and all the other things to where now this 80 acres can be compared to its neighbor's 80 acres to the 80 acres in six counties over and it's giving the farmer a real insight into things like cost of goods you know the way that this startup makes money is by simply having a marketplace that it offers inputs to farmers why can it do that because it knows exactly what the inputs are that everybody's using and then it looks at what the co-ops charging them and co-ops go 20 to 80% price differentials so just as you start to bring illuminance to the field illuminance to these practices you start to put a lot more power into hedging into commodity purchasing than kind of has been available on the farm before so we're talking about phenomes how genes express themselves based on their environment I guess in the old day phenotyping amounted to a farmer kind of walking the field using his eyeballs and you know kind of squishing the leaves between his fingers that kind of thing but today with better computers better sensors better drones you name it all that long hard work that used to be done in the field is becoming much more precise much more automated and there is a world class example of just that right here at Purdue take a look at Purdue's controlled environment phenotyping facility the plants don't sit still for long they are shuttled around imaged and mined for data the goal to understand how genes and the environment interact to create physical traits in plants their phenotype normally this takes many seasons many years plant breeder and geneticist Mitch Teinstra is the scientific director of the Institute of Plant Sciences what the controlled environment phenotyping facility allows you to do is to actually control the environment to understand plant biology and then actually develop hypotheses around what happens to these crops as they are being challenged by different types of environmental factors they capture images of the plants in a broad range of spectra most invisible to the naked human eye it allows them to quickly answer a lot of important questions we have all these computers but we really haven't used them to change the way we manage we've used genetic information to change the plants that we grow but we haven't used environment and management information to change the way we manage to the same degree engineer Jan Jin says hyperspectral images can help farmers identify diseases and pathogens much sooner we demonstrated that with hyperspectral imaging we can detect some of the pathogen diseases in the field and in the greenhouse at least 5 days earlier than before they are visible for human eyes Jin wants to deploy this technology in the field actually over it on drones we use this country to play the role of the drone in his greenhouse at Purdue he uses a mock drone to gather hyperspectral images they can uncover key insights about how well plants are doing with better eyes and a smarter brain we can see it better Jin believes getting these tools into the hands of farmers is of utmost importance in 2050 we may reach 10 billion people for the whole world so in order to feed more people we need to produce more food we need some tools to understand the growth stage of the plants better it's time for another agriculture revolution it's going to be digital agriculture and it's not going to be because it sounds cool it's going to be because we need to enhance agricultural productivity in the least environmentally impacting way so you get the sense this is still kind of early days with all of this and anytime you're involved in science of course you want to control for as many variables as you can and that's part of why you started thinking about growing things in boxes and so it's not just any box though why don't you introduce us to the personal field computer and how that all began for you so I was part of a group that was kind of sent from MIT to go to Fukushima after the disaster and my background is as an architect and an engineer I designed data centers and I designed hospitals professionally before falling backwards back into school so we went out with this kind of motley crew and we weren't supposed to think about food in a long time we were supposed to think about cities and I happened to take a walk while I was there and I noticed these folks that were starting to grow food again in this completely obliterated city and they had a lot of questions about radiation contamination about salt contamination about proximity to the disaster and would the rice be sellable anymore from this area anyway this was the bread basket of Japan and I started following up that more and they import an amazing amount of their own food so there was this real kind of convergence for me of thinking okay well if you don't have the climate that you need could you start to design it and since I'd been designing climate for computers and for humans I thought that this might be possible I got home I was by myself I had no funding very very few friends in the plant community at the time and I started with a couple of Dixie cups and some electronics and I killed thousands of plants you know I'm from a firing background I'm here in charge of every aspect of a plant's life and you fully appreciate how difficult that is and so I started doing that and I thought oh maybe I'll make a product out of this you know the media lab one of the directors of the media lab is the founder on Kickstarter so we thought oh we'll throw it on Kickstarter it'll sell for $29.99 and then I thought it'll probably land up in a landfill and people won't really like it so that didn't really seem good to me and so I started thinking okay but one of the issues was when I got online you had an incredible domain expertise and then you had this kind of general user base that was experimenting with all their own systems but not building any knowledge together so I thought well could I make a device in between so that I could start gathering enough data to just do what I wanted to do so that's kind of the lineage of what we call the personal food computer and that kind of took me on a journey and that journey next stop on the journey was well if we learn something in this food computer you know how will it scale is it four plants or could you do it with more plants at home brewing our own environments with a nice few crazy philanthropic grants at the beginning and you can kind of see this progression from about 2015 to maybe 2017 and beyond where we took these ideas and started kind of scaling them up more plants, more data, more experiments to see what the relationships would be and most recently we've actually been developing much larger units that we were asked by a sponsor at trees and so the idea was on a paper and they said could you build a tree computer and so if we can go to the next slide oh well here's inside of the food computer so it gives you a little bit more of an idea of what's going on and we'll talk about all the tech inside but we'll go to the next slide the tree computer so most recently we've built like these three by three meter boxes these six by six meter boxes our biggest challenges in those cases were can we produce the spectrum from sunrise to sunset can we control the heat when you're producing not much light and could we model the environment that a tree would be living in to learn about it so these are kind of the devices of what we've been building lots more to discuss here a lot more questions on my mind but first let's talk from whence you came here that great agricultural institution MIT how was it back in the day so tell me about the MIT media lab first of all just a few words on that for those who are uninitiated well you're the first person to grow plants there as one of the labs there it's an iconic plastic group of people isn't it the first time I met the founder of the media lab Nicholas Negroponte he came in and said what's a garden doing in my media lab and pretty sure that he was ready to kick me out I had been under the radar at that point but the reason that he asked that is the media lab is a collection of 30 labs and none of us have really anything in common these are new cars, new toys new prosthetics new methods of neural connection all kinds of things but what does unite us is this culture of what we call demo or die and it is that no idea is too crazy no idea it doesn't even matter if it works and is going to create