 All right, well, hello everyone. Good morning. Good afternoon or good evening depending on where you're joining us from today. Welcome to engineering for change or you foresee for short today. We're pleased to bring you the latest webinar in our series focusing on artificial intelligence and data analytics for human development. My name is Yana and I'm the president of engineering for change and I'll be moderating today's webinar. The webinar you're participating in today will be archived on our webinars page and our YouTube channel. Both URLs for those channels are listed on the slide information on upcoming webinars is available on our webinars page. E4C members will receive invitations to upcoming webinars directly. If you have any questions, comments, and recommendations for future topics and speakers, please contact the E4C webinar series team at webinars at engineeringforchange.org. If you're following us on Twitter today, please join the conversation with our dedicated hashtag hashtag E4C webinars. Before we move on to our presenters, I'd like to tell you a bit about engineering for change. E4C is a knowledge organization, digital platform and global community of over 1 million people, including engineers, designers, development practitioners and social scientists who are leveraging technology to solve quality of life challenges faced by underserved communities. Some of those challenges may include access to clean water and sanitation, sustainable energy solutions, improved agriculture and more. We invite you to become a member. E4C membership is free and provides access to news and sell leaders, such as the ones you're going to hear from today, insights on hundreds of essential technologies in our solutions library, professional development resources and current opportunities such as jobs, funding calls, fellowships and more. E4C members also enjoy a unique user experience based on their site behavior and engagement. Essentially, the more you interact with the site, the better we'll be able to serve your resources aligned to your interest. We invite you to visit our website to learn more and sign up. If you're interested in learning more about data collection and other ICT or information and communication technology tools, we invite you to explore the solutions library after the webinar. An example of the type of tech you'll find is the simple Vero scanner and platform, which combines a wireless biometric scanner, a mobile app and the cloud to enroll and match people to their digital records with a touch of a finger. There you can learn more about technical performance, compliance with standards, academic research and user provision models of these kinds of systems. All the information is sourced by E4C's research fellows and reviewed by our community of experts and it's all available to E4C members, free of charge. A few housekeeping items before we get started. I see some chats already, so this is great. Let's practice using WebEx by telling us where you are in the world. So in the chat window, which is located to the bottom right of your screen. Please type in your location. If the chat window is not open on your screen, try clicking the chat icon at the bottom of the screen, which is in the middle of the slide. It looks like a little talking button. So I'm going to go ahead and write my location and we'll see if everybody else is able to see as well. So we have folks from London, from New York, myself from Montreal, from Iowa. Let's see, Panama, Chile, Dhamam, Saudi Arabia, Hamburg, Germany, London. Lots of folks joining us from Pakistan and more. Great. Please do use the chat window to go ahead and continue these conversations. You can also share remarks, share resources. If you have a technical question, you can also feel free to send a private chat to the engineering for change admin. During the webinar, we encourage you to use the Q and a window located below the chat to type in your questions for the presenters. Again, if you don't see it, please click the Q and a icon at the bottom of the screen in the middle of the slide. So I see some folks did answer this question also in in the Q and a window. I see folks from Egypt and other locations. Yeah, just try to keep those kinds of comments, the chat window and Q and a is specifically for Q and a so that we go ahead and we keep track of those questions. For our presenters, if you are listening to the audio broadcast and encounter any trouble, try hitting stop and then start. You may also want to try opening WebEx in a different browser. You first see webinars qualify professional engineers for one professional development hour. To request your PDH, please sign in and go to your member dashboard to access the PDH form. The URL is also listed on the slide. All right. Thank you again for joining us. We have lots of folks here from all over the world today from Kenya to India, Columbia to Tennessee to Princeton, New Jersey. Welcome everyone. We're so thrilled to have you with us today. Obviously, lots of interest in this particular topic. So our first speaker today is Shashi, who is the CEO of the Institute for transformative technologies and the founder of the light Institute at the Lawrence Berkeley National Lab. He is the lead author of the recently released ground baking study to identify the 50 most critical scientific and technological breakthroughs required for sustainable global development. We are thrilled and honored to have Shashi join us and I'm going to go ahead and pass control of the slides to you so you can advance them. And let's jump into it. Excellent. Hello, everyone. Hope you're doing well. Thanks for joining this. What we'll do is talk about AI in the context specifically of the sustainable development goals coming up in 2030. Now, nowadays we live in an all AI all the time world and the conversations tend to range from AI saving the planet and humanity, all the way to destroying everything. And the problem is, obviously, they can't all coexist on all these scenarios. The problem for us really is that the conversations tend to be in huge generalities. And it's very important as organizations working in the social change, social justice, human development space make decisions about where to invest their money and their energy. It's very important to be much, much more specific about where AI can help. And specifically if you're talking about the sustainable development goals, there are a lot of organizations that are gearing the strategies, whether it's funding, whether it's implementation, whether it's policy towards that. And so it's extremely important to unpack, first of all, what we mean by AI and then specifically look at the topics in the sustainable development goals and understand