Harnessing Digital Technologies to Advance Global Precision Health and Development

Show video

Good morning and welcome to the second  session in our frontiers in precision   population health and health equity. Our  inaugural session focused on how we could   use precision population health approach to in  California for studying COVID-19 in California.   In this, in the vaccination campaign in the  fall, we plan to host sessions focused on AI   and cancer. And today's session takes us in a  more global direction. We'll look at how this   approach can be used to advance global health  and development. Our hope in this series is to   stimulate transdisciplinary discussion around  precision population health and we are excited   to co-host this event with the King Center at  Stanford and the Department of Epidemiology and   Population Health. Today's discussion is being  recorded and the recordings will be available on   the Population health sciences website as well  as the King centers early next week. And it's  

my honor to introduce my friend and my colleague  Steve Luby, who will be moderating this session. Steve is a professor of medicine and infectious  disease is at Stanford Senior Fellow at the   Stanford Woods Institute and the Freeman Spogli  Institute. And he also by courtesy, is a member   of our department in epidemiology and population  health. Before joining Stanford, he worked in   Pakistan for five years and in Bangladesh for  eight years where he continues to do extensive   research in Bangladesh. He has led research teams  and have that have advanced our global scientific   understanding of hepatitis C transmission and safe  injections, water sanitation and hand washing,   NEPA virus prevention, typhoid fever,  epidemiology and prevention. I'd like  

to hand this session over to Steve who will be  our moderator and I'd like to thank him and our,   our speakers for today and I'm really excited  to have be for him to be the host and for us   to have continuing these sessions in population  precision health. Thank you for your attendance   and we look forward to lots of discussion and  would really appreciate your participation in this   in this seminar. Thank you Steve.  I'm handing it over to you. Thanks a lot Melissa and welcome everybody  to this really exciting seminar. I have the   privilege of introducing our remarkable panel.  So we have three panelists today. Manisha Bhinge   joined the Rockefeller Foundation in October,  2016 and serves as a managing director for health   Manisha leads program strategy and manages the  portfolio of global partners for the Rockefeller   Foundation's Precision Public Health Initiative  and more recently pandemic response and prevention   efforts. She has over a decade and a half of  experience in social innovation and implementation   science in global health. Prior to this role, she  created and led the strategic partnership office  

at the Tata Trusts India's oldest philanthropic  organization and launched the India Health Fund,   a leverage fund in collaboration with the global  fund. In addition, Manisha was vice president   at BRAC, the world's largest NGO for over six  years, where she developed and managed programs   that promote access to health, education,  economic empowerment, and social justice. She's worked extensively across Asian, across  Africa and South Asia on community-based service   delivery and women's health and empowerment.  Manisha's experience includes management of   field operations, program strategy and design and  partnership for stealing intervention. She spent   her early career in the private sector working  in technology consulting and then at the un   our second panelist is Giulio De Leo. He is a  disease ecologist in interested in investing,   investigating factors and processes,  driving the dynamics of coupled natural   and human systems and in using this knowledge  to identify levers for health and conservation   that is ecological interventions that can improve  human wellbeing and the health of the environment   that underpins it. In the last 10 years, he's  been particularly interested in investigating  

how the development of water management  infrastructures to support agricultural   expansion and intensification may increase the  risk of the transmission of schistosomiasis,   one of the most important of the  so-called neglected tropical diseases. Giulio does not neglect it. Along with Dr.  Sokolow, he co-founded the Upstream Alliance   partners in schistosomiasis reduction and the  Stanford program for disease Ecology Health   and the environment with the goal of developing  ecological solutions to control infectious   diseases with an important environmental component  in their transmission cycle. And then Brigitte   Gosselink who leads Google leads Google dot  org's work to leverage emerging technologies and   Google's expertise to address global challenges.  She's currently focused on how artificial   intelligence can be used for social impact through  efforts such as the $25 million Google AI impact   challenge with a particular focus on crisis  response and sustainability. She previously   created programs focused on how technology can  improve global education and innovation for people   with disabilities. Prior to google.org, Brigitte  was a strategy consultant for nonprofits and  

foundations at the Bridge Group and worked for the  US Agency for International Development and IRD   focusing on innovative approaches in post-conflict  transitions. She has an MBA from the Yale School   of Management and a BS in Systems Engineering from  the University of Virginia. So over to Manisha. Excellent. Thank you Steve. And again,  thanks everyone for your patience. It's  

ironic. Digital technologies practice  sessions still have little of challenge,   but coming back to Rockefeller Foundation, we  started a precision public health initiative   in 2018 and the rubric that we used there  and the philosophy was how do we ensure that   the right interventions read the reach the right  people at the right time? And that was pretty,   that was the underlying premise of, you know, the,  the rationale for engaging in this initiative.   There was precedence to this  work. So when we started in 2018,   I think we'll hear from other participants.  There was a wealth of evidence and a wealth   of work that we built on. And what we realized  very quickly is there the technology that drove  

what we call pre precision public health, and  I'll get to it in the next few slides, largely   settle arrested on three critique,  three categories of technology. We had the frontline technologies and the example  that I wanna show you here is the personalized   frontline tools that were, that are fairly common  in most of the world with the community health   workers largely that allow them to understand who  they are clients are, what are key challenges that   are, are operating within the catchment area, as  well as allow them to do more periphery analysis   and understand workflow modalities in a more,  in a more accelerated and effective manner. The   second body of technologies that we encountered  were what we call visualization technologies. So   there's one thing about collecting data and  having data there is an entirely different   skillset around visualizing and interpreting that  information in a way that aligns with decision   making and brings forward key interventions  and needs that stakeholders have. And again,   we found a number of examples around that. And  the third is also a source of data, but analyzing,   accessing data through remote sensing means,  again satellite technologies, G S I technologies   and alternative sources of information beyond  health information systems, beyond what can   be gathered through the public health interface  were key areas of opportunity that we identified   and where we have technologies on one one end  and where are applications of those technologies.