a benefit right now but if you have the idea you have to build it and so we share all of our knowledge and there's groups that focus on space exploration but also on camera imaging for autonomous cars and so when I need imaging help I go around and we ask each other questions the media lab is 90% corporately funded so we're at that kind of interface of 80 of the world's largest companies but also academic research and so we get a little bit closer to the applied layer of new technology development so it's not just AI for AI sake machine learning for machine learning mechanical engineering it's actually the current director calls us the anti-discipline and I think the institute would say we lack all discipline all together but it is that idea that we're in the white space and we're applying the tools that we have to try to solve problems that are a bit more across a lot of domains so you have a little bit of anti-discipline in your own personal story don't you and that is and you also have a connection to the land your family as you've said your father is in the grocery business and you also have and there you are by the way the red arrow that's Caleb on a horse that was last year you must have eaten some of those vegetables out of your food computer or something in any case tell us a little bit about your back story and how you got into this okay well I was pretty much kicked out of high school I attended less than 40% of my senior year I've never been a very good student but I was always building things and that's why I thought I should kind of become an architect I ended up studying art and architecture and engineering and then one day just showed up at MIT literally without ever applying thought I should just be there and showed them what I'd been doing and thankfully they could take a joke and they let me in but in my background my father's in the grocery business his whole life my great great grandmother came over during the land rush during the original settlement as a mail order bride from Germany and she actually came over with just this trunk which you can see here to her name she's our original explorer in the family and ended up on a large settlement in Kansas farming so we've been farming that land for a very long time but then my family kind of said to me what I think is retold a lot in our generation which is don't do farming you know do something else do technology do something that is different get out of the field and I think that is kind of a message that's been told quite a lot and so that's the path that I followed but I've kind of now come back and that kind of brings us back to the boxes of technology and agriculture give us a few more details now on how it progressed you took us into the scaling component but you know take us under the hood here a little bit and explain what's going on so what you're looking at now is one of the latest versions of what we call the personal food computer it's been designed specifically for STEM and STEAM education so we don't anticipate producing a whole lot of high-grade science obviously the controls are very limited but we've scaled it down for what you saw before to the scale of the audience so that they can learn coding alongside electronics alongside plant science and chemistry all of the data that this bot creates goes into our cloud so anywhere in the world that one of these lights up all of the sensors inside the cameras inside get networked and these kids and other people frankly we'll talk about it later can share their experiences so it's a really distributed network isn't it that's the plan so I know you spend a lot of time tasting the occupational opportunities you have as these plants go into these environments and you stress them in various ways what are you learning about what happens to taste and what it's like what the experience is like so back in the early days when you're starting your own lab and you have no funding and you're by yourself and you're bringing people in and you want them you know in my case I'm like a UFO in the media lab and I'm not a someone that fits in a great and ag school not great in the tech school I'm just silent by myself and people always want to eat things that were in the lab and I would always be nervous that it would taste terrible so I would eat it first and I ate a whole lot of lettuce just a ton amount and I ended up with this tongue that can taste pH within 0.1 or 0.2 so I would be able to say like no you can't eat that today and after evolving that kind of palette I noticed that every time I would take the plants out of the environments and then I would serve them pop up kind of pick your own lettuce patch for nerds they would taste very different when I brought them back the next day and I was learning something that a lot of people in the audience I'm sure know that a plant under stress has biochemical defense and that biochemical defense in this case I was triggering a drought stress and that drought stress was causing a sweetness to be present in the plant the next day and it blew my mind that within one day's time I could shift a basil into a spite more of a spicy state, a sweet state and so on and so we've really been exploring that much deeper now which you'll see in the following research but that's where it all started So stress equates to flavor, that's really fascinating I'm sure you all know that Except in people I might taste different if I was under stress I know you spent some time going to Svalbard, Norway the big seed vault there exploring cultivars which have been kind of orphaned along the way ancient cultivars what that is like when you kind of revive a 150 year old tomato for example and in this slide the journey started by talking with a bunch of food companies in this case it was Driscoll strawberries and General Mills broccoli and we started doing some trials with them with our technology to see how did it compare to field grown how did it compare to greenhouse grown and then I learned this shocking fact I started working with Campbell's in my lifetime I'd eaten about four cultivars of tomatoes most likely and then I started looking into it and there were tens of thousands of cultivars of tomatoes and so I started to wonder why did we pick the things that we picked and of course I think in this audience we would know that that's about shipping that's about disease and pest tolerance drought tolerance, it's not much about flavor at all and it's not much about nutrition at all and so I started thinking could we access our genetic inheritance could we access seed that's built in technology over the last 100,000 years of evolution and that's when I kind of came to this rare and ancient seed bank and so on the next slide we grew out for example these tomatoes these tomatoes hadn't been commercially grown in about 150 years so we were probably the only people definitely in this room that have ever eaten those tomatoes unfortunately we're a bunch of scientists and we don't really know how to cook this was apparently a soft tomato we ate it raw it didn't taste good you know it started us thinking about if we weren't optimizing for a harsh world with genetics could we optimize the genetics with a perfect world and what would that look like these tomatoes for example carry more lycopene than the traditional tomato that you're able to get in the grocery store which has been shown to have a lot of links for brain disease and so the question that we have is can we go through that as a matrix and see what do we actually have from our genetic inheritance that could produce chemicals that would be much better for us you know as a health is it practical to imagine when you're thinking about you know planet with ten billion people to have that many varieties though beyond four varieties is that possible I think it's possible to look through and use the lens of nutrition to select differently and then to use our enhanced ability to design biology to make it more field ready and so no I don't think there'll be ten thousand varieties of tomatoes in the grocery store but I do think you'll go to the grocery store and you will select and pay more for a certain tomato that does have that enhanced lycopene flavor or for a different object that has something that you're looking for so I do think we'll learn to differentiate the things that we eat based on you know what it'll do for us Alright more about