where AI is relevant. Where it's essential, where it's nice to have, and in some cases, where it's a distraction. And what we'll do is start a little bit with understanding AI. The most important thing to understand about AI is it evolves continuously. Obviously, what was AI, what was considered AI in the 80s and 90s is today just considered commonplace analytics. So if you think about financial services going back to the 80s, 90s, obviously there was data collected about individuals and customers, whether it was age, whether it was income and so on and so forth. And financial services companies like banks made decisions on the basis of that, is this individual credit worthy, should we do business with them, how much interest should we charge them and so on and so forth. Now, as the world was able to collect more data about individuals and as computers started becoming more powerful, what we saw is that the algorithms to analyze and use that data also became more sophisticated. So if you go to the right-hand side of the slide, what you see is the early days of neural networks and without geeking out too much about neural networks themselves, what you see is if you compare the left versus the right, you start seeing that layer of bubbles in the middle. In the parlance of neural networks, it's called a hidden layer. And what it allows you to do is, even as you collect lots of data, if you're not able to find a really good fit for the analysis you're trying to do, you can try and cast it into an alternative dimension, alternative sets of dimensions. And what that allowed was really black box kind of curve fitting and analysis. So you're able to start analyzing things that were not remotely possible in the previous generation. Now, the only constraints on how much you could analyze, and again, going back to the 90s, if you think about what computing looked like those days, even the most sophisticated labs had large desktop computers. And the idea of a tiny phone in your hand having a ridiculous amount of memory and processing capacity was unimaginable. Fast forward to today. What you see is that obviously there's data about us being all sorts of actions that we do, data is collected about that, where we are, what we buy, what we wear, and so on and so forth. And at the same time, computing obviously has also become enormously more powerful. And so what you see is, again, just going back to a little bit of the neural network parlance, instead of just one level of depth in that hidden layer, you can have as many as you want. You can have millions and millions of data points, you have incredible computing, and therefore you can have hyper customization of things like services. Now, you can see, you know, if you shop on Amazon or something and that sort, you can see that level of customization starting to be used. On one hand, when it comes to conveniences like, you know, hyper customization, like, like shopping exactly what you want, where you want, when you want it, and so on and so forth. Those conveniences are today made possible because of this, but the underlying question for those of us focused on human development is, well, is this really relevant? And so in that context, the first question to ask is, remember what I said earlier was, the constraint ultimately is the amount of data available and the type of computing available. Now, with respect to data, it is obviously wonderful that the mobile phone, the smartphone revolution has been taking place in low income countries. But if you compare that to the kind of data infrastructure there is in industrialized countries, what you obviously notice is that, you know, it's what is nice to have in a mobile phone data. It doesn't come close to comparing to the data platforms, the data infrastructure that you have in wealthier countries. And in emerging economies like in India, you know, with the systems like the Aadhar, now you're starting to see more and more of a data infrastructure. So what we've done is introduced this idea of the data density index, the point there being that the more data there is about individuals, about communities, systems, countries, the more useful things, and honestly, the more non-useful things you can do with it, but you do need a certain threshold of data before you can start doing any real analysis on it. And what you see there is a smattering of countries along that data density index, and the left end of the spectrum are countries that we consider, according to this index, data deficient, meaning that there is some data starting to be collected, but really you can't do a whole lot of powerful analysis about it. The section in the middle is what we consider data sufficient, which means that it's not ready for real powerful deep learning type of AI, but still you can do some really solid analytics. And then to the right end of that spectrum is stuff where there's a lot of data and you can really, really go nuts with the kind of analysis you can do. The other set of questions to answer when it comes to AI and big data for the sustainable development goals in particular is if you take the idea of irrigation fertilizer, as we know food insecurity is a huge problem, right? And in sub-Saharan Africa, barely 5% of small-holder farmers use irrigation and even fewer use fertilizer. So one really important problem that has to be solved is access to fertilizer and access to irrigation. And so that vertical access you see, it starts with direct intervention. So improving access to fertilizers and so on and so forth is what we would consider a direct intervention. The more irrigation there is, the more agricultural yield there will be and the better the food security situation will be. Then you have what we consider secondary enablers or second-order enablers. So imagine now if a farmer has access to microfinance. A farmer can use that microfinance to buy fertilizer or do something else, but it doesn't automatically lead to yield improvements. It is a very strong enabler. And then we have what we consider tertiary enablers. So this is, now imagine if you had, if the farmer had data, you know, whether it's through AI or some other like push mechanism or an app, if the farmer had information on when to apply the fertilizer and how much exactly and to what kind of crop, that would help, but obviously they'd have to be able to have access to the fertilizer and be able to buy it. On the other axis is dependence on AI and on big data. Some, a lot of stuff can be done without a whole lot of data or certainly not big data. The second thing there is the second threshold along the spectrum is yes, there is data and you can do really good, really good stuff with conventional analytics. And what we mean by that is that