What the global health community can learn from ebola, zika, and other past outbreaks. 

Insights: The Power of Big Data

[MUSIC PLAYING] SPEAKER 1: We started with a collection of collection of energy related data. So our users are scattered in over 80 countries in the world. Even the old surfaces of the buildings look similar. The energy consumption level varied significantly. And we did a survey about our users in this over 80 countries. We found that if you do not have the insulator and are of general level insulator or of very good insulator, the data really varied significantly. I was shocked. And so starting from then, we started to have a 20 centimeter insulator to keep the walls inside that building. The energy data is mostly invisible but is most important, especially the building energy consumption. So 15 years ago, we found it was extremely important. And besides, the building related to energy consumption takes a very big proportion among all the consumption of energy. So if we increase the efficiency of the building consumption, and that efficiency increase can be very high. We now have a high requirement for the comfortness of the temperature in the building. So to reduce the energy consumption of the buildings can become a very or even the most important measure for protecting the environment and the climate. BILL MCDONOUGH: I think what we're seeing here is some of the most advanced approach. one of the most advanced approaches to this. And it definitely will become part of the new normal. SPEAKER 2: Have you used any of this in your designs? BILL MCDONOUGH: Yes. SPEAKER 2: Broad stuff? BILL MCDONOUGH: Not yet, not broad, but similar things. We're looking at all kinds of detection, for example, for air quality. SPEAKER 2: And one last quick question. That one point that you made when you gave your presentation the other day was that this can be done with no added cost. Can this be done with no added cost in the sorts of designs that you're doing? BILL MCDONOUGH: Yes, it gets built into the qualification of the building. It's not a crazy thing at all. 

Insights: Using Data to Slash Carbon Emissions

Uncategorized

Build structures that are more energy efficient.

Using Data to Slash Carbon Emissions

One use is building structures that are more energy efficient. (September 2019)

waste collection off energy related data. So our users are scattered in over 80 countries in the world. Even though surface itself, the buildings look similar. The energy consumption level varied a significant today and the way he does survey about our users in the over 80 countries. We found that if you do not have so into later and all general level insulator all, they're a good insulator that they have really very significantly. I was shocked and so talking from then waste Arteta too, have ah, 20 centimeters insulator to keep the warm inside inside a beauty, the energy data is most invisible, but it's most important, especially the beauty energy consumption. So 15 years ago, we found it was extremely important and besides, the shooting related energy consumption takes a very big proportion among all the consumption of energy. So if we increase the efficiency of the beauty and consumption and that efficiency increase can be very hi, we know have ah ah high requirement for the comfort of the temperature in the beauty. So to reduce the energy consumption of the beauties can become a very or even the most important measure for protecting the environment. and the climb mater. I think what we're seeing here is some of the most advanced, um, approach when the most advanced approaches to this and it definitely will become part of the new normal. Have you used any of this in your designs? Broad stuff? Not yet not broad but similar things. We're looking at all kinds of detection, for example, for air quality. One last question that one point that you made when you give your presentation the other day was that this could be done with no added cost. Can this be done with no added cost in the sorts of designs that you're doing? Yes, it gets built into the qualification of the building. It's not a crazy thing at all.