the box and how it works you talk about encoding, decoding and recoding plants what does that all mean so encoding is what we've basically talked about to this point which is building an environment that can digitize a plant's experience so we're encoding that plant the decoding part is all the different data streams and I know a lot of people talk about big data and what it's going to do so I just wanted to show a small subset of some of the data that we collect so you get an idea the first two buckets you know atmospheric data temperature CO2 spectral composition this is all very cheap very easy not very novel and the cost has been brought way down for all of our environments then we do things like tissue culture water culture to get to the mineral contents we do sequencing of the microbiomes of the plants the root microbiome, the tissue and the surface we do gas chromatography and mass spec which you're seeing on the other side if you're not familiar these are a lot of secondary metabolites from plants they represent flavor compounds and so what you end up with is these two buckets of what did you do to the plant and then what did you get in return and the kind of the thing in the middle is the image and so the next slide yep so this is just a subset of some of the student projects that have been going on we also break it down by spectrometer images hyper spectral images but it's really that idea that you have what did you do and you have what did you get and you have this image in between that can gather a lot of interesting information so this is the basis of all of our machine learning projects which is using images and causes to determine effects and not just determine but predict effects if you can say these combination of variables over time that look like this will result in that chem score or that protein profile or that biomass or color or texture then that's all part of the project that's all part of decode take us to the next slide we'll talk about this one okay so this is what you learn when you really try to do this these devices that I've shown you and they could be any devices the whole point is that they're modular but there's tiny ones and there's middle ones and there's big ones and I tend to design and develop the specific devices based on the funding that comes in and what they want to grow but then you have this kind of modular layer that we call a nervous system or a brain which is all our ability to have microcontrollers microprocessors doing actuation and sensing and getting that information pushed somewhere and then we have to build out this whole cloud architecture it's one thing to talk about big data it's another thing to handle the pipeline of big data which is really challenging so we partnered with Google on this cloud because you know it costs money to put things in the cloud and when you're banking that many images and you're trying to give people real time access to them meaning read writes between all these different data warehousing techniques your bill can go up incredibly fast in fact we were hacked and in two days time our system was bitcoin mining and they brought it it was like a $15,000 bill thankfully Google also has a sense of humor but this is the kind of ecosystem that you have to create for this kind of project and I don't think people really think through the infrastructure layer very often they very much jump to the app they say I want to work on this crop for this reason and it's like great you should do that but to be able to do that for a much broader audience you kind of need to be playing in this game let's see how you visualize all this data explain what we're seeing if you've ever seen like a machine learning visualization what you're looking at is regression modeling so imagine that you have three variables in this case we do I think it was like type of light the UV period and the photo period and you have that amount of data that you've gathered and then you're trying to optimize for some target in this case was red colored and so what we're looking for is the machine is going through all of these combinations and attempting to use deep learning to elicit what the prediction would be about which combination of variables you would have to have to optimize for your target I think we were doing biomass or essential oil and basil in this one so what it's really allowing you to do is skip a ton of experiments because you can see as it goes through and it looks and it starts to find imagine machine learning as kind of like a toddler in a dark room with furniture and it doesn't know where the light switch is so it's going to need to bump into all of the edge cases in fact in our initial studies it returned results that said turn off the lights and if you turn off the lights the plants would die but it needs to know that it would hit a boundary condition to become smarter over time and so each successive experiment doesn't just get smarter towards our one goal it gets smarter around a whole myriad of experiments that we didn't have to do so encode, decode, recode, let's talk about recoding bring it to the world well I guess we're still going to talk about a little more chemistry first so just to give you a concrete example we published this in PLOS recently which was to show the kind of whole framework of control environment to machine learning to prediction to control my environment and validation and so what you're looking at is that surrogate modeling but for a specific set of metabolites so we were able to show that we can use machine learning to up-regulate individual metabolites or groups of metabolites as we want so which brings us to Bazel let's talk a little bit about Bazel what it can do for us Bazel is our lab rat everybody has to have a lab rat so I'm going to show you that the lab rat is a lab rat so I'm going to show you that and then we're going to go back to the lab rat and then I'm going to show you a little bit more about how we're going to use these kind of different materials so I start dealing with all this data and it quickly gets out of your control especially as you want to understand what other people have done in the space and across different disciplines and so I partnered up at Northeastern with the most cited author of all time on network theory all the known associations with diseases from the Harvard Medical School database. So what we found is a simple basal plant has about 300 chemicals that associate with 700 different therapies, disease therapies. And so it became very clear to us, one of the folks on my team is an ethnobotanist. And we all are jealous of the fact that he has this kind of like Indiana Jones, I go down and I find this plant in the Amazon. And so he had all these projects and he brought this knowledge to us that 80% of the world relies on plant-derived medicine. So we started thinking, if we could find the chemicals that were associated with these disease states, and we could start offering practical advice for healers, for, you know, it eventually turned into what you'll see is things like drug discovery, but that you would have a huge impact when you understood the simple thing like basal at that level. But imagine that basal is just that one app. It's just a quick growing plant that we chose. You could do this with almost any other plant. So let's go wind it back just a little bit. The origins of these controlled environments, actually some of the big impetus was out of a desire to figure out ways to grow food in space. And one of the great innovators in this world has been here at Purdue for nearly five decades and he has some amazing insights and how the effort to grow food in space is helping life here on earth. Deep in the vortex of the Indiana winter, Kerry Mitchell has a ready escape to warmer climbs inside his greenhouses. It's very therapeutic, by the way, to go into the greenhouse. It really is. He should know he's been doing it all his life. The son of Illinois greenhouse growers, he was apparently born to be a whole plant environmental physiologist. He first planted roots here at Purdue 46 years ago. And I'm beginning to get the hang of what you have to do around here to be successful. His secret, aim high. So in the 70s, he started working with NASA to find ways to grow food in space. Long before Matt Damon discovered the scatological secret in the Martian. Your hands are Jesus. In the early