And we found there were wide ranging pretty much  the entire gamut of primary healthcare delivery is   a ripe for engagement and action and precision  at large for the simple reason is populations   are heterogeneous, challenges and health  behaviors are heterogeneous and the needs and   concerns within communities also significant  differentiation. So how does one ensure that   delivery of health services are not only targeted  but also streamlined and personalized if you will,   towards specific communities and individuals.  And how does one leverage technology for that   purpose? At the Rockefeller Foundation we started,  and this is again in 2018 with maternal and child,   maternal child and reproductive health as so the  signature focus of our efforts, again looking   very closely at community health workers and the  interface between the public health system and   households and what are the technologies that  work in that interface not only to gather data   off individuals, households, context communities,  but also support community health workers, which   we know are who who we know are paraprofessionals  to upskill their capabilities and provide more   targeted and precise. Again, I'm using that term  services towards those community based on that   data. A second level of engagement that we work  towards is infectious diseases or communicable   diseases and that's largely in the space of tb,  malaria and hiv aids to start with. And as we  

have learned over the past year, COVID has also  been, you know, the remit has increased to COVID. So what does, what do we mean by precision public  health at the foundation? We've again very much   started with, you know, the three factors of  data that we feel that drive precision. First is   availability. Do we have enough of the right  data? Is it timely and is it granular enough   to make inferences that drive health  outcomes or inform health delivery   When we have those three factors, again  availability, acce availability, timeliness   and granularity. And I think the underlying piece  is quality and accuracy being sort of the fourth  

overarching component. We have the opportunity  to solve some of the world's greatest problems.   We can drive scale in a cost effective manner.  Again, remember we talked about the heterogeneity   of health behavior services and disparities. How  does one drive the solutions to scale in a way   that still is applicable? Can we address what were  previously insurmountable barriers to progress? And an example that we, you know, aim to tackle  was just the access to health services writ   large in a world that is deeply constrained  of skilled workforce. Can technology augment   paraprofessionals and or resource constrained  settings to bring services to those communities   and can be used data analytics to understand what  those are and ensure that those are more targeted   in a meaningful manner and broadly address the  health equity divide. And I think knowing who your   customers are is an important factor in delivering  effective services. And if the data on those  

customers and those clients and those communities  are absent largely because they're off grid,   they don't, we we not building  services for those needs. So again,   how does one use the first suite of technologies  that we have available to understand where those   hidden populations are, what we call cold  spots, spots if you will, off the grid,   off the healthcare provision to bring services  and and to them or bring them to the services,   whichever is the more appropriate mechanism. So again, the, the tools and technologies are  important but the purpose and the outcomes   that we want to drive is paramount in term,  especially when we have the foundation set out to   develop this strategy. So what was our approach?  There were three areas that we wanted to focus   on. Again, we were looking at how does one get,  how does one ensure availability, timeliness,   granularity of data driving equity and accuracy  across the board. And three critical components   came to the fore for us to invest and lean  into both from a scale as well as an innovation   standpoint. The first is the data exists, it's  not as though we don't have data in countries,  

it is questionable data. And again I'm looking at  health information systems right large across lms.   Is there, how does one ensure interoperability  within these data sets so that they're not silo   malaria data sitting in one place, primary  healthcare data setting in a separate silo. Infectious diseases are in a third silo, yet it  is the same communities that are recipients of   services largely from same frontline staff and  primary care facilities. How does one create   an integrated view of that community, of that  household and even of that district to ensure   that we are using data and evidence to drive  resource planning, to drive service delivery   and design services that are more customer  and patient focused as opposed to, you know,   traditional modeling efforts while those took us  as far as they have, what's the next generation   to do that in a robust manner, we need to  have digital tools at scale. And again that   the foundation, we leaned into frontline digital  tools because again we felt that the interface   between the primary interface between households  and the health system is the richest interface   where you can gather robust data on specific  needs, households, behaviors and context.  

And finally, what is the processing that's needed?  We have the data, we have it's interoperable yet   to make it actionable, it needs to be processed  and to address critical use cases and challenges   that decision makers need to target. Whether  it is resource planning, whether it is response   to outbreaks and and disease, disease  outbreaks or understanding rationale for a   surge in mortality, mobile morbidity in a specific  community and what we need to do to drive those. So, so that is where our approach leaned into  over the past two years, and this is largely   sort of pre C and early days COVID, we worked  very closely across India and sub-Saharan Africa   and the focus there was again looking at primary  use cases in India we looked at malaria and what   does active surveillance for malaria look  like? How does one inform outreach? How   does one gather the data and how does one ensure  that at the district level we are understand we   are processing and driving analytics to  inform how resources can be deployed in   near real time. So not at an annual basis but  what can we do at a monthly or quarterly basis  