this is not deep learning by any stretch. This is stuff and algorithms and tools that have been around for the last couple of decades. And, you know, it's powerful enough and proportional to the kind of data available. And then at the right end of that spectrum is a set of tools and sort of interventions that really can leverage AI and big data. Now, what we've done is taken in going back to the study that Janet mentioned earlier, the 50 breakthrough study. Now our institute bases a lot of its work on this body of analysis we did a couple of years back called the 50 breakthroughs. What it does is methodically across all the aspects of the sustainable development goals, identify what the most important interventions are, and then puts the technology lens to it saying, well, is technology necessary for implementing this intervention? We do the same thing through an AI and a data lens and granted this is a very dense slide. I'm not expecting you to read the text, but what I want you to see is just the number of bullet points. So here we're looking explicitly at food security, health, energy access and education. And the dimension I mentioned earlier, which is how many of these really important interventions actually need any data, how many of them can benefit from conventional analytics, and how many of them need big data and AI. And as you can see, those numbers, those number of bullet points start dwindling as you go from left to right, which means that you can do a lot of really good stuff without a whole lot of data. Again, examples being things like clinics and clinicians and fertilizer and irrigation and so on. And then there are some really valuable things you can do with solid analytics. So this is around credit scoring, you know, 404 microfinance and so on. And then there are a small number of things that you can do with AI and big data. Now, some of these are incredibly powerful, particularly things around health diagnostics, but the numbers start shrinking as you go from left to right. So if you map all of these bullet points on that two-dimensional matrix I discussed earlier, what you see is that most of the really powerful interventions actually don't, they're first of all direct and a smaller number of them are second order and a considerably smaller number are tertiary. And similarly, most of them require either conventional analytics or no real data at all. And again, as I said earlier, there's a small number of very powerful things that can be done with AI. So in summary, as we are individually and collectively deciding how much to bet on AI and where to place those bets, the most important thing we would suggest is let's make sure the foundational things are done well. People have to have access to medicines and to things like, as I mentioned earlier, irrigation and clinics and teachers. Then there are a lot of wins we can get using conventional analytics without going too crazy on AI. Importantly, there will come a time, not necessarily in the next few years, but there will come a time when powerful data algorithms can really add a lot of value. And in anticipation of that and to prepare for that, it's very important to start building underlying robust data infrastructures. One example of that being India's Aadhar system and India stack. The more that is replicated and the more the world catches up with a lot of the foundational interventions, the easier it will be for us to deploy big data and AI over time. Thank you. And over to you, Yana. Let me pass the ball back. Thank you so much, Shashi. So without we're going to transition to Zia Khan and for those of you who don't know him, he is the Vice President for Innovation overseeing the Rockefeller Foundation's approach to developing solutions that can have transformative impact on people's lives with a focus on innovative finance, data and technology and science. He writes and speaks for Columbia Leadership, Strategy and Innovation, as you will hear today. Mr. Khan has served on the World Economic Forum Advisory Council for Social Innovation and the U.S. National Advisory Board for Impact Investing. He's an investor and an advisor to a range of impact-oriented enterprises. And he'll share with us a little bit regarding Rockefeller's perspective on data. So over to you, Zia. Thank you, Yana. And thank you very much to the Engineering for Change group. It's a real pleasure to be here and share a few perspectives. And one of the joys of my job is I get to work with smart people like Shashi, who I think provided a really good analysis of AI for development. I just want to share a brief perspective on what all this means for development. I'm not an AI expert by any means, but we at the Rockefeller Foundation who have been in the business of driving big social change in health and agriculture and employment opportunities for over 100 years are incredibly excited about the developments that AI offers us to help have more impact to more people around the world. Now, what I'm showing on the screen right now is actually a graphic that I discovered in one of our annual reports from 1918. And the point I'm trying to make is that data was actually at the heart of how the Foundation has operated for over 100 years. I won't go into all the details, but basically this was an effort in Arkansas to eradicate malaria. And what the team would do is they, in a controlled way, would go into different communities, try different interventions, and measure the spread or the preponderance of malaria. And you can just sort of see visually there's a little bit of a before and after fact that makes a pretty compelling case to identify what is it that's working. And once the Foundation would find what worked, they would then quickly scale it up. Now, back in that day, there was a lot of investment that had to make using a hypothesis-driven approach in collecting the data and gathering the data. And frankly, not much changed over 100 years in how the development sector worked. Typically, we would have a few hypotheses around what interventions could work. We'd spend some money to create the data collection approaches for that and assess what's working, what's not working, and try and scale those things up. But with the admin and AI and big data, we can take a fundamentally different approach. And now we move on to this second chart. And what is being shown here is a very fine-scaled map somewhere in Northern Kenya. This is about a 5 kilometer by 5 kilometer view. And down at the pixel level, it gets to about 10 meters by 10 meters. That shows crop yields for maize. So how much maize is a certain area producing? And you can see the legend in the bottom corner, which is tons per hectare, where red is fairly high and blue is fairly