2020 Back to the New Workplace post COVID 19 - Ramesh Raskar
Today. We have the second half of our back to the new workplace post. Koba 19 webinar, with. Professor, ramesh Hrothgar, and three, of our MIT startups, in. The. First session last week professor. Kokum and dr., dairy talked, about the bigger picture of, going back to work around. Social and policy considerations. As well, as employee. Experience. We. Were mentioning that Co result that the technology, solution, is a key part for what you and your company are focused, on for, returning to work. The. Good news is today we'll be delving much more into that practical, technology, solutions, aspect, with. More real-life back, to work scenarios. Let. Me share today's agenda with, you. First. We're hear from Professor, Roscoe, speaking. ANCOVA safe paths and paths. Check the, digital compact Racing's non profit is spun out from MIT. There. Were here from the three of our startups. Palbok. Tulip. And human eyes who, focus on different aspects, of that returning, to work solution. Then. We're here have a panel, with Professor Roscoe the startups, and Jeff. Dell more of FM, Global representing. The corporate voice that. Panel will be moderated by, Michael, shriek from MIT, Sloan. Let's. Get started. I've. Been reducing, professor Hrothgar who is speaking first. At. The MIT Media Lab professor. Roscoe's research focus, is on, AI imaging. For, health and sustainability. These. Interfaces, spend researching, physical. Digital. And global. Domains. He. Has received a Lemelson, Award for, his groundbreaking inventions. In a number of other distinguished. There were awards. He's. Worked on advising, special, research projects, at Google Facebook. And, at. Apple, so. He knows the ins and outs of these tech giants as well. He. Has also co-founded. Multiple. MIT startup yesterday. You testify, and a congressional, hearing on the same topic. Take. It away professor Bhaskar, hello. Everyone what, a pleasure to be here and thank you Rebecca last. Time when we spoke on this platform which, is getting started, and. As, Rebecca said we have spun out the. MIT. Safe paths work into a non-profit foundation, path, check foundation, so I'm going to share some slides and and. Walk you through some. Of these some of these ideas.
But. I'm going to focus mainly on the learnings so I'm supported. By Chris Christ who, you. Know is a doctor, and a software entrepreneur and also. Part of the path check foundation, so let's talk about reopening, businesses. And. Patchak foundation, is building. Solutions not, only for public health but, also for this new start for countries, States and employers, that are open source privacy. First and interoperable. The, first versions were based on GPS the. New versions, are based on google apple bluetooth solutions, and the, versions of coming we have other case. Management, and public health solutions as well, just. A brief summary of what, we discussed last time which is on the Left we have the app that's. Going to the employers, or citizens, for students, exposure. Notification. Personal, guidance and, on the right we, have the safe places dashboard. That's, used for public health and businesses, so it can improve the tracing conversation, your, contact tracing dashboards. Heat, maps and so on and then. Our code base which is the largest nonprofit. Open. Source code base for. Cobie 19. Supports. The whole range of api's GPS, Bluetooth Wi-Fi. QR. Codes and everything, else and the challenge really is how do you do this in, a privacy-preserving way, so, we have three arms the technology, arm the think tank and. Implementations. In. Technology, we are doing the app the web tool the algorithms, SDK, everything, is open source think, tank they have interoperable. Standards all. To work with stakeholders, landscape. Analysis, so. You know check our website and implementation. With Public Health and employers and Laura, corporate relations. So. We, are also delighted. With, that we have a fantastic alliance, in fact, next it will be making an announcement with, some big, fortune 500 companies, who. Have joined us in in various ways there are tech players or implementation, players or, contributing, talent or using, our our sources, so on so, keep a watch on that for for next week. Some, of you probably had questions about the Google Apple, proximity, API and the, way it works is if. Two people Alice and Bob and let's say Bob is going to get infected at some, point when. Alice and Bob meet maybe they bug did not know that he's infected, at some, point Bob gets infected and he uploads. The keys the. Bluetooth keys that they have exchanged, over a broadcast, channel and then. And then, Bob uploads them to the server and at. Some point alice downloads, on the skis to see all, the keys that she has heard over the, Bluetooth broadcast, is the same key that she received from the server and if there's a match then, she knows she, was exposed, by Bob so. It's a pretty clean simple, technology. And we, do the same thing for GPS we do the same thing for Wi-Fi and so on so, only the infected, person has to upload some information, the, 99.99%. Of, the people who are not infected never. Have to upload, anything at all and that sub privacy, is maintained and if they do upload is all encrypted. So. When you think about the. Experience, flow, from individual, from, the moment they get exposure notification. To, the moment they might be infected is, infected. And have to upload the information you, know it's an emotional, rollercoaster for, most people and so. Our team is made up of not just technologists, but epidemiologist. You, know behavioral, scientist, legal. Privacy, and so on and we all work together to figure out what the solution should be and we put it out in the public domain but. We are very fortunate with the few million dollars of philanthropic funding, so, we have hard core software, engineers, building. Extremely, high quality code, as well so. From exposure notification, you, come to triaging. Symptom. Checking, quarantine. Or not test, or not what. 14 days after you have come out of those those, conditions are not so we look at this holistic. Picture and again as a software company anybody, can build this the, challenge is how do they do this in a privacy-preserving, way, so.