that drives more effectiveness  and efficiency in those spaces. We're right at time now so we really have  to move on. Can you close quickly please? Yes I do. I can. So very quickly we did similar  programs in Uganda looking at essential health   coverage during the pandemic and similarly  in several parts of Africa and this is in   Ethiopia, look using g I s services to find where  there are graphs very quickly summarizing where we   have challenges and I won't let read them out,  but fundamentally data is still paper-based,   countries have different maturity levels and  capabilities are limited. So there's a lot of  

work to be done to reach the goal that and the  aspirations that we have in front of us. So with   that I'll stop and thank you for your patience  while we fixed the technology issues. Thank you. Thank you very much Manisha. Our next speaker  will be Giulio Deo Giulio. Over to you. Thank You very much to much Nisha for your introduction.  I want to follow up with a specific case study   that is about integration ecology withdrawn in  satellite imagery to the, for the control of   neglected tropical diseases. While we are ramping  now with the vaccination campaign and we start to   have some hope in ending the devastating and  tragic COVID-19 pandemics that just today has   exceeded 180 million certified ca cases worldwide.  Well there is a silent epidemic that has been  

going on day after day with no relief for decades  actually for centuries affecting nearly 40% of the   world population. 3 billion people typically  living in tropical andSo tropical countries   with limited access to clean water, to healthcare  and to Sanitation. And I'm talking about a family   of diseases caused by pathogen and parasitic  wars that share free features. They have an  

environmental component in their transmission  cycle, whether freely stage of a parasites or   an animal reservoir for this disease contingent  occurs through infected bites for instance from   a vector such a mosquitoes or a tick or simply  by stepping into contaminated soil or water. They are a disease of poverty affecting  disproportionately population with limited   access to resources in the healthcare and they  generally do not trigger long-lasting immunity.   So most importantly, for a lot of these disease  we do not have vaccines and sick people can be   treated which is great, but treatment does  not protect from brain infection. Therefore  

targeting also the environmental reservoir or  preventing exposures are two key approaches   for five these diseases. And the government might  talk to guys is to show how digital technologies   can help to address this problem. And I want to  show you that by using one of the most important   of theor tropical disease as a reference  example, I'm talking about histo disease,   rehabilitating parasitic disease of poverty.  They in four continents, 74 country in tropical,   tropical regions affecting more than  200 million people worldwide. The vast   majority in sub-Saharan Africa, the  vast majority school aged children. It is caused by parasites of the blood vessel  which is represented here in the lower left   panel. And this parasites as a complex lifecycle.  It required two host, the human host that you can  

see here and a limited set of fresh waters, nails  the obligate intermediate host of the parasite.   These nails are actually really tiny, they looks  harmless but they are hijacked by the parasite.   They transport them in a factory of infectious  stages called saria. They are released in the  

water for several weeks after infection. So  people get infected by stepping into parasites,   contaminated water. Now there is a drug to  treat infected people quantal, it works well,   it's cheap, but after removing the parasites  from people, it does not anything to target   the environmental reservoir of the disease.  Since they continue to shed infectious stages  

for weeks after drug treatment and like on  a treadmill, this lead to an endless cycles   of treatment and reinfection to interrupt this  cycle, the what the organization has recognized   that OSI can be controlled only by complementing  medical treatment with environmental intervention   and cause we always operated in condition  of limited resources is to, it is crucial   to identify and target snails, asna habitat  to fight that EAs to show how we can do that. I want to bring you in a quick journey to my field  size in Senegal here. So we are in Western Africa   in the lower Beijing of the Senegal River up  here at the border with ma, with ma Lithuania.   And since 2017 I've been fortunate to lead seven  field missions at Senegal. We've been conducting   survey of nearly 1500 school aged children. We put  a a lot of effort to try to monitor the infection  

dynamics in this hyper endemic area of in the  world. And this to this end, we've spent countless   hours collecting nails from the river in the lake  where people congregate for fishing, for bathing   and other daily tasks. We have collected actually  nearly 50,000 nails and Metis three recorded data   about all of them from the sub special location  where they came from, what type of vegetation   they were founded to precise miser of their shell  and of course which parasites live inside them. And this is what we found first  we found very patchy distribution,   very ephemeral with snails aggregated in  clustered ear and they are kind of difficult   to find. But we also saw that there was a strong  association of snails with certain type of aquatic   vegetation like phylo very easily to identify  visually like for instance from this picture.  

We also actually ended up finding some  unexpected mismatch in some villages   that were no near shore's nails but high  infection levels in people, which was   something really odd because we know that the  life cycle requires nail and parasite. And so   we basically realized that the regular sample  scheme was failing to capture this nail where   they are. We needed to change perspective.  And in order to do that in January, 2017,   we bought a simple, cheap recreational drone, A  D J ui, phantom four with a simple air GB camera. We started to take shot of our field site and  what we found a lot of more suitable habitat   for this nail that we could see from the shore.  And specifically we even realized that the area   that we could safely sample by stepping water  up to one meter, one meter and a half at most   was just the marginal areas of these large  distribution of floating vegetation. And   this vegetation could be easily identified  from our high definition subter imagery,   drone imagery and classified in different  habitat of different quality forces. So we  

decided basically to move from these occasional  observation to a systematic flight plan of the   series of overlapping drone photos that we took  a every water access point about 50 to 80 meters   above the ground level. I will spare you with  some of the technicalities and the details,   but we took hundreds of images weed  to identify our much vegetation. There was different areas and this basically what  we found. So the extent of aquatic vegetation  