low. Now, what's interesting about this chart is this is data that's actually readily available and is being converted and processed to understand what is happening on the ground. In this case, there's a company that the Rockefeller Foundation helped start called Atlas AI that's based and founded by some Stanford professors based on some research that they did. Where what we're taking is satellite imagery data and using some other training data sets and machine learning algorithms figuring out how to get really fine-scaled resolution on what is happening on the ground. And you can imagine this being useful for a range of folks. We're in conversations with people at the World Bank who are very interested in understanding the net effect of their various interventions. There's the Alliance for Green Revolution in Africa that's very interested in understanding how can fertilizers be used and other inputs more productively. And then a lot of private companies that are interested in understanding this from a marketing and business development perspective as well. And the point here though is that we no longer have to invest in producing the data as much as we had to. We have all this data around us and AI is able to help us use this big C of data and reveal really important insights so that we can get better at predicting what will have an effect. We can get better at understanding the cause and effect of different interventions and really figure out which ones are the ones that are worth scaling up. And we're just tremendously excited about the potential. As Shashi mentioned, there's a whole other conversation when it comes to the use of AI in terms of what effect it will have on workers and what effect it will have on jobs. That's an area that we're exploring as well. There's also a whole conversation around privacy and what rights do people have to their data and how do we create data trust and data commons. But all this to share that from a development perspective and from the business of driving social change, we're incredibly excited about the potential for AI and people like Shashi and others are helping us understand how we can use it appropriately and most effectively. Thank you so much for that overview, Zia. So you actually teared up very nicely for the conversation that we're going to have today jointly about some of these other nuances related to AI and data analytics. So what everyone should be seeing, the attendees should be seeing on the screen right now, some questions that we're going to explore together with our panelists today. And if you have additional questions, I encourage you to enter them into the Q&A window so we'll tackle them towards the end. So as you mentioned, this is an area that is riddled with learning to be done. And one of the first questions that we are curious to explore is whether development players are really considering the right criteria generally before investing in AI or data analytics interventions. Shashi, obviously, you shared some thoughts on how that should be thought about, frankly, and what the initial options are. But what has been your perspective generally or what have you observed in the development sector? Is the right criteria being applied? One way to use data is to select the right intervention out of a menu of interventions. Another way to use data is to do an X post-facto assessment of how an intervention is taken place, how well it is done. The third question is whether data is the intervention unto itself. And as long as we are very clear how we're using data, not confusing one for the other, meaning that the fact that whether it's AI or just another data-driven tool, we should make sure that we're not confusing an assessment tool for an intervention. And as long as we're clear about that and about the role of data, I think we'll be okay. And right now, obviously, because it is a bright new shiny object, I suspect there are a number of cases in which there is overinvestment in data, but that will correct itself over time. We're not overly worried about it as long as we can make sure we're using the right lens, the right filter, and not putting too much effort into data, assuming that it will naturally lead to the more foundational investments. And Ziya, in your experience, I guess this also can apply to impact ventures. Are they kind of jumping on the shiny new tech bandwagon or do they think about the right parameters before pursuing an AI or data analytics strategy? It's a great question. And I think I would say there's two questions that you need to consider, I think, in this space. What can you do with data and AI, which tends to be a technical question, and what should you do with data and AI, which is not a technical question, but people conflate it as well, and we're seeing all sorts of evidence with that with the various social media platforms as well. You know, one of the challenges in this space is the unintended consequences are still something we don't fully understand. And we at the Foundation are... we're trying to figure out what is the right middle balance. On the one hand, we don't want people to get too conservative about the potential consequences because we still have a lot of poverty in the world. We're still facing climate change, and we need to move with some urgency on all these, and we think there's enormous potential. On the other hand, we want to be conscious of what are these issues. And so for us, the right approach is really to try and integrate different perspectives together. How do we make sure that technologists are having the right conversations with civil society, that they're having the right conversations with government, and that we can come up with some guiding principles or frameworks around how to use technologies like AI data and also blockchain. Data is particularly interesting because of the privacy questions, and different countries are taking different approaches. China has an approach where the government has a lot of access to data, and the U.S. we take one where the private sector has a lot of access, and India is trying to develop an approach where the user has a lot of access. There's always going to be a tension between what is good for the individual and what is good for the public and society. And I think public health is a classic example where the more data we have, the more we can do around public health, but medical data and health data is a very personal and private thing. So I don't have any answers to this. I think we're starting to understand where the tensions are and our role that the foundation hopefully can play and other actors is how do we create the right bridges and conversations to come up with guidelines around this. That's truly interesting. And