All The machine learning all, the encryption, all the decentralization, has to work without, the name or email, or employee. ID or a national, ID anything's, identifiable. Should, not be used for this process and that's why some of the MIT magic, comes. In so. Just want to make sure that when we mean by privacy, it's. Not about the individual, privacy but, is also about consent, it's also about regulations, often you, cannot share, the data because of HIPAA regulations sometimes. It's about trade secrets, you may be an employer and you, cannot share your data with insurance. Player or the pharma players or, its national, security you might have people in Singapore and Asia and Europe and is there travel across boundaries. You simply cannot use the information that's. Kind, of held by. Those countries in, the systems and again what's incentive for an employee or citizen, to participate, they, can just switch it off so these are all very complex issues from, human behavior. And, your incentives, that, we that we work on as well so. Let me share with you a few learnings. On on, the, last three months we have spent this implementing. With many fortune 500 players working with many governments, and we, all have contracts, for several US states and territories and, nations. So. I'll. Be first one to admit that it's a humbling experience because we, thought we can build something and it'll work it's, a very complex situation and I, have built many startups, from my own bare hands, which teams but there's nothing, that, matches, the scale and importance, of, what we are doing and so we'll be very honest of what works and what doesn't doesn't, work so. When it comes to businesses, what they're noticed is that, businesses. Do not want to access, the, health care data of their employees and even location. Data is considered. Very, sensitive, because it's personally, identifiable, and. Without, consent is out of question so that's all kinda considered invested at the same time businesses, do want some control and some command. And control dashboard, like concept. And they also want notions of Passport like who has which, symptoms but, they have tested an art exposures, and, what you are noticed is that everything. That people are doing right now is is. Israeli scattershot like symptom, checking is very.