matters, the more aquatic vegetation, the more  snail and snail cluster we found. We confirm   the fact that these days are extremely patchy  and theoral. They fluctuate in time lot and in   fact it's much more efficient to understand  that the potential risk of transmission to   track treatable habitat, which we can spot from  the drone imagery that samplings nails as we   did side from side. And the relationship  between alu vegetation and prevalence of   infection is robust enough to understand  just by looking basically a drone imagery,   whether the site is a high transmission site or  a low transmission site. So this has been great   cause then we know what we can do in a single  village. But the problem as Manisha was saying  

before is what happened when we scale up, when  we do map the large special scale, you know,   we have to understand where do we want to  prioritize intervention tailor, what do we   want to basically deploy and distribute product  quantum force do environmental intervention. And this scale we're talking about something  like 20,000 square kilometer. There is no way   we can fly a drone over such a big area.  And so we rely and we moved to satellite   imagery. The quality is not as good as  drone imagery but as improved dramatically   in the last five years. And this allowed us to  detect natives avatar from satellite imagery.   There is no way we could parse manually such a  big area and we need to automatize the process and   thus we rely basically on some of the techniques  of machine learning, what we can see by as we can   train and machine learning algorithm to do for us.  And so we basically use the satellite imagery and  

the drone imagery to ground truth in the satellite  imagery and ultimately we apply a specific type   of convolutional layer of network that has  become widely used in biomedical application. And we use this on a really limited number of  imagery, surprisingly limited. And the training   is a training dataset and the the results were  excellent. So despite the limited dataset,   we are able to identify 95% of the floating  vegetation, the treatable habitat for ISNA   by using this type of algorithm. So what we see  here is that we can integrate the field ecology   with drone and satellite imagery to identify  potential hotspot of transmission. We can use this  

information to find tune, imp, epidemiological  mathematical model of disease dynamics where   people movement for instance is calibrated  by using mobile phone data and we can explore   different scenarios and have a sort of a civilian  system where every year we can span this area,   identify where the hotspot are and which are  the best intervention to control the disease.   To conclude, I want to stress and to emphasize how  likewise precision medicine there is an untapped   potential precision mapping of transmission risk  for environmentally transmitted disease. And we   need to integrate the old fisher disease  ecology, which is the foundation of what   we do with the superpower of the digital  technology. And with there I stop my talk,  

I want to thank you all for your attention and  thank also the broad set of collaborators that   provided an invaluable support, crucial support  to develop this research. Thank you very much. So thank you very much Giulio. And  so we've now had two excellent setup   presentations and we're getting ready  to now move to our keynote speaker, Clinicians in remote hospitals, counselors for  crisis, L G b, youth smallholder cotton farmers,   online fact checkers. These aren't necessarily  the people that you, you think are working at   the forefront of cutting edge AI applications, but  why not? Why shouldn't those working on the front   lines of our most challenging societal problems  be using technologies', most advanced tools to   enable their work? And we have an opportunity, can  you please put the slides back to the That's fine,   thank you. Have an opportunity to ensure that  really everyone can benefit from ai, advanced data   analytics and digital technologies. and@google.org  we work to ensure that that's possible.   I'm really pleased to be here today  to share more of that work with you.  

Next slide. Google's is google.org is  Google's philanthropic arm. We work   to bring the best of Google's resources to  address societal challenges. We do that in   the form of funding, so every year directing  1% of Google's profits to support nonprofits. We also think about how we can use our technical  expertise mostly in the form of volunteers who   spend anywhere from an hour to a full-time  secondment with organizations that we're   partnering with, bringing their expertise  alongside that of frontline organizations   who are working on these problems day in, day  out. And then of course thinking about how we  

can bring advanced technology, connecting with  expertise and really bringing the, the tools and   we use to power much of our work to organizations  that we think have the opportunity to change the   game for some society's biggest challenges. Today  I'd like to share a couple of examples of those   organizations and then I wanna talk also about how  we're working to enable more organizations around   the world to use machine learning in particular  and where we see future opportunities. Next slide. The first example that I wanna share is out  of Uganda. So air pollution, as many may know,  

contributes to a significant number of deaths  every year. Some estimates up to 4 million   deaths in Africa. And recent studies even show  that that impact is greater than malnutrition   water and Sanitation sort of problems that I  think we, we often hear more frequently about   in continuous air pollution. Monitoring is  a critical tool for combating and addressing  

that problem, but sensors are quite expensive,  typical sensors, and they have to be installed   across many specific point locations in most of  the models that we might think of being used in   some higher income countries. The researchers  that we are working with and proud to support   at Macri University have developed air quo,  which is a project where they have built a   low-cost sensor. You can see one here on the  page that is uses solar power and a lot of   local components and they're deploying them  across the city, not just in point locations,   although importantly they are, but also on  motorcycle taxis, on boat bodas that drive   throughout the city of Kampala and are able  to provide an additional layer of data without   necessarily having to have the quantity of, of  sensors that you might need in a point system. They're then applying AI and machine learning  on top of that data to both interpolate between   the data points that they might not be able to,  to get and also to do forecasts and predictions   that citizens can use through an app that's  now available throughout the country and that   governments are starting to use to inform policy  and also organizations like schools and others   being able to take action on days perhaps when  the air quality is bad, not having students play   outside. We think that this tool and, and the  project that they're working on has tremendous   potential to change the game for air pollution  throughout the continent and perhaps around the   world. And they're looking right now to think  about how already it can be scaled to Nairobi.   We've supported them obviously  with funding and, and also AI   expertise. But I think that one of  the most interesting pieces of this  