days, the problem was more in the realm of Star Trek. Scotty, where's that power? Coming, sir. The lights we had to work with were like hot street lights and they were very wasteful of energy. So when you do put pencil to paper and you calculate what it would take to support a crew of eight people in space, it just did not compute at the time. But that changed with the advent of a much more efficient source, the light emitting diode. LEDs really are a game changer for controlled environment ag in all of its forms. Over the years, he has helped change the game even more by designing novel ways to deliver LED lights to plants. He first created something he called lightsicles and now is refining vertical bars of LED light that plants grow up and around. The term of art, intracanopy lighting. Putting light on a bar like this vertically, what's the idea behind it? Well, it's so that we don't have mutual shading of middle and lower leaves by upper leaves, which is what happens when your sole source of light is the sun coming from overhead. Along the way, he started playing with color, understanding what hue can do. It turns out LEDs don't just save, they can also enhance depending on the color. This allows one to almost develop light and growth prescriptions for growing different crops indoors. Carrie showed me some of the leaves of this labor. Some reddish lettuce, they are growing and savoring on the International Space Station. Interestingly, if you don't have blue LEDs on them, it just grows green. But if you have a little bit of blue, it brings out the pigmentation. And it makes it more nutritious. All of this thinking about how best to grow food in space has triggered a terrestrial revolution. It's the dawn of a new age of controlled environment agriculture. One of the reasons I'm still in the game after 46 years is that I'm finally starting to see some of the fruits of the efforts that we started back in the 1970s and 80s. It's been a long time, but it's here. All right, so we've been thinking about a lot of future ideas, talking about the future a lot. Let's a little more brass tacks with you and some of the projects you're involved with right now. Tell us what you're doing. Sure, so a few years ago now, one of the member companies of the Media Lab came to me, it's one of the world's largest cotton manufacturer in India. And they said, the water table in India is dropping feet per year. The quality and consistency of Indian cotton is threatened. We've also know about this cotton variety. It's kind of an ancient rare seed that produces this amazing fiber that you cannot grow in field. And so we have this idea that maybe someday we would have a factory that grows cotton, gins cotton, mills cotton, product makes and shoots out the other side. And it won't be for all of the cotton products, certainly not the basic cotton product, but for an advanced textile that would have a natural fiber within a kind of advanced fiber length, staple strength, all that stuff. So we started off, whoop, we go back a little bit. So we started off with simple design of experiments, proof of concept. Could we grow cotton? In this case, aeroponically, what would be the quality of that cotton and how would it compare to things that are already grown? So I started with a quick dose of cotton education going down to the Mississippi Delta, working with University of Arkansas and then using some of their genetics so that I would have something to compare against. And so we grew out the cotton, did all the tests and used some of our modeling to actually prove that we could enhance the fiber length, fiber qualities, micro and air, a bunch of different attributes of the cotton that they were already growing in field. And so once we got past that hurdle, which was, can you grow it this way? And of what quality and of what yield? We've now deployed this technology out with their factory on the border of Pakistan and India, which is about, you can be in their factory, for example, going 45 minutes straight in a golf cart and never turn and you'll be indoors the whole time. And this is where 70% of the sheets and towels for the United States are produced. And so what we've transitioned to now is another core tenant of the Media Lab, which is deploy or die, which means you have your research, you've demoed or died, it has some significance, now get it into the world for testing. And so what we've done is built up a small cotton lab onsite their factory, which now they're training employees on how to run and the next step is 10,000 square meters inside of their factory for production. So ideas like this are great. It's kind of, you know, it's a specific kind of pinprick. Give us the big picture. Go from micro to macro. How does this affect the whole world? Well, I think a lot of what people think about in this space right now from a commercialization perspective is what I would call climate manufacturing. What can you grow in a box at a profit? But there's a whole other use case that we were interested in a while ago now of what could you learn in a box that you could apply? So we did this very basic test, proof of concept, where we took all this weather data. What you're seeing is little weather satellite stations that have been collecting weather data for 30 years called the EPW Public Weather Database. We scraped all that weather data and we formed it into what we call our recipes or the input states to our machines. And then we passed it through our cotton model, which we'd been developing. So if you're following now, we've got a fake model using real climate data from the past 30 years and the visualization of the learning looked like this map. And it said, these places in this kind of whitish yellow color should be the places that you look for highest yield and quality of cotton. Now, we didn't know if that prediction had any weight at all except we had the FAO database that is also open source in public for the last 30 years of cotton yields. So when we cross-referenced what our model has suggested would be the thing to do with the data from the FAO over this time period, we were about 75% correlated. Now, anybody in statistics knows that's not great, but this was just a quick test with a few smart people to see what do we think about this idea of climate prospecting? Could we use these machines to build links out into the field? So is it possible to, for example, come up with better Nutella? Yeah, so the good thing was we did that little test and then we found an incredibly open-minded and supportive sponsor in Ferrero. So if you know Ferrero for Nutella and for Kinder chocolate and for those precious Ferrero Rocher candies, and if you wanna learn some crazy stuff about Rocher candies now special that Hazelnut is, I know that's an enticing offer. I'll tell you about it after the talk. But basically here's what Ferrero was facing. They buy 50% of the hazelnuts in the world. 70% of them are produced in Turkey. Turkey is experiencing a lot of climate issues, a lot of financial pressures, and so one idea was could we use our machines to predict forward and assist that Turkish farmer in production? And then there was a second idea. Hazelnuts have never been produced in the Southern Hemisphere. So if we could identify a cultivar, that could withstand the edge cases in the Southern Hemisphere that would still be productive. This would give Ferrero two growing seasons, which may sound like a very simple thing, but that's billions of dollars for that company. So it's actually a huge problem. So we took our idea, we started with a white piece of paper and that was the tree computer project that I showed you. We started with all the places that they thought they might like to grow hazelnut. This is kind of from the company perspective where they said, look, this is where we think land is available, cost of labor, cost of inputs, all of that. These are the places we like you to look at. So then we identified kind of the edge cases of those conditions, these kind of high temperatures and these low temperatures. And then we ran a genetic screen of different cultivars of hazelnuts inside of the tree computer. And we mentioned the LEDs and Kerry Mitchell's work and it's something that one of my colleagues who's hanging out here today spends a lot of his time working on, which is emulating sunlight, so that we could actually get, I think we're about 95% in