Annoying Because people do it for three days and the fourth day they just click random buttons to say I'm fine I'm just gonna come into work if. They have already done the test you, know you have to close the loop with testing, facility, so. This tension, between centralization. And decentralization of. Employee. Data is a, major issue and what we have been able to convince, many. Employees employer, sorry is that, they should use decentralization, they should not worry about capturing, so much data about their. Employees that. Employees, start playing games and this stop using the apps they stop using you, know the pendants, and the you, know Bluetooth. Tags. Or UWB tags that are being sold on the market for 100 or 200 dollars it's, all technology, it doesn't work for you know when the, human ingenuity the human laziness starts. You know getting around it there's. Also the confusion about should we use Bluetooth R or GPS because Bluetooth. The Google actual exposure notification, can only be used by states and nations and not, by businesses, so what do you do so. We're also working through those issues and. Then finally I would say this is this is the most important, employers. Simply, do not an employer, X doesn't. Want to give an app to, their employees saying here's, an app by. X our company's so download it and use it because. Employees, exhibition, large companies will, not allow. The, infringement, on their, personal, freedom and, we have noticed this again and again so the, relationships, we have built with a lot of fortune 500 companies and others is that they would like us they would like path check to. Be almost like the lead certification or, Fairtrade kind of organization, it's, open-source non-profit. Auditable, so, we provide the front end we, take the mantle of privacy, and personal freedom, and security, and, we, provide that solution to, the employees, and the, employers, then. Talk to our app to create the dashboards, and when, we upload any micro. Or macro. Aggregation. Of data from this app we. Take care of all the privacy constrain, so, that even, the most malicious employer, will, not be able to reconstruct anything about the employees, at, the same time the, workers will feel safe and the, employee can create dashboards for. All the policies, what's working and what's not working so it's a good symbiotic, relationship, where, path check takes care of the pesky issues of privacy. Personal. Freedom and and. Data protection and the, employers get the benefit of of having - force and. So you know here are some slides I want talking to them about them in detail but. A classic business experience, would have the same you know employee shows up there, to do tests on the site what happens you, know if there's a 24, hours to 48 hours delay or somebody was infected, or not if. They're infected what do you do if they are not infected what. Does it mean to still. Allow them to practice social, distancing, staggering, schedules and so on so, some of the things we are doing ourselves in past check for, others we you, know they're partnering with other innovators, in the startups or other organizations, they, provide the other part of the stack and together we, make an offering, so that is privacy-preserving but. Still allows you to have this command, and control structure. Around it and. So in that sense in a safe pass is working with many stakeholders not, just governments, and public, health but, also with workers and labs and healthcare, and insurance and, schools. And. We are learning a lot and we're putting out our documents, you know on a regular basis of, what you're learning we just put out the document yesterday, on the. Relationship, between manual. Contact, tracing and digital. Contact tracing because, you, know most organizations, will have to set up some, small. Stuff it could be one person or, hundred people who, are just call center command control, and how, do they interface with the apps that are out there again. There is no literature on it there is no playbook on it so. We're putting out our experiences, and sharing it rest, of the world. And. Then on the MIT that's on the path check foundation, site and at, MIT we, continue to do cutting-edge research, in mush, relearning and encryption. And digital. Health so. These are some of the papers that are on our website how do you do is Wi-Fi colocation, cheapest.
Intersection. How. Multi-hop, analysis, on bluetooth how, do you preserve encryption, how do you do hashing, how, you do machine learning in presence, of. Encryption. So, please take a look we're, putting the papers out literally. The day after we feel, proud enough we put it out we don't wait for the. Review cycles and we submit them for review so you know we are getting out of the academic mode, and making, sure I focused, on, the goal here so when we talk about the kind of machine learning aspect. Of, Kove. 8:19, you know the classic challenge is, again. The debate between privacy, and utility, on the horizontal axis you know it's, you. Know it progressively, more privacy, so anonymization, when, we think about three, main. Pillars of of, privacy. Confidentiality. And, nanami, t and privacy. Right. And confidentiality. Simply, means some, people can see everything and in. Today's world that's not just not going to work anonymity. Helps to a little bit but if it's an overlapping data set anonymity, can also you. Know you can crack open anybody's, identity, so, anonymity. Is we think of basically no privacy and. But, that's great for machine learning so, the vertical axis is tentative. Statistics, on the data, can, you run any inference, on the data or, can you do full-fledged machine learning can you learn the AI and for. Machine learning if, we are you, know it's going to do questionnaires, from your employees, or you have sensors, or video or audio all, this data coming in and you, want to train some networks neural, networks to train that how. Are you going to deal with that information but it's not a smart phone or some, other edge devices and. So, so, what we realized is that anonymization, is for the best in terms of utility but, worse in terms of privacy, encryption. Is great for privacy but. Suffers, in machine learning so, two techniques, have emerged, we. Call it smashing, called federated. Learning and smashed. Learning, sorry it's federal learning and split learning and split. Learning is from our group at MIT and, it allows you to give its sufficient. Privacy at, the same time allows you to do full-fledged machine learning so, it's almost magical, that data stays