[MUSIC PLAYING] BETH KOWITT: What kind of data are you collecting from those farmers? And how are they using it? Maybe you could give an example of that. BETH FORD: Yeah, we have applied research plots, for instance, where we're working with farmers to help them understand how they should plant, what kinds of things they should plant, different soil types. We also use predictive models. And we focus specifically on in-season technology or in-season models. And I think that that's really important this year. This year, many of you have seen we've had dramatic issues with weather across the country. Many farmers couldn't get planted. And that means that they have a huge issue with a gap in production. And more importantly, these farmers have been in a situation for multiple years where they have had low commodity prices. They have had struggles in the rural communities. They have had struggles on the acre. So these predictive models, this data that we're capturing-- and we're capturing farmer data, what they planted, what soil types, what their processes are on the acre. And we use that in conjunction with an alignment with our applied research platform to provide a model for them for what they may want to do, what they want to do in season. Do you want to apply some kind of a spray? Do you have an issue? Do we need to get a soil sample? Do we see a disease state? And so this model, we're using all sorts of imagery, analytics. But their data is their data. We capture their data in a data silo. We pair it with our applied research plots. We use predictive models. We use satellite technology. We use drone information. Many of these things that allow us to advantage a farmer. And never has that been more important in the United States, especially in this year, where we have had dramatic issues-- drastic issues with weather. BETH KOWITT: OK. Are farmers wary to give you their data? BETH FORD: You know, it's a great question. You know, nobody-- thankfully, we are a co-op. We are farmer owned. And the data is their data. And we make it very clear to them, while we will capture their data, we will not use their data unless we have permission to use their data. And we may use it from a modeling perspective. But they own their data. Of course they're reluctant, right? It's like all of us. Well, what are you going to do with the data? What kind of data? We've been using it and leveraging it by the way as well on sustainable production. We have a Truterra insight model that allows us to work with a farmer to improve their sustainable production. So are they wary? Yes. Do they understand the value of it? Yes. And we've been working with them for many, many years. And so we have a significant amount of data. And we make it clear to them that they own their data. BETH KOWITT: OK. So Sara, you have an interesting model. You work with everybody but the farmer. So explain why that is. That would be really helpful, I think. SARA MENKER: Yeah, well, we work with everyone but the farmer. But we definitely work for the farmer in the sense that the reason we exist as a company as grow intelligence is to make the system around the farm a lot more efficient. So what we do is we take large, disparate data sets that have anything to do with agriculture, that have been captured anywhere in the world. We transform and translate all of that data, normalize it, and essentially build a series of predictive models that let us you know forecast anything from supply to demand to price for any agricultural commodity anywhere in the world. One of the things that we have seen in the past couple of years is just the sheer volatility of climate events that occur everywhere in the world. So we're talking about floods in the US. But southern Africa, this year, had one of the worst droughts on record, right? And so you have these extreme shifts. So what it enables us to do is that we actually can look at a combination of satellite imagery with data that comes out of national statistical agencies to model real time-- model and build out real time indices on whether it's drought or flood. An index that, globally-- that lets people to start manage that risk. The thing is, we're not going to necessarily get away from the fact that it's happening right now, because we're just being confronted with it. But we can manage it better. That means if you are a food and beverage company or you're a retailer, you have massive supply chain risk that's tied to the way that you procure. So California drought, you know, 70% leafy greens come from one tiny portion of California. The next time there's a drought, where do you diversify that risk from? You can actually structure of financial instruments that let you do that and manage that risk, so you as a company can get paid out on a policy that then lets you use that money to go and essentially purchase product elsewhere in the world for potentially much higher prices. Because that's what happens, right? Markets react to a decline in supply by increasing price. And organizations need to be positioned to be able to react to that and manage that, so the consumer doesn't also suffer as a result of it. And so this is what I mean by that there needs to be a real re-architecting of the financial system that we have. And I think data can play a massive role in enabling that communication across organizations and the transfer of risk across organizations. BETH FORD: You know, to add on that, one of the things we're trying to do-- because change will happen in terms of building a robust and resilient food chain at the farm level. And so we've invested in something called the our Truterra insights engine, where we're working on the acre with the farmer. We're using their data. And we're using-- it's a over a trillion data points. And what we're trying to do is help a farmer understand what investment they can make to improve sustainable production on their farm while improving their profitability, so variable rate application of fertilizers or stabilizers or no-till or things like that that will improve the resilience of the acre on the farm on which they're working on, but at the same time, improve profitability. And I agree. What we need to understand is what are the incentive structures that we want to put in place to incentivize a farmer who wants to continue to improve production on their farm but at the same time is trying to make sure that they can be profitable in order to do that? So insights and data and analytics will help us. And at the same time, we're working with our master agronomists very directly with the farmer, with their data, to help them understand how they can improve work and production on their farm. [SOUND EFFECTS]

Insights: How Big Data Can Help Agriculture Respond To Climate Change

Farmers work more efficiently and battle climate change.

How Big Data Can Help Agriculture Respond To Climate Change

Land O’Lakes CEO Beth Ford is joined by Sara Menker, Gro Intelligence CEO, to discuss how tech is helping farmers work more efficiently and battle climate change. (July 2019)

What kind of data are you collecting from those farmers and how are they using it? Maybe could give an example of that. Yeah, we have applied research plots, for instance, where we're working with farmers to help them understand how they should plant, what kinds of things planned different soil types. We also use predictive models, and we focus specifically on in season technology in season models. And I think that that's really important this year. This year, many of you have seen we've had dramatic issues with weather across the country. Many farmers couldn't get planted, and that means that they have a huge issue. We have a gap in production, and more importantly, these farmers have been in a situation for multiple years where they have had low commodity prices. They have had struggles in the rural community. They have had struggles on the acre. So these predictive models these the's, this data that we're capturing and we're capturing farmer data, what they planted, what soil types, what their processes are on the acre and we use that in conjunction with in alignment with our research, provide a model for them for what they may want to do what they want to do in season. Do you want to apply some kind of a spray? Do you have on issue? Do we need to get a soil sample? We see a disease state on dhe. So this model, we're using all sorts of imagery analytics, but their data Is there data? We captured their data in a data silo. We pair it with our applied research predictive models. We use satellite technology. We have drone information, many of these things that allow us to advantage. Ah, farmer. And never has that been more important in the United States of specialists in this year where we have had dramatic issues, drastic issues with with weather, our farmers wary to give you their data. You know, it's a great question. Nobody. Thankfully, we are co op. We are farmer owned and the data is there data and we make it very clear to them. While we will capture their day that we will not use their data unless we have permission to use their data. We need from a modeling perspective, but they own their data. Of course they're reluctant, right? It's like all of us. So what? You gonna do with the data? What kind of data? We've been using it in leveraging and by the way as well. On a sustainable production, we have a true terra Insight model that allows us to work with a farmer to improve their sustainable production. So are they wearing? Yes. Do they understand the value of it? Yes. And we've been working with them for many, many years. So we have a significant amount of data, and we make it clear to them that they own their data. Okay, Sarah, you have an interesting model. You work with everybody but the farmer, so explain why that is. That would be really helpful, I think. Yeah, well, we worked with everyone but the farmer, but we definitely work for the farmer in the sense that the reason we exist is a company, as grow intelligence is to, you know, make the system around the farm a lot more efficient. So what we do is we take large, disparate data sets that have anything to do with agriculture that have been captured anywhere in the world. We transform and translate all of that data, normalize it, and essentially build a series of predictive models that let us, you know, forecast anything from supply to demand to price for any agriculture commodity anywhere in the world. One of the things that we have seen, you know, the past couple of years is just the sheer volatility of climate events that occur everywhere in the world. So we're talking about floods in the US, but Southern Africa this year had one of the worst droughts on record, right? And so you have these extreme shift. So what it enables us to do is that we actually can look at that. A combination of satellite imagery with data that comes out of national statistical agencies to model riel time in a model and build out real time indices on whether it's drought or flood, an index that globally that lets people to start manage that risk, that the thing is, we're not going to necessarily get away from the fact that it's happening right now because we're just being confronted with it. But we can manage it better. That means if you are a food and beverage company or you're a retailer, you have massive supply chain risk. That's tied to the way that you procured California drought. You know 70% of leafy greens come from one tiny portion of California. The next time there's a drought, where do you diversify that risk from? You can actually structure financial instruments that let you do that and manage that risk. You as a company can get paid out on a policy that then let's use that money to go and essentially purchase product elsewhere in the world for potentially much higher prices. Because that's what happens, right? Markets react to a decline in supply by increasing price, and organizations need to be positioned to be able to react to that. I managed that so the consumer doesn't also suffer as a result of it, right? And so this is what I mean by you know, that there needs to be really kind of a really architect ing of the financial system that we have. And I think data complete a massive role in enabling that communication across organizations and the test for risk across organization. You know, to add on that one of the things we're trying to do because change will happen in terms of building a robust and resilient food chain at the farm level. And so we've invested in something called a true terror inside mention where we're working on the acre with the farmer were using their data, and we're using it over a trillion data points. And what we're trying to do was help a farmer understand what investment they could make to improve sustainable production on their farm while improving their profitability. So variable rate application of fertilizers or, um, stabilizers or no till, or things like that that will improve the resilience off of the acre on the farm in which they're working on but at the same time improve profitability. And I agree. What we need to understand is what is the incentive structures that we want to put in place to incentivize a farmer who wants to continue to improve production on their farm but at the same time is trying to make sure that they could be profitable in order to do that, so insights and data and analytics will help us. At the same time, we're working with our master of Rana's very directly with the farmer with their data to help them understand how they can improve work on and production on their farm