emulation now inside of our environments in color and intensity from beginning of day to end. So all of that is to say, we've been designing and developing these kind of robotic control environments. We've been screening these genetics against the edge cases. And what we found in this little graph, it's most important to just see the little orange line. What we were following was photosynthetic rate of all of these different cultivars. And we identified one of the cultivars that actually seems to thrive, seems to not only survive but thrive and of course the other ones decline. This is a cultivar that has never been planted in the Southern Hemisphere, was not identified previously as a good candidate. But Ferrero and other companies doing this type of work usually have to, this is a five year growing tree crop. So they have to make a bet on some cultivars that are unknown, build up a whole plantation and wait five years to see if it pans out. In this case, we're able to identify and de-risk their ability to start doing outgrowing projects in these different geographies. Sweet. All right, so one of your key ideas also is to democratize all of this. Tell us a little bit about OpenAg and what that is all about. So I think a lot of what we were talking about from the consumer side is driven, I said those three questions, but if you summed up those three questions, it would be trust. Everyone's trying to acquire trust from the company side. As far as taking money out of R&D of the big five food companies moving into venture capital buying startups, it's all in this quest for belief and trust and getting that consumer back. So one of the things that I think is if we're gonna progress with this science and the current stigma of science and food are bad, we need to bring the public along with us. We need to democratize where this technology is going. So on the very simple front, a few years ago, we launched out on a simple Wikipedia page, all our bills of materials, all of our diagrams, the horrible video support tutorials that we had at the time, and we simply made it available. We didn't do any marketing, we didn't give anybody any money, and all of a sudden this little community sprung up inside of a discourse forum. So if you're not familiar with discourse, it's kind of like Reddit. Every user can come online, it's for free, they start their own threads. If you look through these threads, they're as simple as, where are you building your food computer? Or there's this one down there called Monkeyman TV Build Series. It's this guy that finds enjoyment about building the food computers in a chipmunk voice on YouTube. Tons of followers, right? Like, imagine that his gift is communication. And he's using his gift on this thing as part of the community and producing a lot of awesome videos and inspiring a lot of folks to get involved. But it goes down to the granular level of folks just interested in the electronics or just interested in the bio aspects. And then they find community with each other and then they start sharing their knowledge, their journey together, and they start answering their own questions. So it's the nerd farmer nation. And part of that, big part of that is getting these devices into classrooms. Tell us about that. Yeah, so as we start releasing this, they start popping up everything from K through 12. This was in Boston. In the middle, that's, I think it was like a ninth grade in Baltimore. On the bottom, this is a professor in South Korea that's adapted a whole food computer curriculum course for mechanical and electrical engineering. It is really start to spread on its own in a very individual way. Let's say just driven by these passionate and curious people. And then after that individual thing happened, let's say for a couple of years in the community, there started to spark kind of this more collective learning. And so what you're seeing here is what we call homebrew food computer club. So if you know homebrew computer clubs, this was like the 60s and 70s, people in their basements, weirdos working on things that people didn't understand and sometimes they couldn't articulate the value of. But it was driven by passion and curiosity. And you're finding that same thing in young people now. They wanna be involved in food, but they don't exactly know how to be involved in food. So as you build this device that creates this ability for them to experiment and explore really awesome things are happening out of it. So I'll just let one speak for itself. One day, I was having a philanthropic donor come in and I said to my community, I need videos. I need your best videos as quick as you can. And so I got all these videos and some of them were from Doomsday Preppers and were pretty weird and not the best thing. No, no, not the best thing for donors. You didn't wanna show the chipmunk guy. Yeah, well, I mean, he's awesome. But we got this video come in. This is a 12 year old doing this project. And I think I'll just share his journey a little bit to give you an idea of what it's like. Building a food computer was an incredible experience. I worked in almost all the building steps and I learned many things. The difference between AC and DC and between sensors and actuators. I worked with wires, relays, and the voltmeter. I installed sensors for measuring water and air temperature, light intensity, CO2, pH, and so much more. Wiring the mega proto shield, now that was tough. My dad did all the solding. We even had a little burning due to a short circuit. I installed the LED panels in webcam. I had my first experience with Arduino and had lots of fun testing all the sensors and actuators. And my dad took care of the Raspberry Pi and its software. It was my first experience building a piece of hardware. So, you know, that's magical though, when you see that, right? Yeah, we sent him the t-shirt because we were like, thank you. Yeah. But he deserves a swag. Yeah, yeah, 100%. But, you know, this generation has access to things like thingiverse, you know, has access to things like, you know, big tutorial websites to teach you how to build anything. You know, it's very much ingrained in that kind of maker culture. And so, bringing that into agriculture, you know, if I would have asked half the kids that we work with, you know, do you want to be a farmer? Do you want to do agriculture? I mean, they don't honestly, most of the time, know that they ever eat plants. They say that to me all the time. But then when you show them this and there's consequence to code, there's consequence to engineering. I mean, I kind of wish I had that and I probably would have gone to bio-class. Yeah, me too. Well, give us an idea of how far it's sped. That gives us an idea, right? Yeah, so in the last couple of years, and like I mentioned, not giving anybody any funding. You know, people constantly tell me that I can't, that our group can't do what we do. And one of the first ones was, you'll never get this across language barriers. And you're only targeting this at this demographic of people with this disposable income and it'll never go anywhere. So we just did it anyway. And across, in three years time, we have food computer builds in 65 countries across, you know, all kinds of language barriers, not just coding language barriers, but actual language barriers and how they get the parts and where they source everything. And this is about 3,000 people. That's exciting. And tell me, going back to the classrooms for just a moment, I know it can be difficult to integrate something like this into curriculum. Is it hard and obviously there financial barriers as well? Yeah, I think, you know, the difficulty has more been that the way we teach is so isolated, that each of these disciplines are taught, you know, in and of themselves. And so this device really sitting at the middle of chemistry, biology, coding, electronics, mathematics, you know, that's one of the bigger challenges is how do we develop the curriculum to integrate this into schools today? But I can tell you that we have no shortage of teachers, of PTAs, of school councils, of national STEM and STEAM outreach programs coming to us to help. So much so that recently we had to launch a nonprofit, a foundation that specifically handles kind of the interface of curriculum development, deployment of schools and so on. So you really, you know, it's a lofty goal, but part of the idea here is to kind of create a new generation, a new kind of farming generation. Is that practical and realistic and do we need that? Yeah, I think, you know, the average age of the farmer in the US is 65. 