you kind of bridged the other question that we were trying to explore around data privacy standards and practices. So, Shashi, do you want to build on that and, you know, give in some of these regional approaches to privacy? Is there anything, any guidelines that you want to share or, you know, have the audience consider as there may be exploring this for their own projects? Yeah, you know, Zia hit the nail on the head when he said it's important to distinguish between what can be done and what should be done. What should be done. And then there's another corollary question to that, which is how should it be done? And that is still emerging. To be told, even the most, I live in California and even the most data savvy part of the world, so to speak, is struggling with questions about privacy and standards. Again, I don't consider myself a privacy protocol expert, but clearly that balance between an individual's privacy versus kind of metadata about the individual that can and should be used for the general good. That's something that currently, you know, people working on data solutions have to wrestle with. And the challenge is they're wrestling, they're being forced to wrestle with that question in the absence of a robust policy framework. Now, the thing with data is, particularly private data, is once it's out there, it's out there. It's very, very hard to take it back. So there's a real urgency around making sure that the right protection mechanisms and the right policy frameworks are in place. My worry is things aren't moving fast enough. But then just going back to that question you asked, I guess the next question, which is, you know, is it worth it, right? Are there risks of the widely deployed data acquisition backgrounds? Do they outweigh the benefits? I don't think so. In any technology, any generation of technology, there is no free lunch. So my sense is that data will start adding more and more value. I mean, right now, as I mentioned earlier, there are limits to what can be done in countries that we consider data efficient, data deficient. But, you know, even if we put the perfect protocols in place, people who want to exploit data for less than savory purposes will continue to do so. So this challenge isn't going away anytime soon. But overall my sense is that like with any other technology advance, it's a double-edged sword. Of course. Do you have any recommendations around resources or case studies or some examples that really highlight good practices in terms of data privacy by key players that you might be able to point the audience to or we can follow up with? Well, you know, there are a couple of examples that we can build on. So if you think about health data, you know, in the U.S., there was a lot of concern about individual health data being exploited by insurance companies for denying insurance and things of that sort. And so a few years back, quite a few years back, the HIPAA policy framework was put in place. It's essentially a framework for protecting individual health, even though, say, the broader health trends and so on can be used for important decisions. Now, what we're seeing is in our health work in Kenya, for instance, right, where there isn't necessarily a fully evolved set of data protection standards, it turns out that something like HIPAA is being considered a gold standard. Now, similar standards are there for protections about on the financial data and so on. But what you don't see is stuff that people post themselves. So if you put data out there voluntarily, there really isn't anything to protect against that yet. The most troubling thing is stuff that companies collect with your implicit consent, stuff that you're signing off on the dotted line, allowing them permission to collect this data. But you're not actually aware about the extent and the volume of data that's being collected about you. For that, unfortunately, there isn't any example yet. And actually one of our... Oh, sorry, go ahead, yes, sorry, apologies. Yeah, I just wanted to add a couple of points. I think there are probably some good examples around policies. I think what's important is whether to be frameworks of the right questions to ask yourself in your situation. So, for example, we worked with the Beck Center at Georgetown University to produce an ethical framework for the use of blockchain, but it doesn't have any sort of answers, but it almost gives a decision tree of these other questions you should be asking yourself. The second point I'll make is privacy itself is a pretty fuzzy concept. It doesn't have a specific definition, and it's very culturally shaped. So, for example, if you look through the history of the United States, we don't have a general concern around a police officer walking around and writing down license plate numbers on a notepad near a scene of a crime, bothered by the idea of a CCTV camera capturing license plates of everyone going down a street. So, we find that in conversations with folks, the idea of identifying what specific threats are you worried about and how do you respond to those threats is probably a more productive way to have a conversation about privacy than just privacy as some general abstract concept. Right, it's a very good point. One of our listeners actually, I could kind of contribute into these examples that should be thought about. So, Eki noted that here she actually, I can't tell for the name, so apologies. A Zilla listener noted that people use Google to research symptoms about their health, for example, which is data can be used towards diagnostics, apps on smartphones that use the phones capability to gather their pulse, blood pressure, etc. So, she noted, thank you for clarifying, she noted that privacy can be opt in and opt out. So, that's a very good point and I think contributes to, you know, to your conversation here. So, we've been hearing a lot about how the risk of bias is a big concern in development and applications of AI. What steps are being taken to prevent further marginalization of the underserved and whoever wants to jump in to talk about this first is welcome. Yeah, after you, my friend. I'd like a couple of quick points. Thank you very, very generous of you. I think bias can happen in a few ways. One is in terms of, and I think you framed it well, like there's the application and one big effort of ours is just how do we make sure that social sector organizations get access to the modern techniques and approaches that can really help them out. Because data scientists are, you know, it's a hot commodity these days and it's hard both because of the nature of the work but also frankly because of the labor market for a lot of these social sector organizations to get access to them. So, we're