 [MUSIC PLAYING] MICHAL LEV-RAM: In an ideal world, how do companies go about making sure that privacy and security are, first and foremost, top of mind when they're developing these technologies? And I know you've talked and written about how patients, consumers, should own their data. What does that world look like? And is that the only solution? ERIC TOPOL: I think, ultimately, it's the only solution, that people have to own your data. Because what we're seeing now, of course, is the worst case scenario where data is sold, it's stolen, hacked, cyber thievery is rampant. And our medical data is at least 5x the worth on the dark web as our personal financial data. So we have a situation now which is not acceptable. And if you talk to cyber gurus, they will say, you have got to get it down to a unit of one or the smallest possible units. Whereas, of course, it sits on massive servers that have indeed been hacked. And so many health systems have been held hostage for their data. So that requires a fundamental change. We've seen it in other countries like in Estonia, and Switzerland, and other countries are adopting this ownership model, where people can share parts of their data with certain-- whether they be physicians or medical research projects or that sort of thing. Eventually we will go there because this will have to be considered a civil right. That is your data, you paid for it, and all the other things that it could change your life. But in the era of AI, it's the inputs. If you don't have all this data, which is, in the US, it's all over the place from various doctors and encounters throughout one's life, then you have a compromised input and you can only expect so much as an output. So that's why we have to move to a different method and we're a long ways from that today. JESSICA MEGA: If you think about a pacemaker, or an ICD, or a continuous glucose monitor that is giving medically great information, to a person who may change their dose of insulin, for example, we have to be 100% sure that this is actually their data. So again, I think understanding cybersecurity in this age is critical, and you're seeing groups across the board championing that. I think on the ownership, I would even say and take it one step further, how do you actually give value back to people? So not just surfacing a lot of medical data, but saying, this is what it means for you. Here is your continuous glucose data. You ate a certain food and here's how you can make healthy choices. I think the more we think about, not only the data, but value, that's an ambition that we've had from the beginning. MICHAL LEV-RAM: And would you say, I mean, are you taking steps in that direction early on to make sure that that data is owned and stored with people themselves? JESSICA MEGA: So if you look at our baseline initiative that we started several years ago in conjunction with Duke, Stanford, and the American Heart Association, as well as volunteers who sit on the committees, the idea was, how can you ethically return information to people? And so that has been a philosophy that is something that underscores the work that we do. ERIC LEFKOFSKYl: My take on it in general-- MICHAL LEV-RAM: It's not that obvious, I mean, we don't own most of our data that's out there. ERIC LEFKOFSKYl: I think the Supreme Court has weighed in. But the bigger issue, I think, is coming from kind of the tech world, actually, this is one of the viewers I think the federal governments has done a decent job in terms of defining PHI, and there are laws like HIPAA, there are a litany of requirements in terms of what you can do with identified data as opposed to de-identified data, none of which, by the way, exists in the non-health world. We have breaches all the time in e-commerce of millions of patients where the credit card data-- and if you had that same breaches as a health care a company, it would be radically different. So I agree, we have a long way to go. But in general, I think this is an area where we've done better than we have in some other areas. And I think the real challenge is, one of the biggest problems you have as a patient in terms of ownership of your data is what do you do with the data? First of all, it's a huge amount of data. And second of all, it's largely unstructured. And what ends up happening is you, as a patient, when you get your hands on your data, you're often e-mailing it to a physician. And they have no ability to import it back into their EHR, EMR system. So there is no interoperability or transferability of this data. And so I think even though a patient should own their data, and they certainly have a right to it. And even the provider should be able to house that data seamlessly. Probably the bigger challenge we all have to tackle is the underlying infrastructure that allows us to move around this kind of rich, clinical data, seamlessly across providers, across physicians, all of which is entirely today broken and siloed, but again, is likely to be fixed by these really massive technology currents that are just there. So with NLP and OCR and all these great technologies, we will likely be in a place in a very short number of years where this data can be structured on the back end and moved between an Epic, or a Cerner, or an Allscripts, or an [INAUDIBLE],, or pick your provider. And that I think is going to be one of the biggest unlocking of patients being able to port their data from one doctor to another. 