2% of the population is involved in farming. As Miles alluded to before, the beautiful minds that are in this room, that are in the generations from this room, that have been cultivated in the field, it's very hard to translate that knowledge if that person isn't there. And so thinking about, you know, it's not just the US, in Africa, half of the population is under 18 and 80% don't wanna be farmers. And we spend a lot of time thinking about small shareholder farmers in Africa when in reality, a lot of those kids are moving into the mega cities. And so across the globe, you keep hearing this constant refrain. Where's the next generation? What are they gonna be? You know, they wanna be involved in food, but they don't know how. So to offer a 21st century interface, to say you can be a data scientist farmer or a roboticist farmer or a nutritionist farmer or an educator farmer, all these other things, you know, that go along with that effort is what I think is part of building that next generation. So I'm about to go into the audience and get some questions and we'll take your questions the old-fashioned way through the microphone, but also Caleb would love to engage with you on Twitter. Caleb grows food is his handle on Twitter and we can do that now or later or anytime. Should be Caleb sometimes kills food. That's the alternative site, yeah. So while I walk into the audience, I want you to ask me one important question. Is the idea of controlled environments really something that is scalable in the sense that we're talking about feeding billions of people? I mean, we're not gonna grow certain crops in boxes, right? Right, so I think this is, you know, the rise of the kind of vertical farm culture and we're specifically talking here about one use case, climate manufacturing. And of course, for the last 10 years, we've seen a bankruptcy every other year scaling up from 10 million to 30 million to 50 million and now we're looking at hundreds of millions in the VC space that I think is finding failure. And do I believe that that's, you know, and I'll just stop with this slide because it's kind of awesome, you know, this was a factory that produces a million heads of lettuce a week in Japan, sells the product labeled Toshiba. So you're buying Toshiba lettuce. So as we sit in this room and we think about how we think that this might be crazy and when I go to Europe, trust me, there's some pitchforks in the audience about this. Like you need to realize that in different places, the fear of your food and the provenance that you are able to attach to that food, this lettuce sells for three times premium because they say where it was, where it came from and kind of give it that background history. But even that premium, as we're finding in the space of bankruptcies, is, you know, the aim at these kind of low margin, high throughput crops, I think it's gonna take a long time to get there because putting the world in a box, first of all it's difficult and all the thought leaders in the world at Purdue that have been working on this know that, and then attaching a business model while not sharing anything. You know, all the venture capitalists wanna see a proprietary patent portfolio, a proprietary data set, a proprietary process, something that can try to find scale. Well, since all these people, I imagine it in my mind and I go visit them all, usually after they go bankrupt so they'll talk to me. And they're all, I imagine them all as different circus animals. Like there's a tiger, there's a lion, there's a bear. They all have four legs, they all have heads and eyes, but a lot of other things are quite different and nearly incomparable. So if you were to work on solving one of their problems, it would almost be a useless effort because it wouldn't port to any of the other ones that are going on. So I think this kind of points towards that need to collaborate, that need to break down these silos of information. But where we are finding tremendous value, I think if we go to the next slide, is in somewhat, maybe it's not the first thing you think of. This is a video, but I don't know if it'll play. But when you look at it, it looks like what I've shown you before. It's some racks, it's some crazy lights that are pink. It grows plants. This is actually growing a tobacco plant, infecting the tobacco plant with mosaic virus, using the mosaic virus to create a protein that's resistant to Ebola, to create a vaccine. And so I think as we look for near term in this space, especially on climate manufacturing, we're looking more at pharmaceutical natural derivatives, nutraceutical natural derivatives, biochemicals that have an active ingredient component, even cosmetics, I've had a lot of interest in the cosmetics world, for something that has that margin, that can justify this infrastructure. And I think it'll start to trickle down, but to get there we need standards, we need information and knowledge sharing. I'll just, I think there's one more thing after this and we'll open it up. Slide? Well, I guess one of the questions I have in my mind is, we're not gonna be growing wheat in a box, right? I mean, so let me answer this one and then I'll go to wheat in a box. No, go back, cause it's awesome. So as most of my stories start, I didn't know what I was doing when this person called me and they said, this is the world food program. We'd like to see what you can do in a Syrian refugee camp. And I said, absolutely nothing, because I'm not gonna send unproven technology that's too expensive to grow, too little food, for people that really need it. And they said, okay, but we have camps that are on 30 years of 100% maritime food relief. We can't get fresh produce into these camps and they are not allowed to grow, because growing starts the process of colonization and land ownership. So what are you gonna do? I said, well, okay, like, as long as we all are under the notion, we all know that this is not gonna feed anybody, but we can start to see what would happen if we grew things. And so the world food program came online, the State Department, previous, called the Office of Global Food Security, Secretary Kerry announced it, it was super cool. We all thought that they'd grow fresh vegetables and things to eat. That's because we often discount human ingenuity. And so when we brought these food computers into that camp, what happened is what I hope happens much more often, which is we enable someone with a technology platform to do what they do best. There happened to be a person in the camp that was a PhD on St. John's wort. Studied this plant's whole life, knew that St. John's wort had an active ingredient that once made into a tea was a depression, a levion. The camp was rife with depression, as you can imagine. And of course it's not culturally appropriate to take Zoloft and Paxil, and it's probably not being dropped in. So this person, with their knowledge, on this platform was able to create somewhat of an operating business that provided a net benefit to this community that needed it and provided an economic benefit for the person that came up with the idea. So I think we see more potential along that space. And then to answer your question about, would you ever think that I would grow cotton in a box? No, but there's a certain reason that someone wants it. Corn and soy, that's the, I think the other part of the equation, which is what do we learn in the box that affects the field? And so as we start to build new models, and I'll tell you, I'm the first one to admit it's an incredibly difficult task. There's people in this room and at this university working on it, much more in depth than we are, but to connect what we can learn in a control environment to what it teaches us in a greenhouse, to what it could potentially link to in the field, that's an amazing thing to be working on. Because if we can do it in a control environment and we can run 10,000 experiments using machine learning and that that can point us towards what to do with some of the commodity crops in the face of climate change, in the face of increasing pest resistance, whatever it is that we're looking for, I think that's how this work makes its way into the field. All