trying to do a couple of things. One is we have invested in a group called DataKind, which is an amazing group. They're sort of like doctors without borders for data scientists where they take data scientists working in different companies and try and connect them to social sector problems. So, for example, a data scientist at Netflix who focuses on demand prediction can get connected to a water utility that's struggling to manage for demand to do some work and also that water utility is saving $25 million a year. So, we're trying to encourage and figure out how we get the application more broadspot. The development is a really important point there and there's all sorts of things that are coming out, particularly when you look at the justice system and how when based on the data sets but also frankly the cultural mindsets and approaches of the developers are coming from a more homogenous community that's anchored in Silicon Valley, what are the unintended consequences? We don't have any good solutions to that right now. We're trying to encourage diversity in scientists around AI and trying to encourage and support their work and highlight these issues but we don't have any good answers right now. That's a great summary, Zia. This is my biggest, one of my biggest worries about the blind use of data. So, if you take the simple example of people using mobile money right, you know, MPSA in Kenya for instance obviously in Kenya MPSA has made made has achieved tremendous penetration but imagine if you turn the clock back a few years and if you were to use some of the early data generated by MPSA users and try to extrapolate to the broader population on the basis of that obviously, you know, you would only get a very biased sample and you know, the that's going to be a problem in data deficient countries where it is quite likely that the initial rounds of data will not be all inclusive and it even won't be representative of the broader population. So, there will be significant limits on the kind of things you can do to extrapolate from a limited sample of data points and making broad generalizations and policy decisions on the basis that can be quite troubling. And this is where data, the concept of data science is extraordinarily important. You have to know how representative you know, a set of data points about the broader population. To be told, the mathematics of this is actually there. The only question is, will the people using data both through analytics and then broader extrapolation and policymaking, will they use those common sense mathematical tools that have been around forever? That's a golden question right there. Interestingly so, we have time to take questions at the end but I think this one is quite aligned to the discussion so I'm going to go ahead and introduce it now. So, with something like blockchain or the quantum internet do away with the bias and the business of sites like data kind or organizations like data kind. That's interesting. I can't fully understand the question because I'm not sure I understand the comparison but maybe I'll offer a comment in case it's relevant to the question if it's not an answer. I think one thing we shouldn't fall into the trap of thinking is that everything can get automated. We are complex people we're working for the impact of people and so something as straightforward, something that sounds as straightforward as going into an organization understanding their data science needs is actually a pretty complex social process. You know, what is the problem how are they going to solve it. It could be threatening to some people who are invested in doing it a different way it could upend the power dynamics of the organization. So the idea of introducing data science just like the idea of introducing any kind of capability into an organization is not sort of an automatic process. So I have low confidence that there will be some technology that emerges in the next 5-10 years that actually replaces that and that way I really do think there's always going to be a human interface in shaping the problems and understanding and determining what it is that we need to go forward on. That is actually the essence of what DataKind does. They invest a lot of time to figure out what is happening in organizations and the way it's been described to me is which I like this metaphor is it's like thinking you're going to bake a cake and then just walking into the kitchen and finding out you don't have flour you don't have any eggs but you do have a few other ingredients and as you rummage through the kitchen you come up with new options of what it is that you can make. It's very hard for me to think about how you automate that process so if that is what the question is getting at then I'm not so optimistic. What I am optimistic though about is how do we use network technologies to improve that matching between supply and demand. Once we have shaped the demand for various organizations and once we expand the network of volunteers who want to connect and offer their help and are motivated by having social impact then I think there are some technologies that could help us. Shashi, do you have any perspective? Yeah, yeah, you know relevant to the last bullet point on the page you have with questions it's if you any policy school before you graduate with a degree in policy you study statistics and the reason you study statistics is that it's a very important tool kit for you to use to understand how much you can extrapolate from data for making big policy decisions. Nowadays lots of people are both subject to decisions about them being made with data and an increasing number of people are using tools black box tools to make decisions using data and so what that means is it will be extraordinarily important for a much broader swath of the population to become data literate and what that means everything from understanding you know how to protect your own data parameters and the other end of the spectrum is to how to use data and how much to rely on whatever data you're using to make your decisions and to Zia's point there's not going to be any general tool that will help you automatically ring fence in a good data versus bad data but it's very important for broader society and individuals to become data literate so that we are much smarter about this very powerful tool kit that we're starting to use. I heartily agree with that particular point I think we're all still kind of capturing up on that front so was that in mind that perspective what is necessary to really leverage these powerful tools for those who are interested in entering this field for the purpose of solving some of these intractable problems how would you recommend they get involved