Insights: Who Owns And Protects Our Medical Data?

Who keeps and safeguards this information?

Who Owns And Protects Our Medical Data?

New technologies are generating enormous amounts of data about our health, but who keeps and safeguards this information? (April 2019)

in an ideal world. How do companies go about making sure that privacy and security are first and foremost, top of mind when they're developing these technologies? And I know you talked and written about how patients consumers should own their data. What is that world look like? And is that the only solution, I think, ultimately, is the only solution that people have to own your data? Because what we're seeing now, of course, is the worst case scenario. Where date is is sold, it's stolen packed, cyber thievery is rampant and our medical data is at least five X. The worst on the dark Web is our personal financial data. So we have a situation now, which is not acceptable. And if you talk to cyber gurus, they will say, You got to get it down two units of one or the smallest possible units. Whereas, of course, it sits on massive servers that have indeed been hacked in. So many health systems have been held hostage for their data, so that requires a fundamental change. We've seen it in other countries, like in Estonia and Switzerland, and other countries are adopting this ownership model where people can share parts of their data with certain, but they be physicians or medical research projects, that sort of thing. Eventually we will go there because this will have to be considered a civil right. That is your data. You paid for it and all the other things that it could change your life. But in the era of a I, it's the input. If you don't have all the data which is in the US, it's all over the place from various doctors and encounters throughout one's life. Then you have a compromise input and you can only expect so much as an output. So that's why we have to move to a different method. And we're a long ways from that. Today. If you think about a pacemaker or an I P D or continues glucose monitor that is giving medically great information to a person may change their dose of insulin, for example, we have to be, ah, 100% sure that this is actually their data. So again, I think understanding cybersecurity in this age is critical and you're seeing groups across the board chancing that I think on the ownership I would even say and take it one step further. How do you actually give value that two people not just surfacing. Ah, lot of medical data, but saying this is what it means for you. Here is your continuous glucose data. You, uh you ate a certain food and here's how you can make healthy choices. I think the more we think about not only the data but value, that's that's a mission that we've had from the beginning. And would you say I mean, are you taking in that direction early on to make sure that that data is owned and stored with people themselves? So if you look at our baseline initiative that we started several years ago in conjunction with Duke Stanford in the American Heart Association, a swell volunteers who sit on the committee, the idea was, How can you ethically return information to people? And so that has been a philosophy that is something that underscores the work that we do my take on it In general, it's not that obvious. Well, don't own most of our data that's out there in the Supreme Court's waiting. But the bigger issue, I think, is coming from kind of the tech world we? Actually, this is one of few hours. I think the federal government's been a decent job in terms of defining pH I, and they're lost like hip. There are litany of requirements in terms of what you could do with identifying data supports the identified, none of which, by the way, exists in a non health world. We have breaches all the time. Any commerce of millions of patients with a credit card dated and if you had that same breaches of health care company radically different. So I agree we have a long way to go. But in general I think this is an area where we've done better than we have in some other areas. And I think the real challenge is, um, one of biggest problems you have is a patient in terms of ownership of your data is what do you do with the data? What a huge amount of data. And second of all, it's largely on structured, and what you end up when it's happening is you as a patient, when you get your hands on your data, you're often e mailing it to a physician, and they have no ability to import it back into their e h r E M R system. So there's no interoperability or transferability of this data. And so I think, even though patient should own their data and they certainly have a right to it and even the provider should able, How's that data? Seamlessly? Probably. The bigger challenge we all have to tackle is the underlying infrastructure that allows us to move around. This kind of rich clinical data seamlessly across providers across positions, all of which is entirely today broken and siloed. But again, it's likely to be fixed by these really massive technology currents that are just They're so with NLP and OCR and all these great technologies, we will likely be in a place in a very short number of years, where this data can be structured on the back end and moved between an epic or Cerner on all scripts on Uncle Liam are pick your provider and that I think it's gonna be one of the biggest, you know, unlocking of patients, being able to support their data from one doctor to another

[MUSIC PLAYING] JAMES PARK: So for us, we clearly don't want to monetize data directly, meaning sell it to third parties, et cetera. It's not in our interests. And there's already a lot of regulation, including GDPR in Europe, that's going to heavily regulate this for the protection of consumers. But we do see the opportunity in data. Managing data flows between different entities is one area, and then secondly, providing insights or diagnoses, some value-add on top of the data. So for instance, a couple of the algorithms that we're working on through the FDA's new Pre-Cert program is algorithms to diagnose sleep apnea, or AFib. So those are some ways that we could monetize the data, by providing algorithms on top of it. SPEAKER: That's great. And it's a subscription model, or is it a-- JAMES PARK: It could be subscription, or it could be some type of risk sharing model as well with payers. SPEAKER: And that is to the payers, right? JAMES PARK: Right. OK. 