right, let's get a questioning. My name is Juliana. I've been following your work since 2016. And... T-shirt for her? Can you get her a T-shirt? All right. You mentioned that a lot of people don't want to become farmers nowadays and that's actually true, but one of the projects like yours made me actually want to go into this field and know that, okay, I can be a farmer, but not like my grandpa was for 65 years. And that leads me to my question that I'm currently an agricultural and biological engineering student and I would like to know if there's any academic background that would be helpful for someone to have to work in this area. Yeah, great question. First of all, I did not pay you, right? Okay, we've got that out of the way. You know, thank you. You know, having one of my team members with me here today, it's those kinds of things that make us think we're not crazy, which we get called a lot. So, in terms of skill sets, you know, if you look at my team and who I've put together, there's definitely a biocide, right? Plant biochemistry, plant physiology is really important because that tells us what we're trying to optimize for. The rest of the team is more like embedded systems and hardware engineering. Which is a lot of what we would design the system to do what we would do it for. And then the kind of software infrastructure of the team. So I think you can use the training that you've had. You're just going to have to collaborate outside of your discipline. And I think, you know, finding ways to take your knowledge, make it digital and make it available. You know, when we look for data sets to do training on, there's not a lot of public domain data sets in our space. Especially well curated, well tagged with metadata that makes it useful. So I think thinking about whatever it is that you've learned which I'm sure is awesome and how it might work in a pipeline towards things like machine learning would be really valuable. I don't know if that answers your question. But I mean, I have a flavor chemist on staff. You know, like it really isn't necessarily about like which discipline it's, how collaborative can you be across different disciplines? And what's your tolerance for pain? Because it's taken quite a bit to stitch these pieces together which traditionally don't talk much. Okay, another question here. My name is Victoria and I'm in the food science program here. And I was wondering what effect would this information have on like field farming as opposed to urban farming as opposed to like the Toshiba lettuce that you showed earlier. So I think the kind of two clear things that I can see, let's say reduce the signal, reduce the noise to some signal is climate manufacturing. And that I would say is a higher category than urban farming or vertical farming or plant factories or so on and so forth. It is this idea that we engineer a climate to produce a biological product to bring it to urban farming, like let's say eat fresh, pick fresh things. I think that is possible and not necessarily near term in a profitable sense. I do think from a community development perspective and a community farming perspective and an education perspective, it has amazing potential right now. But yeah, I think eventually that climate manufacturing could account for 10 to 20% of what a city eats, especially in the things that carry enough margin to justify it. On the field side, that's where I would call it climate prospecting. And that's really this link into phenotyping. Rapid phenotyping to produce models that we can run simulations on with real climate. To say, if we know X is coming, how could we prepare for that either genetically or from a farming practice perspective? And it really is a somewhat of a, could be a de-risking tool for our infield farming practice, if that makes sense to you. All right, another question back here. Hi, thanks for such an awesome talk. My name is Shomen Dukumar Ghosh and I do research in deep learning. So I wanted to ask you a question about what type of challenges are you facing because you are dealing with such so much tons of data. As you also mentioned in our answer for that lady there, that you are having shortage of data in this area as of now. So are you trying to collect, because you have a cloud, so you are anywhere in the world, if someone uses data, you are able to get that data, right? So are you trying to use that data for training of your models? And in case you don't have enough data, are you also going into reinforcement learning for applying where you don't have curated data or using those techniques? We have this project and that's actually, I'll start there. A couple of weeks ago, we convened a meeting with the Rockefeller Foundation, with the Schmitt Family Foundation, with the Walton Family Foundation, with the USDA all about this idea of building a research network of people sharing this kind of data. For the specific thing that you just described, we know we can't produce enough on our own. So we have to use tricks for small data, which you started to mention. The biggest hole that we had in our proposal is we didn't have an ag partner. And they were like, what do you know? You're gonna repeat things that have already been learned in control environments for the last 30 years, and they're probably right. So I'm actually here, meeting with folks and anyone that's interested, to think about how we could at least start with two nodes on a research network that developed standards together. I think I've visited a lot of different phenotyping operations in the US and abroad, and they're all pretty awesome. I mean, the one here is fantastic, and I'm jealous of the conveyor belts and the cameras. But to this point, they all exist as data islands, and they're all operating off their own data standards. And it's really not been much of a priority to make that data that's being collected publicly available as a commons, which is actually a very difficult thing to negotiate with a university. I was lucky enough that the director of the media lab was one of the founders on Creative Commons, and so I have a big support underneath me to make things open source. So I was able to convince, actually it took the presidential level of the university approval that I could give away the intellectual property for free. And so being able to break into that, where we can at least start by having networked research experiments on the university level, we know that's gonna take a lot of time and a lot of hard work, so we're prototyping that on the education network. Up until, let's see, you last, I don't know, six months ago? Six months ago? Up until six months ago, we didn't have a cloud that was operating. I lost you, there you are. And so we partnered with Google to develop that cloud. Everybody was hosting data locally to their machine. And then we had to get them to upload them, and then it was a very crazy process. And so I would say we're six months into the ability to have a network that is running experiments, and again, the personal food computer is not very controlled. It works off of the ambient. We can affect the ambient a little bit, but we do have a number of different sensors and a number of two cameras in a lot of the units, one at least, to start doing aggregate learning from that community. And it's really the idea that this project could become like a SETI-like project, that it could become a citizen science project. And that as we were recently lucky enough to raise some funding for STEM education, that each one of those bots that goes out, independent of what they do, could be useful towards illustrating this idea of the network advantage, which I want to then extend into the research community. First, the control environment ag research community develops some standards together, shares some information that's useful, and then take it farther from there. And I think we're seeing that happen on the field side, not so much in the public domain, but definitely from infield data acquisition efforts. So I'm just kind of counting on that the field becomes more described, and I'm trying to figure out the models that can interact kind of with that data. Arshad, back here. I'm Pradham, and I'm a current computer science major here. And I'm wondering like in the personal computer, like I meant the food computer, like what programming languages do you, does it use, does it use like a