project is actually that the implementation  itself is really where the innovation lies. The low cost sensors, the installation of those  sensors on taxis, and then they're also working   to engage the community around particularly  high risk spots of the, of the city. And also   in places where the data has not really shown to  be predictive, they're actually asking folks to,   to provide crowdsource information. So on days  when they might, you know, see a fire burning or  

notice a lot of pollution adding that data in.  And I think that community engagement approach   is something that is really kind of a hallmark  for a lot of the work that we're working through   and something that we wouldn't necessarily, I  think have seen or they had not even necessarily   proposed in the initial part of this project.  And so innovation in that implementation part   of the way that we deploy digital technology  in addition to of course the cutting edge tools   like AI machine learning that they're applying  is really advanced the, the opportunity here.   Next slide. And on this page you can see that  app that I was mentioning, you can see the points  

there zooming in on them and then also an ability  to see a forward looking forecast and prediction. Next slide. The next example that I'd like  to share is, is focused on the problem of   antimicrobial and antibacterial resistance.  Obviously bacteria, viruses are always changing  

and some have adapted to ensure that medical  treatments that we typically use are not   necessarily able to be used. And some reports  suggest that the number of deaths attributal   to Inca microbial resistance by 2050 could be  as high as 10 million if we don't take action.   And these are particularly challenging cases in in  war torn countries where there are common injuries   and bacterial infection can actually cause  significant loss of life. This is obviously a lot   of the countries where M s f al frontier doctors  about borders does their work. And this project   is a really interesting origin story. So the, the  challenge of dealing with antimicrobial resistance  

requires prescribing the right antibiotic for  the bacterial infection that's being presented. And to get that information, we typically  use expensive lab equipment and very   highly trained microbiologists to interpret  essentially the Petri dish that you see here,   a susceptibility test that shows how the  vi the bacteria is reacting to various,   various antibiotics. Those tests are hard to  interpret as you can see here. Then perhaps we   hard for, for someone on an untrained eye to to  use this tool or to to use this petri dish and   sort of understand what to do. And it's very  hard to get both expensive lab equipment and  

trained microbiologists especially into perhaps  a country like Yemen. And that was the case for   M S F A couple of years ago. One of their lead  microbiologists was, had been recently there   and she was interpreting these images being sent  to her by an email overnight in Paris and their   home office. She connected with a friend who was  a computer scientist and said, you know, we've   seen all these advances in computer vision, isn't  there a better way we might be able to do this? And this project was born, they have now  developed a free mobile app that you're   seeing the demo here on the, on the left. It  supports those sort of non-trained technicians   to measure and interpret these important tests  so that clinicians can then prescribe the right   antibiotics. It works by taking a picture of the  Petri dish and then measuring and, and looking  

at those inhibition zones. So that's supported  by computer vision. There are then interpretive   questions that help guide the clinician to  interpret the results, resizing the area,   for example, around this particular dot, and  allows them importantly to layer in some of   what they may have seen in terms of patterns or  ask questions as they're interpreting. And then   they're able to see a recommendation that they can  then share with the clinician and also send in for   peer review as they've been trying to make sure  this tool continues to, you know, deliver results. As noted here on this page, the the way that the  studies initially that they've done in in this   tool is just in trials at the moment, I should  say, is showing a really high level of agreement   against both the standard auto automated tools and  human interpreted results, which I think is really   telling and, and shows the opportunity that we  have to layer in these tools to clinical settings.   You know, as Manisha was mentioning, where there  might not be some of the technical expertise   available on the ground. And it's also a really  great example of a human in the loop system. So  

we talk a lot about AI and, and concern about how  it might interpret results that may or may not be   accurate. And here you see AI augmenting really  the most error prone and challenging portion of   the task, this kind of initial assessment, but  ensuring there's multiple points throughout this   system at which people can layer in their own  interpretation and do the necessary checks. Obviously when we think about  machine learning in particular,   prediction is sort of the ultimate goal for many,  for many cases. And if we can get information to   help us act in advance of adverse events,  that I think is something that we've all   been striving for for many years in the social  sector. I wanna share this final quick example,  

which is work that's done actually not necessarily  only by google.org and our third party partners,   but also through work that our crisis response  team has been leading internally. So this is an   effort to really get better flood forecasting  predictions, floods are one of the most   devastating economic and harmful events in the  world, and they are notoriously challenging to   predict, especially during heavy monsoon seasons  in places like South Asia. We've been able to   combine advanced map mapping capabilities using  some of our mapping tools and AI to really predict   a path of a flood and generate detailed maps  that then can alert through our online systems.

And those alerts are also provided to government  who often alert through their own systems.   But I think this is a really good example  of combining digital technology with   offline solutions because as you  can imagine, many of those who are   at greatest risk for impact from floods are  not necessarily the folks who are getting a,   a smartphone alert with a map showing them  where the flood susceptibility is highest.   And so we've partnered with the International  Red Cross in in India where this has started and   expanding now into Bangladesh, into other  countries to really build out a community   alerting system alongside this. So how do we take  the information that these models are outputting,  

which is critical information and try to get it  into as many hands as possible. And obviously,   you know, digital tools in all forms, you know,  s m s and WhatsApp groups being formed, but also   just word of mouth and using a lot of the, the  ways in which community information flows already. Some communities use flags to raise down the  river to say, hey, a flood might be coming.   Some communities use literally musical  instruments to try to, to try to get   more information out. And so partnering with  this network of people alongside the technology   has been a really critical input. And as noted  here on this page, we are seeing through some  

studies done with, with folks at Yale, a two  day improvement in rural alert De dissemination   during the pilot that we ran alongside our flood  alerting. And that two days, as you can imagine,   could be the difference between loss of a  home, loss of livelihood or, and you know,   worst case loss of life. So we're really excited  to see how this might be able to scale to many   more places, working closely with governments  and working closely with our N G O partners.  