Insights: How Fitbit is Considering Monetizing User Data

The company doesn't want to sell your personal health data.

How Fitbit is Considering Monetizing User Data

CEO James Park explains why the company doesn't want to sell your personal health tracking data to third parties, but he does want to make money off it. (March 2018)

way clearly don't want to monetize data directly, meaning selective third parties, et cetera. It's not in our interests and theirs early. A lot of regulation, including GDP, are in Europe. That's going to have We regulate this for production of consumers. But were we Did you see the opportunity in data? Managing data flows between different entities is one area and then secondly, providing insights or, you know, diagnoses some value out on top of the data, for instance. Ah, couple The algorithms that we're working on with the FDA is through. The FBI's new Priester program is algorithms to diagnose sleep apnea or a fib that those are some ways that we could monetize data by providing algorithms on top of that. Great. And it's a subscription model. Or is it a It could be subscription. Or it could be some type of rich risk sharing model. Well, and that is to the payers, right? Okay,

[MUSIC PLAYING] DANIEL GRUNEBERG: McKinsey put out this study recently. And they said that agriculture is the least digitized industry of all industries. But that's rapidly changing. Last year, there was about $3 billion invested in new technology. So now ag tech has been born and is top of mind, because it's enabling us to escape to scale farming more sustainably. So that said, what's Sensei up to? Well, we are building a global network of hydroponic farms. For those of you who aren't as familiar, a hydroponic farm as one where you grow fruits and vegetables directly in water beds with nutrients in a climate controlled environment. So this can be a greenhouse or a building. And you have sensors and software that climate control the environment to create the ideal conditions for plants to thrive. Now, our first farm is on the island of Lanai, which is in Hawaii. Now, this is a state that imports 85% of its food and where, on average, it takes about 10 days for food to get to the islands. So it's degrading that entire time. As part of our mission, we are looking to make our operation fully sustainable. So we use about 1/10 of the water as compared to conventional farming. We use no harmful chemicals. And I'm happy to say that we are a partnered with Tesla to provide solar energy at our facilities. But there's a bit more to our farm. So the conversation in hydroponics today has mostly been around nutrition per acre-- or yield per acre. So how much yield can we get per square foot? Sensei's focus is different. We focus on nutrition per acre. So to properly nourish the world, we're collecting and analyzing data on all the fruits and vegetables that go in our facilities, then creating a continuous feedback loop to then help us better understand how to make food closer to the way that nature intended it. And so from black trefelli tomato panette we'll grow all the way to komatsuna mustard greens. We'll reproduce the best of the best. And then we'll tell you the exact nutritional breakdown at the time of harvest. Because we believe that you deserve to know what you're eating. With that said, our door is wide open to collaborate. We're just getting started. And thank you so much. [SOUND EFFECTS]

Insights: Using Data to Operate Hydroponic Farms

SENSEI suggests the future of agriculture.

Using Data to Operate Hydroponic Farms

SENSEI’s Daniel Gruneberg explains how his company sees the future of agriculture. (March 2018)

Mackenzie put up to study recently, and they said that agriculture is the least digitized industry of all industries, but that's rapidly changing. Last year, there was about $3 billion invested in new technology. Now AG tech has been born in a top of mind because it's enabling it to scale farming more sustainably. So that said, What? Sensitive to well, we are building a global network of hydroponic farms. For those of you who aren't as familiar. Ah, hydroponic farm is one where you grow fruits and vegetables directly in water beds with nutrients in a climate controlled environment that could be a greenhouse or a building. And you have sensors and software that climate control the environment to create the ideal conditions for plants. Now our first farm is on the island of Lanai, which is in Hawaii. Now. This is a state that imports 85% of its food and where on average it takes about 10 days for food to get to the island. So it's degrading that entire time assed part of our mission. We are looking to make our operations fully sustainable, so we use about 1/10 of the water is compared to conventional farming. We use no harmful chemicals, and I'm happy to say that we are partners. Tesla's to provide solar energy at our facility but there's a bit more to our farm. So the conversation in hydroponic today has mostly been around nutrition for Acre, our yield breaker. So how much yield can we get? Per square foot senses focuses different. We focus on nutrition per acre. So to properly nourish the world, we're collecting and analyzing data on all the fruits and vegetables that grow in our facilities, then critic, creating a continuous feedback loop to then help us better understand how to make food closer to the way that nature intended it. And so from black, definitely tomatoes submit will grow all the way to come up through. The mustard greens will reproduce the best, the best, and then we'll tell you the exact nutritional breakdown, the time of harvest, because we believe that you deserve to know what you're eating. But that said, our door is wide open to collaborate. We're just getting started and thank you so much