on the phones or H devices, and we, can still create a global AI for. Any task you may have and this, technique is useful in others, as well they're, applying it to, getting back to work for. Kobe 19 and Canada, like I would have time to go into detail of the technology, but, the key idea here is instead, of sharing raw data from. Those edge devices like smartphones and cameras and so on we, just share the wisdom we, just share the features, of the, data as, opposed to sharing the raw data and that's how we maintain the. Privacy of the, data. So to conclude we. Think contact, tracing case, management, and, advising. Employees. And citizens and, students, on. Our premises is, very, critical to reopen businesses, and whether, it comes to machine learning or encryption. Or digital, health paths, check as a foundation, is committed to creating open. Source extremely. High tech you know research and development, and getting it out there and MIT. Safe paths is committed. To doing high quality research and. Contributing, the open source and working, with our partners at MIT thank you so. Thank you so much Rebecca and thank you Ramesh we you have done quite a bit of hard work with your team um, we. Heard you testified, yesterday on, a panel before the congressional, taskforce on exposure notification. And contact, tracing would. You care to share some, of the issues, and how they could be addressed yeah. Thank You Cheryl I think I think it was a wonderful experience. To. Do the congressional, testimony on, on, contact. Tracing and. I. Think the points that kept on coming about the one of the four panelists, yesterday, and the, question that kept on coming up was first of all wash with National Response which. Right now is in in disarray. But maybe the second point which is very important, is, this. Notion that government will likely. Likely. Consider. Nonprofit. And open, source entities, to. Do contact, tracing, and, that's. An interesting subtle. Point because if you're not in nonprofit or open source then. It's very easy or. The contact tracing app to, get misused. You. Know for, any large employer and if states and nations are deploying it then, you could imagine if 60 70, percent of your, state on the, smartphone is using this one app is extremely, powerful you, know even.
Some Of the most popular social. Apps you, know are, not don't have 60 70 percent penetration so, given that it's extremely powerful and they're, thinking about making it from nonprofit, and open source because. Even though using privacy-preserving techniques, they're, fancy things we can do on the server side using, metadata that. Could be problematic and the third point that keeps coming up is how, do we the manual, contact tracing with digital content reasons because, the government has spent tens of millions of dollars to. Create an army effectively. Call centers, and case management centers, but. They are becoming very inefficient. Just, today Boston Globe has an article where, about half the people were laid off from. The 1,700 people they had hired to do manual contact tracing because the number of cases in Massachusetts, are coming down so. Ramping, up and ramping down man, and contract racing is just not possible and, so the digital apps are going to play a very critical role so again kind of a coordinated response. Nonprofits. And open source doing apps and, merging. Manual and digital contact tracing. Thank. You so on another question related to that. There. Were so many organizations, developing, products for contact, tracing can, they can. You characterize the landscape, within it and where they might be sergey's with safe paths and paths check yeah, so so six-plus MIT, research, part shake is the foundation, that deploys the solution, you, know we are here to, not. Compete but collaborate, in. Fact we are here to enable everybody else so, in sum we already, have are building our own primary, app and server. And dashboard, in many US states and territories and Nations, so, we are delighted that will be the primary brand, in those states and countries at, the same time there's some other players who already are trying to build and the glue bit confused, because, the rnd required, is too complex or, as, as. I said earlier and Chris is online that's great as, we said earlier they don't use their own brand. For. You know because too risky for them to be out there. So either big from an rnd point of view or from a. Brand point of view path check is helping those, entities as well so whether you're state or a nation or large employer you, can use the path check brand Chris do you want to add something to that. Sure. So we've, been talking to a lot of businesses since the very beginning both business is looking for help and businesses, who'd like to work with us and partner with us and we actually have a formal, partnership program, so, you can join us at a basic level an ecosystem member, that just means you're part of the team you, can go join us at a gold or platinum level, which means that we're going to give you some support at varying levels and you, can also join us as a technology, partner, you're literally using. Our technology building it into yours or vice versa. Where, you're, actually helping us to build our technology, so we have several options for partnerships. And working with us. Great. Thank you Chris um. So, it's been proposed that government agencies might. Be requiring, code related, data from companies how, might save pads, dashboards. Help a company transfer, this information I I, think I have not heard that so far but. I would say again there's, confidentiality, anonymity. And privacy and privacy, means. Other. Than the individual, other, than the worker nobody. Else not, path check not the employer not the government nobody, gets to know anything. About this individual, any. Personally, identifiable data, not their location, you, know not their name none. Of that not even their IP address, and. Ideally. And. But. All the services are provided and, you think this seems kind of a, mismatch, how can you do that how can you provide services to somebody without, knowing anything about them and and. That's, actually that's already possible the technology, for that from a computer science point of view is already, available it should that we have to deploy it in the, right way and deployed, at scale because. We're talking about deploying it to billions the people now that, has never been done in a matter of weeks in matter of months and that's the role a path, check is playing like when we see that green padlock, in.