So I wanna zoom out and talk a little bit about  how do we ensure that more and more problems   might benefit from machine learning and artificial  intelligence in the advances that we're seeing. And I think Lacuna Fund is a great example  of that. So we, with Rockefeller actually   some of manisha's colleagues alongside the  German and Canadian governments have started   a collaborative effort to fund labeled data sets  that address problems that are underrepresented in   currently available data. There's three areas  that that work has started agriculture. So   thinking about things like crop disease or soil  moisture, not the, the data that we're using,   you know, here in the US where I'm standing, but  importantly represented across Sub-Saharan Africa.  

Language is another area. How do we ensure  greater representation of language? And then   we just announced actually earlier this  week, an ex first round of expression of   interest for health data sets. So really  looking at how more representative data,   both in the US and in low and middle income  countries could drive more e equitable outcomes.   And we realized that this is a critical  input and I think really call on everyone   to be thinking about how we can continue to, to  close this gap and ensure that we have the right   data that we need to drive equitable outcomes  and to create new innovation across the board. Finally, I'll just leave with these three areas  we've recently published with ID insight as of an   article which you can find in S S I R that really  points out, you know, was a framework for how do   we think about more areas where AI and machine  learning may be most impactful. And there are   three that I'll leave you with as I think we can  all continue to point our attention to do more. So  

the first is, of course, point of care diagnostic  tools, really thinking about how with especially   better data, we can see that differential impact  as we just talked through in the M S F example,   communication tools that support marginalized  communities that really lean into language.   So thinking about human translation, translators  being as augmented or with those new languages as   we mentioned coming online, how we can ensure that  people are getting the information that they need. And the final area that I think we feel  there's a lot of opportunity is agriculture   yield prediction, obviously with a significant  portion of the world relying on agriculture for   their livelihood, for their health, thinking  about where we can see more tools that actually   increase productivity, forecast yield, help inform  harvest, help inform decisions that are being   made. I hope this has been a, a tour through  where we see some of the, the fascinating work   that's happening with AI machine learning. Again,  really in in partnership hand in hand with those  

who have worked on these issues for decades. And  we'll work on them for many to come. Thank you. Thank you very much, Brigitte. It's really good  to hear about Google's work in this area and,   and, and how this connects with broader public  health efforts. We've had a number of interesting   questions come in and I wanna start with a  first question to Manisha. I think that we   have all watched with dread as the surge  of COVID-19 ran through India recently   contributing to as much as a, to a couple  of million deaths. So first I'd like to  

ask Manisha her thoughts on how digital  technologies might be used to help manage   the crisis. And after we hear from Manisha,  perhaps Brigitte could also talk about the   perspective that Google might take in terms of  where contribution might be. So first Manisha, Thank you Steve. Yeah, I mean it was a crisis and  then, and you know, unfortunately a tragic one,   one of the challenges with COVID-19 at large India  and the world broadly is that we were flying blind   and there was limited information in terms  of transmission, community transmission,   emerging variants and availability of supplies and  critical essential supplies. So I think there were   failures, information failures across multiple  fronts. I think that the key lesson here is two,   two things. You know, there's certain things that  worked in certain things that didn't. India's  

crisis is enormous because its scale is enormous.  You have 1.3 or close to 1.3 billion people packed   in highly dense geographies. So I think elements  like social distancing and some of the, the, the,   so public health in interventions that we have  Western in countries are harder to implement.  

You know, every event, every step out of the  house, it's a, it's a potential super spreader. Having said that, there were certain  things that did work. The AR app,   which was rolled out as, as a  contact tracing app was, is fairly   well implemented. I mean, I I'm saying fairly  well with the, the caveats alo in terms of what   that actually means, but in terms of the number  of people that went on it were Monitoring, it   was much more than expected. So there were certain  things that worked. I think we need to acknowledge   the role that digital technology can play and  acknowledge the fears and behaviors that surround   those technologies. So I'll, I'll sort of break it  out into three things. First is, you know, contact  

tracing individual applications that monitor  people their behaviors and their surroundings   that has challenges in terms of uptake largely.  So I think it's not just the r o GTO app,   I think even the other efforts that Facebook and  Google put in place, you know, had challenges. The second is scientific information. And this is  around, and this is the work that foundation has  

really is leaning into going forward is what  is, how does one implement and support a cost   effective scale surveillance platform around,  you know, genomic sequencing. So again, driving   precision in terms of what are the pathogens that  are circulating, what are the variants of concern   and what is the bioinformatics capabilities  that we want to build in the country to actually   pulled together an early warning system. So  again, what have we learned, what are we doing?   And again, you know, we were late, India was  late in terms of identifying the Delta variant,   which was circulating months before, not because  we, the, the infrastructure wasn't there,   but the Coordination and the ability  to pool, aggregate and analyze data was   limited in the early days. So again, what  is the system that we can put in there? So you've really highlighted that  there were true information gaps   Yes. That really contributed to our  suboptimal response to this crisis That is across the board agnostic of country.