beyond say educating themselves and becoming more data literate generally what else how else would they equip themselves to be effective I think you like that sure I can go first so you know there's this phrase hammer looking for a nail so for those of us in the technology for good space whether it's data or whether it's any technology it's much much more important to understand the underlying problem we wish to solve than it is to be a deep expert in any particular technical field whether it's data or otherwise the worry with approaching this from a pure data expertise lens is that everything looks like a data problem and so without investing a lot in understanding the underlying problem if we try to force fit any solution whether it's whether it's data or anything else I'm afraid that we'll be guilty of hammers looking for nails yeah I couldn't agree with that anymore there is and it's well intentioned but there's a whole range of efforts largely by technology companies to showcase their technology and show how it can have positive impact and I think Shashi's entire presentation was around you know sometimes we might have overkill and you need to get the right tool for the right nail or the right problem at hand so I think the interesting space for us on how I think people can get involved is how do you create that bridge between the supply of interesting data and data analytics and the demand from the sector and so I think the right interface is you know asking questions of to solve this problem is there a prediction problem if we could predict some things better would that be really meaningful important or if we understood the cause and effect is that sort of the question that we're struggling with or are we better able to compare effects or different interventions I think if you can think of those bridging questions and people from the field who want to use data science can translate their field expertise and knowledge into those questions and then data scientists can understand what are the most applicable tools and techniques for those different kinds of questions but there needs to be some kind of space for those to get come together and too often the development sector I think there's a lot of people hearing a lot about AI and they just kind of want to walk into the AI store and find themselves some shiny AI and apply it to a problem and the reality is it's going to take a joint effort a team effort an integrated effort to make the best out of the data supply and data technique supply and the demand of the very real problems we're all struggling to address. Yeah I I'm curious about that because you talk about the spaces and there's some forms that for example come to mind to me where those kinds of conversation can happen like here in New York City there's the Bloomberg data for good exchange which often is a form for that kind of interface. Do you have any recommendations or kind of your go-to's for these kinds of forms where these questions are explored and you know there's an opportunity to learn and not the pigeon hold into the AI store mindset. Either one of you. Well I'll just share that I think there is a meeting of the middle that is starting to happen so I think more and more of the tech conferences are including social sector actors early to frame and shape the conversations around what are the most important applications versus what are the most interesting applications and so I think and again I don't know the entire space but you have conferences like Code for America or others who are much more conscious of those efforts and then similarly I think a lot of the development conferences that happen are no longer trying to treat it like a science fair in terms of showcasing the technology but applying some of their internal techniques so I don't really have an answer other than optimism that this is happening and we ourselves are trying to invest in you know we're calling them charats but kind of like how do we take a new approach to meetings where people can come with various assets and spend real time together to develop meaningful solutions or at least catalyze a new sort of problem framing or new use cases and the use cases is actually a term we're trying to use more and more because we think that is the way to think of how to integrate both the demand that is relevant to users but also the supply of solutions right and Shashi I know that ITP has authored white papers or briefs on this topic are there other recommendations that you have for forums or opportunities for these kinds of conversations to really explore these questions in more depth with other organizations that are also applying these techniques No recommendations per se but just an observation we are on the cusp of something a really interesting decade where as if you think about hardware solutions there aren't that many organizations that can very easily jump in and introduce very powerful hardware solutions but with software with data it's led to a real democratization of the solution space anyone with some basic understanding of coding and access to data can start doing very powerful things so part of me is actually quite excited about this democratization so that we will start seeing a much broader set of solutions being offered to solve the big problems we care about so I'm looking forward to that That's great to hear the optimism there so I do want to address one question that came from our audience regarding the plans for using AI to assess the ramifications of I guess in this case impact the ramifications for example they know that the agricultural intervention of providing fertilizer may and often does lower water quality or cause eutrophication and so forth so any examples or something you want to share regarding the application of AI to understand potentially unintended consequences as you mentioned yeah so it turns out that that was a big consequence of the green revolution which was a big effort through the 1950s and 60s where a much more scientific approach to agriculture was taken in Latin America and Asia and on the positive side it is largely credited with having saved a billion lives and saving entire countries like India and Pakistan from mass starvation there were some unintended consequences to that ranging from concentrating the financial gains to larger land holders and leaving small holder farmers out of it and some environmental consequences so I mentioned that not as an excuse for how we go forward on AI but just to say there are unintended consequences I am optimistic though that AI and big data will allow people to simulate and assess a wider range of consequences than they were able to in the past so while I certainly do not believe that the nature of these