specific one, or is it like multiple different languages together for like the sensors and stuff? Yeah, so great question. I have been through this one. Picking a language picks a community. So if you expect to have community development of software, you have to have a language that the community that you're targeting can speak. I would say we've been through ROS in the early stages, which is robot operating systems, and we hit some roboticists that weren't really super interested in food. Then we decided to go to C, and we were gonna like bare metal all this stuff so that it would be really efficient. And we actually got somebody from the NSA that came over to our group to start helping us build that out, and then we figured out there's not enough people that could help us. So actually right now the current kernel is all Python. This is a common language. It's being taught starting in kind of the middle school, going into high school. It's a common language being taught at the university level. So that's what most everything is written on. You, in particular, should go to our GitHub, where you can find documentation about this, or our forum, or our Wikipedia, so that you'll see the diagrams of our whole infrastructure, where it stands today, and hopefully where I can recruit you to do work for free that would be interesting for you and beneficial for the community. All right, and a reminder, you can always continue the follow-up questions with Caleb at Caleb Grows Food on Twitter. We have a few more questions. We'll try to squeeze in here. Here's one more. Thank you so much for coming to speak with us today. My name is Maya Rodamaker. I'm in the Agriculture Communication Program, and I think it's incredible what you were able to do with taking that technology to the Syrian refugee camps. I'm very interested. Is there any initiative to implement that same sort of technology in American food deserts, or like very urban, very poverty stricken areas in the US? Yeah, great question. So this kind of gets into what I would call like the application layer, right? Like you have this interest because you have this background. You have a knowledge of a need that exists, and then you wonder, can this technology stack be used for that? The answer right now is from a community benefit standpoint, from a nonprofit standpoint, I think it's interesting to start experimenting. I think the claims that you could grow enough food financially, viably, to affect a food desert from a production and consumption point of view are very, very hard to make, and probably not anywhere close to true. But using this to work on issues of kind of education and food literacy to gather a community, it's really fun. Like we do with one of our farms, we grow certain leafy greens and herbs, and we have had a history of donating them to low-income grocery stores. So there's a grocery store started by the founder of Trader Joe's in Boston that specifically offers for a low-income demographic. And so we do donations from our shipping containers to that store because most of the time they don't get very much fresh produce. It's all shelf-stable stuff that they've been able to get from other grocery stores that was close to code date and move it into this format where it sells out quickly. So I think some of those solutions are probably more near-term. I forget what it's called, but it's awesome. You should look it up. Trader Joe's low-income grocery store. Doug Roush is the guy that runs it. Anyway, so just by bringing this food into that community, telling the story of where it came from and getting people involved in it, we've done donations to food pantries. I think that that's a really interesting place to explore, but you're gonna have to start from a non-profit perspective and from a perspective of experimentation and probably best to have it around an educational goal right now. Doesn't mean that we can't get there. I'm just being honest with you about how I see the current state of the technology and the cost of implementation and all of that. All right, one more question here. Not a question of food safety. Hey, Caleb, thank you so much for coming to Purdue and for this talk. I really learned a lot. My name is Naman and I'm a graduate student at the Purdue Computer Science. I was really interested in the deployment phase of what you spoke about, especially the example you have of India of the cotton farmers. And if I understand correctly, like you were able to take it up to scale and produce enough cotton to fund 70% of the linens made in America. Big misunderstanding there. So I was just wondering like how do you, what is the role of automation in this scale of deployment and what challenges did you learn from that specific product and how we can have cities that are kind of dependent on a big vertical farm and they're being able to sustainably feed their constituents? So we do not produce the cotton for 70% of the sheets and tiles in America. I wish we did. That would be sweet. They produce the sheets and 70% of the sheets and tiles from their factory that's golf cart 45 minutes along that thing. What we've been able to do with them is deploy this technology in a prototype format where they're now starting to do kind of advanced textile production, small quantities until we prove out that it makes sense for them. And the next step for them is to do 10,000 square meters of cotton. But the real question that you had which is a super good question, and let's suggest it from the point of view of why are vertical farms failing? Of course all the kind of big reasons that I talked about, but if you want to get more tactical, labor, cost of labor, that's the number one operating cost for a vertical farm, especially in the United States. So automation obviously replaces cost of labor. So a ton of effort going into right now full automation of the system. These systems are relatively new. They're all from venture-backed startups. So their ability to integrate highly costly automation is pretty low, but it's starting to happen, and I would say the Netherlands leads the way on plant production automation, not necessarily in the space, but that does provide something that is absolutely necessary, especially in our context. The next biggest cost is actually operational energy, but it's not maybe what you would think, which is like lighting and all the other stuff. It's heating and cooling. These things tend to be built in old warehouses because they're like, it's an urban renewal project, we're gonna put it in this old warehouse that's been reclaimed, and that warehouse is leaky and horrible and has no distribution of HVAC. The primary objective, what I used to do designing data centers was heating and cooling, was basically taking the heat off of the computers, getting it out and keeping them cool enough to operate. This is the mindset that we have to take towards vertical farming, which is every kilowatt to transition to every BTU has to be as efficient as humanly possible. You need to treat heating and cooling as incredibly precious because that's gonna drive the majority of the cost in this space. So I think innovations around distribution of effective heating and cooling, innovations around automation, and of course, you go into these environments and from the biologic perspective, they're pretty rudimentary. These are mostly off-the-shelf technologies. I go around to all of them and they all say they're super special and then you get in and they're like, where'd you get this rack? Home Depot. I think we're pre the time where there's a skilled workforce of crafts people across plumbing and HVAC and electrical and all these things that actually have an idea of what to do with vertical farming. So I think those are two of the biggest pain points that I can see and I hope that answers your question. Well, I think there are a few more questions out there. I think Caleb will hang around a little bit if you wanna come on up and ask him a few more questions and of course at Caleb grows food if you want to continue the conversation on Twitter. But Caleb Harbour, my friend, colleague and always a lot of good food for thought for sure. Always a pleasure to be with you. The puns are never ending. And there's no end to it, yeah. But thank you all for being with us tonight and give him a round of applause and thank you for being part of the idea stuff for us. Good job.