 CLIFTON LEAF: How does a company like Pfizer, or any drug maker, use artificial intelligence to help create better drugs, drugs that target the targets better? Just take it in the simplest terms if you could explain that. DR. MORTEN SOGAARD: Well, I could maybe give a couple of examples. One is kind of electronic biomarkers where particularly for neurological diseases, actually diagnosis one of the issues for us. Maybe you have 30% treatment response, 30% down response, and then you have 40% placebo response. So how do you design clinical trials? So having much more precise diagnosis, for example, having someone call in to a in a phone line, having the voice pitch recorded as a proxy of mood or depression. Yeah. So that's kind of one example. We have actually collaborated with IBM on Parkinson's electronic monitors where basically we characterize the phenotype much more precisely. I think another example I could give is deep learning. So that was I guess the technology pioneered by Google to identify cat faces as I'm sure you're all aware, right? So we use that for mining electronic health record and kind of associated lab data to understand disease trajectories for some of these chronic diseases like fatty liver disease or autoimmune disease to really understand what is likely to happen to two different patients. Of course, it's at the population level, but we also hope that it gives some productivity at that individual. 

Insights: How Pfizer Is Using Data To Improve Medical Diagnoses

Pfizer uses artificial intelligence to create better drugs.

How Pfizer Is Using Data To Improve Medical Diagnoses

Dr. Morten Sogaard explains how Pfizer uses artificial intelligence to create better drugs. (May 2017)

Ah, a company like flies or any drugmaker used artificial intelligence to help, you know, create better drugs, drugs that that actually target the target. The target's been, you know, So what? We'll just take us in the simplest terms. If you could explain that Well, I could maybe give you a couple of a couple of examples. No one is, You know, kind of the Elektronik by markers. You know where you know, particularly for neurological diseases. Actually, diagnosis one of the issues for us is you know, maybe you have 30% treatment response, you know, 30% don't response, and then you have 40% placebo response. So how do you design clinical trials? So? So, having much more precise diagnosis, For example, having someone call in to a, you know, in the full line, you know, having the voice pitch recorded, you know, as a proxy up. You know, mood, a depression, and yeah, so that's kind of one. One example we have excellent collaboration with with with IBM on Parkinson, you know, kind of electronic electronic monitors were basically characterized the peanut type much more precisely. I think another example I could give, you know deep learning, you know. So that was against the technology pioneered by Google to identify, you know, cat cat faces. And I'm sure you're aware, right? So we use that for mining, electronic health records and kind of associate. It left data to understand deceased trajectories for some of these chronic diseases, you know, fatty liver disease or or immunities really understand. You know what is likely to happen to two different patients? Of course, it's at the population level. But you also hope that it gives some productivity of the individual.

[MUSIC PLAYING] CATHY O'NEIL: The problem is that, actually, all algorithms, all big data algorithms and models, they're only as good as the data that goes into them. So if that data has bias, then the outputs have bias, too. Not to mention that it's the humans that decide what success looks like. So for any algorithm, to build an algorithm you need this data. Which again could be problematic. But you also need to define success and then optimize to that success. And depending on who you are, success might mean one thing or another. SPEAKER 1: Right. And so you worry about how not all math-based decisions are bad. Some algorithms are good. Some become weapons of math destruction. How do you tell when one is not doing what is supposed to be doing. CATHY O'NEIL: Well, sometimes it is doing what it supposed to be doing. And it's still bad. I define a weapon of math destruction to be one that has large-scale implications if it's important. It's also secret, and it destroys people's lives. So those three things put together make it a really destructive weapon of math destruction. The example I think about is the teacher value-added model. Where we get teachers fired by this stuff, they don't understand it. But at the same time, it's just a bad statistical model. Some people get a 6 out of 100 one year, 96 out of 100 the next year. And they didn't change the way they're teaching. So it's actually a meaningless thing, but it has strong consequences and it ruins people's lives. [MUSIC PLAYING] 

Insights: Identifying and Avoiding The Dangers of Big Data

Author Cathy O’Neil explains what could go wrong.

Identifying and Avoiding The Dangers of Big Data

Author Cathy O’Neil explains what could go wrong. (December 2016)

but with the problem is that actually all algorithms, all big date algorithms and models, they're only as good as the data that goes into them. So that data has biased than the outputs have bias to. Not to mention that it is the humans that decide what success looks like. So for any algorithm to build an algorithm, you need this data, which again could be problematic. But you also need to define success and then optimize to that success. In depending on who you are, success might mean one thing or another. So you worry about how not all math based decisions are bad. Some algorithms are good. Some become weapons of mass destruction. How do you tell when one is not doing what it supposed to be doing? Well, sometimes it is doing what it's supposed to be doing, and it's still bad. I defined a weapon of mass destruction to be one that has a large scale implications. It's important. It's also secret, and it destroys people's lives. So those three things put together make it a really destructive weapon of mass destruction. The example, I think about is the teacher value out of model, where we get teachers fired by the stuff, they don't understand it, but at the same time, it's just a bad statistical models. Some people get a six out of 101 year, 96 out of 100 the next year, and they didn't change the way they're teaching. So it's actually a meaningless thing, but it has strong consequences, and it ruined people's lives.