problems just by definition means there are unpredictable consequences and such I am pretty optimistic that we'll get a better handle on what are the scenarios that can't play out because we can just do much faster analysis using much wider range of data points than would have been possible before one particular example I'm very excited about it's an organization that Zia works quite closely with it's called Planet and I don't remember Zia mentioned it earlier today but it is in my opinion one of the more remarkable companies out there today they've figured out a way to get tiny satellites up in the sky right and they've now deployed the largest fleet of satellites in history and they so Planet takes an image reasonable resolution image at every point on the planet on Earth every day and so that's an enormous amount of data that can tell a lot about whether it's whether it's pollution whether it's deforestation you know I was at an event hosted by Human Rights Watch the NGO a couple days back and they were able to document in a tremendous detail the atrocities that were being perpetrated by the Rohingya population in Burma and and so you know so that that one tool in particular and I'm not sure what there where the policy is they're on making their data available but you know when it comes to specific things that are happening and observable on the surface of the planet actually think that these satellite images can be incredibly powerful and so for those of you who want to start looking at stuff in your area so let's say you're working in Country X where currently nobody else is doing anything I suspect that you can use Planet Satellite imagery to start doing analysis of your own whether it's to track pollution whether it's to track development and construction of roads and so on and so forth that's a really great piggyback because one of our listeners wanted to know how to make a data poor country a data rich country how to move them on that index that you shared earlier Shashi he notes that or again she I have got to stop using pronouns that time and resources that are needed for that may not be available so leveraging you know these kinds of organizations like Planet Satellite Data if it is available assuming that it is maybe one option are there other pathways yeah you know so my own sense of that journey between data deficient and data rich a lot of that stuff has to come from the government partly because the data will increasingly become a public good but also some of these things are foundational and just too intensive for any individual company to do so the foundation of stuff has to be I think government led on top of that you can have private companies do stuff whether it's you know whether it's the M-Pesa mobile money smartphone stuff or otherwise and then the third layer is citizen data and obviously you need some sort of a platform for citizen data to to be added but you know again if you go to the planet stuff for instance I can imagine someone doing some really interesting analysis using the planet imagery and posting that data in a public forum making it available to everybody the only challenge there of course is the moment you start having citizen data scientists who are collecting data and posting it out there to be very careful about selection bias and so on and so forth but currently my sense is there's enough platforms out there for citizens to start getting involved and start creating their own public data platforms right so interestingly a question also came in which is that is built on that or what global data sets are available to social providers to use to apply AI to develop solutions to global problems that they are addressing beyond planet do you have any other examples I can go and you can jump in again I would rephrase that question right which is what's the problem you want to solve I mean poverty is a pretty problem right and social problems it's a pretty broad category my sense is advice would be rather than phrase the question in terms of how can I use data in AI work forwards from the problem you're trying to solve and then understand the applicability of data driven AI driven tools and then figure out what data you need to build out those tools what it highlights for me though is because I'm drawing a bit of a blank on the answer to the question I think it's actually your framing it in the correct way but it seems to me something that would be useful is some kind of central directory of all the different data sets that are available so if people did have kind of a sense of what problem they are trying to solve they had a place to go to and that's a very helpful I love that question because it sparks some ideas for me around what we might be able to do to be helpful in that it's a very exciting one for us as well I've come across various data sets many of them are sector specific as they should be and region specific again as they should be and this may be an opportunity for us to work even jointly to build at least a starting point kind of guide for some that are already available and freely freely open for folks to use so we have reached time and I really apologize for those of you whose questions we may have not addressed but I do want to thank our Shashi and Zia for their time I know both of you are incredibly busy with a lot of travel and we're really grateful for you setting aside this hour to explore these questions with us the really critical questions and these conversations there are not that many forms where they can be had so it's great that we can use this platform so with that I'd like to thank all of our attendees as well for joining us today for those of you who are interested in your professional development hours do use the link that's provided on the slide or go to your member dashboard to actually fill out the form to get those PDAs if you have questions that we didn't address and would like us to direct them towards the speakers feel free to email us at webinars at engineeringfordchange.org and for those of you who are still looking at the chat one of our listeners was kind enough to share a link to some data from the World Bank so feel free to pull that and last but not least please do become an E1C member to get more information on upcoming webinars and resources that we're pulling together for you so with that I wish you all a good morning, good afternoon, or good evening wherever you may be and we look forward to seeing you on the next E1C webinar as you can see my own technology failed me my earbuds went out during the middle of this call so just a reminder that technology doesn't solve everything but it can be really useful to addressing some of our critical problems thank you everyone have a great day bye bye