And it's interesting to me because this  is actually some of the same things that   Giulio was talking about saying, I don't have the  information I need to recognize where the snails   are. So maybe just Brigitte, could you talk a  little bit about how google.org sees this? Are   you completely agnostic on where you're working  or do you see like with this big health crisis   of COVID-19 and in India that  there are particular areas where   AI and Google strengths really  can lean in and contribute? Yeah, we've been doing a lot of COVID related  work over the past year. One, one piece of that,   which unfortunately to Manisha's point  I think didn't quite work for India yet,   but is hopefully in place for the future  is, is work that we've done to support a   global specific case line. It's called  line list data in epidemiology world   that provides case information across 140  countries in those early days, you know,   to to some of what Manisha was sharing, getting  that kind of information to say, hey, we're   starting to see something that looks a little bit  different and we actually don't mean aggregate   data. We need to know specifically what were the  cases, where were they, that dataset is being   run through a university consortium now under the  umbrella of global health, also their U R L with   India in particular though i, I just a hundred  percent agree with so much of what Manisha said. And I would say one, two other things to  add. One is that I think we also saw the  

need for information. At the end of the day,  you sort of end up with frontline workers   sort of stuck holding this challenge themselves  and, and very poorly equipped to do so because we   have been flying blind, having the right  protocols, having the right information,   even just having the right information  to share with families who in, in many   cases we're actually providing that  care in their homes for their loved   ones. It's really challenging. So I think that  information gap also extended down to that level   and, and just points again to the need to  strengthen the infrastructure of those systems   in times of not crisis. And we've also, we've been  working with an N G O called Armand who supports a   lot of government community health workers. They  luckily had already stood up a, a mobile academy,   which they had been running for, you know, the  last year or so in partnership with government,   which provided a vehicle to get, you know,  just creating a new set of content through   the same channels. If you have to also  be establishing all of those channels at   the same time that you're creating content,  it's just impossible. So I think, you know,  

and that that even was probably inadequate  for where it would've ideally been, but I   think again to this point of how do we invest  in the channels infrastructure that we need now,   how might we use this as a motivator, right? To  do some of that work is, is just been critical And this, this is this no notion of where do we  invest. We have some capacity that we think can   help out, but there's been a lot of concern raised  by questioners here too about how can we guarantee   that we can actually get this into the hands of  people who need it. That if we think of the most   marginalized communities, the type of communities  where Giulio for example, is working with and   what RA and, and, and, and where the Rockefeller  Foundation is, is trying to reach out. There are   severe shortages of personnel, there's shortages  of IT equipment and, and how can we adapt these,   these potential tools in the setting  of such resource scarcity. So maybe,   maybe I'll ask Giulio to answer first, but I would  be really interested in others' responses as well. Yeah, absolutely Steve, thanks a lot. I'll  give you of course an academic perspective.  

So thinking about what is in the real of  capacity of the work that I can do and,   and I know that Manisha and JIT have, you know,  a different scale of operation in action. From   my perspective co-design, the research that  we're doing with the local community and the   specifically the academic communities there  that are operating locally is one of the key   element. And, and so is there involvement  in our specific project we're working with   two no-profit organization as Spark for Lacent  has been operating in the RF for 25 years with   the university Teton Berger in San Luisi and the  African Institute for mathematical biosciences.  

And so we working with them, we can be sure that  part of the knowledge that we so passionately   want to develop will be co-developed with the  new generation of leaders of these countries. Thanks to Sparko Lasante specifically our  no-profit who's working there are, we are   also trying to work elbow to elbow with the local  community and will chiefs and with the people who   are there on the ground to be sure that they  need are met. And so this is basically our,   we operating, we do also invest in, we will do  in the future in training launching undergraduate   courses, for instance in drone mapping,  which extend from our specific goal of   controlling the greater tropical diseases, but  expand also to agriculture mapping because this   is an area where there is potential for jobs in  the future and we want to share these basically   technology with them. Last comment is that when  we brought drones the first time in Senegal,   we were concerned of bringing something  that was potentially scary or you know,   sounds that completely out of the  context, but it's been on the contrary,   a fantastic way to engage with all the  villagers. I think everybody, you know,  

as included looking where we live from the bird  eye view, you know, is a, you know, it's a new   experience and so we've been able to interact with  the people there in a way we didn't even expect. Thanks Giulio, Manisha, I'm interested in  your perspective both in terms of how we   reach out and also is this a role where NGOs work  really well, but does that also give us limits in   terms of long-term engagement by governments  and use of these technologies and all? And,   and Rockefeller thinks a lot about skills,  so I'm interested in your perspective. Yeah, no, I think that's a really good question.  And you know, for a lot of these solutions to   work and deliver on the impact that they promise  they need to operate at scale and they need to be   institutionalized. So it's, it's, I would argue  it's not an either or. It is an, and there's a   role that NGOs and social enterprises can play,  especially in the leading edge of demonstrating   what the innovation or the technology is, how is  it applied and really unpacking the implementation   science behind the application of that technology  to solve a specific set of problems or problems.  