[MUSIC PLAYING] CLIFTON LEAF: So curing cancer seems like something that would happen in the hospitals or the clinics in your doctor's office, not in computer labs. In this space here, we saw the for Watson for Oncology, but applying analytics to human DNA and DNA of cancer cells is a promising frontier of cancer research, as we've heard and seen this week. It can help patients get the best treatment for the type of cancer they have, minimize the negative impact of that treatment on them, and ultimately, save lives. The huge size, though, of the data sets and the complexity of sharing those data sets among researchers threatens to put the brakes on progress, but the experts who are committed to the cure are racing to the finish line. GREG SIMON: Big data is right now the big yellow object everybody is in love with. But we still live in an information scarce medical world. Compared to in finance, in the 1920s, they created-- I mean, going back to the 1920s at the University of Chicago, you can find the-- every stock transaction ever done between then and now, and yet, you can't get your medical records out of your hospital. So big data is an idea, and it's an idea that's happening in a few places, but genomics is a very small part of it. Proteomics takes you from 30% knowledge about what could happen to 60% or 70%. It's very expensive to do. It's very personal to do. When the vice president cut the ribbon, so to speak, on the genomic data commons at the University of Chicago, that's just raw genomic data from the Cancer Genome Atlas. There were 14,000 people. Now there are 32,000 people, soon to be several thousand people. It's been accessed 5 million times since June. That's the good news. The bad news is when they started the Cancer Genome Atlas, they spent $100 million to get data from 50,000 people-- 24,000 people-- half of it was not usable. So big data is right in the middle of bad data being created by lack of pathology standards and lack of medical standards, and lack of transmission of the data ubiquitously instantly the way we do with financial data, the way we do we do with weather data, you name it. So can it make a difference? Yes, but we have to change the culture of sharing information, because our ability to share is far outpaced our attitude about sharing. DR. KEN ROBERT SMITH: We have a lot of genealogical data, and we have been able to bring millions and millions and millions of records of genealogies into electronic format, link that to two cancer records, and as much as we want to and we and we need to be focusing on genomic information, genetic testing, a very important component to big data in cancer is risk stratification. So I want to know whether I'm in a high risk family or am I in a normal risk family. And if I can do that, in many ways, that can be as valuable-- perhaps even more valuable-- than providing my DNA to a sequencing platform to determine whether I carry this variant or that variant. They can both be used in conjunction, of course, but it's an aspect of a big data that I think we need to be talking more about. CLIFTON LEAF: Has there been more of an effort into categorizing all of this information that doesn't come out of the lab? GREG SIMON: Well, we've been part of creating a lot of collaborations around data. There's a protease genomics project with the DOD and the DOE, and the NCI. With the genomic data comments I mentioned earlier, we have a very simple program on Facebook called Cancer Base. All of our cancer statistical data is five years looking backwards. So if you go to cancerbase.org on Facebook, and it says, do you have cancer, what kind do you have, what stage are you in, where do you live, and it creates a global map of people with your cancer, where they live, and the second and third phase of that program will tell you how those other people are being treated so that you can compare what your situation is with them. Because some of the most important big data is what happened to your parents and where do you live. Those two pieces of information are more important than all your genome, unless you have a single gene disease, which you probably got from your parents who had the same thing. If you don't know what happened to your parents and we don't know where you live, then we're not going to intervene in the right way at the right time. CLIFTON LEAF: It's funny, if you follow some of the election coverage, there has been reports of companies like Catalyst, which have down to the level door bell analysis of every voter, whether they voted, what priority registration, but also where they voted in previous-- what magazines they subscribe to. We have that for voters, but we don't have that for cancer history. I mean, Ken, you've been working with the most, not homogeneous, but a group of people that have consented to a lot of data information and personal information about themselves. How unusual is that? How do we export that model to other public research? DR. KEN ROBERT SMITH: So what we've tried to do in Utah is to, both for the purposes of addressing cancer, but really for most of the serious diseases is to try to marshal all of the data that we can legitimately have access to. Interestingly, you mentioned voting, we have agreements with our voter registration office in Utah to get all of the voter registration data from the population. We also get it for all the driver's license holders in the state of Utah, and we have a legal system for doing this, and I mention this only because the insights about where the risks are for cancer or any other disease can really come from almost any type of data. So you mentioned about knowing who your parent-- what your parents risk was and where you live, well, we need to be able to look wherever we can look to get that information. And so our strategy has been to develop these relationships with our departments of health and other organizations so that we are-- they are able to share the data with us. We create this amazing connection of all the data, and you can create that big data sphere that you can query, and there are peculiar stream-- I'll call it peculiar streams of data that you may not have thought about that would go into it for the purposes of doing cancer surveillance, but they all contribute to it. So if I know where you vote, I know what district you're in, I know where you live, and from that, I can attribute a great deal of information about you, and that's another piece of information. So we're trying to go from the genome to the menome, and learn as much about you as we can, and just to be clear, we are very sensitive-- you mentioned consent earlier, so we have all sorts of protections about how we do that in a way that makes everybody comfortable and provides the benefit. [MUSIC PLAYING] 