The role of the government is in terms of the  enabling environment and institutionalizing   those that implemented that, that that model of  implementation. You know, we talk about access   to technologies in rural and remote areas.  One of the fundamental areas we leaned in a   couple of years ago is access to frontline mobile  tools in remote and, and, and rural rural areas.

And for that to function, we need a couple  of things. We need network, right? And that   is the role of the private sector and government  understanding how does one create the Economics   for network providers to engage in say, rural  Sierra Leone, how do we work with a ranch, how do   we work with mtn, et cetera. The second is cost of  data. I mean, we don't think about this in the us   but fundamentally, yeah, who's paying for data in  these countries and how much data are you using?   So that fundamentally constrains what you can  do and who you can't do. So there is a multiple   strand me, you know, approach here where we are  trying to use relevant technology to the context   as it stands today, but also influence and and  advocate for the enabling environment to move in a   direction that allows for smart technology, allows  for better bandwidth that could deliver on those,   those analytics both at the, the facility level as  well as the handheld level with community health   workers. So I mean the, the cost model changes  substantially and if you don't address the cost  

models, then scale becomes a fundamental  constraints. So you'll have a set of very   well-designed, beautiful, highly impact small  projects that basically end when the financing   ends. And we have probably, you know, millions of  those over the last several years in that space. Yeah, I mean, Manisha you make a  really good point and something   that we often don't bring to the forefront  that with focused effort and some funding,   non-government actors can come in and collaborate  and make really remarkable interventions,   but taking those interventions to scale so that  they can impact the trajectory of a pandemic   is, is another task and, and is a difficult  task. And I think part of the reason why   the public health community is so attracted to  Google and to the tech companies is that they do   have a funding model that allows them to put some  money in to generate data, yet at the same time,   obviously they're a profit making company  and trying to figure out how this all plays   in ways that can generate data at scale  that can matter for public health. And so   Brigitte would just welcome your thoughts on, on  how, on what you see as the contributions there,   the limits, what steps we might take to make  this a more productive and enduring partnership.

Yeah, I think that scale is often something  that we, we talk about as though it's like   not something we really understand how to do  when actually we have seen many things go to   scale and picking them apart, you know,  a little bit to say how did that work   and getting the right ingredients from the  start with projects would go a long way.   I think we, you know, gov we know government  isn't in needs to be an involved actor. We've   heard it throughout the examples that everyone's  discussed here today. And yet too often we do   something in a corner and then ask government to  look at it without having had them there in the   beginning. It's not a pilot if no one's watching,  right? We're not actually demonstrating anything   to anyone. And I think that is actually  something that tech companies have learned,   right? We don't build products in isolation,  we build products for users, we build them,   we know quickly check them and  sort of run that process forward.

And I do think there's, there's value in  that thinking and that mindset that that,   that is under, under invested in I think sometimes  in the projects that we collectively support. And   I think a lot of that ingenuity and is, and  clarity is actually there in the organizations   that we're working with. They're just not  necessarily, they're just not being offered   the opportunity to put the peel puzzle pieces  together in the best way that they would want to   because they're hemmed in by funding  constraints. They're hemmed in by, you know,  

this particular Ministry or or agency that,  you know, says that they can't do the thing or,   or no one's just willing to try. And I  think those are things that actually we can,   we can do a little bit better job of. I wanna on  the, on the role of, of kind of Google and tech   and on the data front in particular, you know,  I think the pandemic has been an interesting   opportunity for us to try some new approaches  to releasing more data than we have previously. Obviously we of course take privacy  very seriously as a company and so   it's quite challenging to think about  how we can release information that   would maybe be beneficial while  still of course preserving privacy   in the way that we want to and need to. We did  release this year community mobility reports  

for the first time offering aggregate data in a  fairly small and a sort of county level equivalent   geographic area that showed sort of the, the  movement based on Android phone data that we   were able to connect for those who had turned  location sharing on. And we found that those have   been quite helpful for a lot of epidemiologists  in modeling. So we're asking ourselves questions,   you know, about how we continue to do that  kind of work going forward. Similarly symptoms,   search data sets, so looking at search  and saying what are people searching for? How might that be correlated with, you know,  upcoming outbreaks and another data set that's   now out there that is available to researchers.  And I think these are actually helpful having that  

forcing function of, hey, this is a big global  emergency, how might we be able to contribute?   Then brings back the question of what does that  look like in an ongoing way. But I do also think   sometimes there's a challenge of, of like tech  companies and the data will just solve it all   when actually so much of the other components are,  are critical as, as everyone's been speaking to,   right? That that will be one important input as  will other data sets out there. But at the end   of the day, if governments and other actors  aren't there to act and we haven't sort of   brought them into that process, then we likely  won't achieve the outcome that we wanna have. I think thanks a lot for the comment.  I think my, I think my favorite   piece of, of what you were saying there  was that a pilot is not a pilot unless   somebody is watching. And this whole notion  that as enamored as we are with big tech and ai,  

we still need to work with the broader political  and social ecology of public health if we are   going to work towards solution, and  this is, this is a potential tool,   but solving some of these very difficult  issues like revenue model and Coordination and   figuring out how we're gonna manage privacy are  actually really tough nuts to crack. And in fact,   when we think about precision health, usually  what is thought about with precision health   is more precision medicine. That is when you  have one patient coming in, how can you get all   of the data and all of the genetics and all of  the exposure to optimize for that one patient. So, so we've really been optimizing this notion  around precision medicine and using precision   public health has been, I would say, a less  common framing. Cer

2023-10-15

Show video