Insights: Medical Experts on How Big Data is Being Used to Treat Cancer

How Big Data is Being Used to Treat Cancer

They say one problem is that there still isn't enough data. (November 2016)

curing cancer seems like something that would happen in hospitals or clinics in your doctor's office, not in computer labs in this space here we saw before Watson for oncology. But applying analytics to human DNA and DNA of cancer cells is a promising frontier of cancer research. As we've heard and seen this week, it can help patients get the best treatment for the type of cancer they have minimized the negative impact of that treatment on them and ultimately save lives. The huge size, though, of the data sets and the complexity of sharing those data sets among researchers, threatens to put the brakes on progress. But the experts who are committed to the cure are racing to the finish line. Big data is right now, the big yellow object everybody's in love with, but we're still living in information scarce medical world compared to, you know, in finding it in the 19 twenties they created, I mean going back to the 19 twenties at the University of Chicago. You confined the Trans every stock transaction ever done between then and now, and yet you can't get your medical records out of your house. So big data is an idea, and it's an idea that's happening in a few places. But genomics is a very small part of it. Proteomics takes you from 30% knowledge about what could happen to 60 or 70%. It's very expensive to do. It's very personal to do when the vice president cut the ribbon, so to speak, on the Genomic Data Commons at the University of Chicago. That's just raw genomic data from the cancer genome Atlas. There were 14,000 people. Now they're 32,000 people, soon to be several 100,000 people. It's been accessed five million times since June. That's good news. The bad news is, when they started the cancer genome atlas, they spent $100 million to get data from 50,000 people. 24,000 people. Half of it was not usable. So big data is right in the middle of bad data being created by lack of pathology standards and lack of medical standards and lack of transmission of the data ubiquitously instantly the way we do with Financial day, the way we do with weather data, you name it, So can it make a difference? Yes, but we have to change the culture of sharing information because our ability to share is far outpaced our attitude about sharing. We have, ah lot of genealogical data, and we have been able to bring millions and millions and millions of records of genealogies into into electronic format. Link that to cancer records and as much as we want to. And we and we need to be focusing on genomic information, genetic testing. Ah, very important component to big data in cancer is risk stratification. So I want to know whether I'm in a high risk family or a rhyme in a kind of a normal family. And if I could do that in many ways, that can be as valuable, perhaps even more valuable than providing my d n a, uh, to a sequencing platform to determine whether I carry this variant or that various. They can both be used in conjunction, of course, But it is, isn't it? It's an aspect of a big, big data that I think we need to be talking more about. Has there been Maur Oven effort into categorizing all of this information that doesn't come out of the lab? Well, we've, um, been part of creating a lot of collaborations around data. There's a propio genomex project with the D. O. D in the D. O E. And the guy with the genomic data comments I mentioned earlier. We have a very simple program on Facebook called Cancer Base. All of our cancer statistical data Five years looking back. So you go to cancer base dot or GE on Facebook and it says, Do you have cancer? What kind do you have? What stage are you in? Where do you live? And it creates a global map of people with your cancer where they live. And the second and third phase of that program will tell you how those other people are being treated so that you can compare what what your situation is with them. Because some of the most important big data is what happened to your parents and where do you live? Those two pieces of information are more important than all your genome unless you have a single gene disease, which you probably got from your parents who had the same thing. If you don't know what happened to your parents and we don't know where you lived and we're not going to intervene in the right way the right time. You know, it's plenty if you follow some of the election coverage. There's been reports of companies like Catalyst, which have, you know, down to the level doors, doorbell, analysis of every voter. You know, whether they voted, what party registration, but also where they voted in previous. You know what magazines they subscribe to you. We have that voters. But we don't have that for cancer history. I mean, can you? You've been working with the most not homogeneous, but but a group of people that were have consented to a lot of data, information, personal information about themselves. All right, How unusual is that? How do we export that model to other sort of public research? So what we've tried to do in Utah, um, is to, uh, both for the purposes of addressing cancer. But really, for most of the serious diseases is to try to marshal all of the data that we can legitimately, uh, have access to interestingly, uh, the you mention voting. We we have agreements with our voter registration office in Utah to get all of the voter registration data from the population We also get it for all the driver's license license holders in the state of Utah. And we have a legal system for doing this. And I mention this only because the the insights about where the rifts are for cancer or any other disease can really come from almost any type of data. So you mentioned about knowing who your parents, what your parents risk Woz on DDE where you live? Well, we need to be able to look wherever we can look to get that information. And so our strategy has been to develop these relationships with our Department of Health, uh, and other organizations so that we we are they were able to share the data with us. We we create, uh, this amazing connection of all the data and you can create that big data fear that you can query and there are peculiar steam. I'll call a peculiar streams of data that you may not have thought about. That would go into it for the purposes of doing cancer surveillance. But they all contribute to it. So I know where you were. You vote. I know where what district you're in. I know where you I know where you live. And from that I can I can attribute a great deal of information about you. And that's another, uh, uh, piece of information. We're trying to go from the genome to the mean ohm and learn as much about you as we can. And just to be clear, we were very sensitive. You mentioned consent earlier. So we have all sorts of protections about how we do that in a way that that makes everybody comfortable and provide the benefit.

Insights: The Power of Big Data

Insights: Using Data to Slash Carbon Emissions

Latest video

One use is building structures that are more energy efficient. (September 2019)

Videos (8 videos)

Insights: Using Data to Slash Carbon Emissions

Insights: How Big Data Can Help Agriculture Respond To Climate Change

Insights: Who Owns And Protects Our Medical Data?

Insights: How Fitbit is Considering Monetizing User Data

Insights: Using Data to Operate Hydroponic Farms

Insights: How Pfizer Is Using Data To Improve Medical Diagnoses

Insights: Identifying and Avoiding The Dangers of Big Data

Insights: Medical Experts on How Big Data is Being Used to Treat Cancer