A Browser that makes us feel comfortable typing or credit card address so, the prowl browser doesn't know it you, know your Google Chrome doesn't know it even, the merchant, doesn't know your credit card number but you're still able to buy those things from. That notion is the same same philosophy here. That's. Great. I think, we'll have time for a couple more questions since. We're running a bit ahead yeah actually, I see I'm Joe hit Joey hasty asked, about tourism and. Travel and I just want to answer that that's a great question because many of our earliest adopters are, actually, tourist, locations. And they're very concerned about getting started as soon as possible obviously, for the summer season and so. We are working with them closely and we do have solutions for them what's really interesting is that sometimes it's challenging because you, have to host these solutions in various locations and you wouldn't always be able to cross the data over so we actually are working on interoperability, solutions, that could allow a traveler, from one country to, still be covered when they land in a new country. That's. Great. Another, question is are you working with any behavioral. Scientists, to test out best practices, for use in adoption. Definitely. That's something, that we did right in the beginning so. You, know then really that some of you might know from rationale, yours it's. Part of a team and. And. Julian's. Not a very, famous professor from Harvard also. Behavioral scientist it's, also on our team and. So there's some amazing people and not, just for employers also for nations, we're running pilots, because as you know behavior, is very culture. And regions, specific, and. So we're learning a lot of pilots and, hopefully we'll be able to share some results from. That we also have a separate program on incentives. Not. Just carrots were also stick so not just sticks but also carrots, in. How, people. Should use these solutions, hopefully. We'll have a lot more to share in the coming weeks but the, early, engagements. Only. Thing we know is this is not a trivial problem solve so if, your company is thinking about hey we have an IT department they, can build a solution we can use it for our employees, that's. Not how this is going to get rolled out there are a lot of complexities of privacy, it takes, and behavior. As I said that. Clear oh. Okay. And. I guess. Other. Questions. What. Can you suggest, as, the key, learnings. Of what. You've been doing over the last month, how things have pivoted and changed, that we should know about. Yeah, I think I think you know we, started, with GPS, because. The, new Bluetooth is very challenging when, I, took a sabbatical leave from MIT I was at Facebook for two years. 2016. And actually, built a very large team to, do bluetooth based you. Know proximity, analysis, and, we realize this is actually you, know very very pesky, problem so, we talked to Apple and Google recently, and they said you, know there's. Not much we can do and then, a few weeks later fortunately, they released the API so once that is the API fortunately, we knew in advance they're, also using the Apple Google Bluetooth API so. There they have been lot of P words around, dad and also we realize that working. With States and governments is easier, for us because we, are nonprofit and and. Have the brand at, the same time working. With employers is great for us because it's a very quick turnaround, cycle, for us especially.
For A large employer in a state and you have some influence on the, state Chris. Can tell you more but we have been very happy that has been a great relationship for us you work with a large employer and. Ask. Them to help us navigate the, regulatory. Environment in, that state they, just have a sentence or two yeah. Yeah. Sure so we definitely been approaching from all angles going, straight to government when possible but going to businesses, and schools we're not possible, and actually trying to form coalition's between. Schools, and even businesses to help us put pressure on government because government is not always responsive, so we are taking these kind, of going. Around the back yeah. And if the International scenarios is also great because you. Know a large employer, is. A very, important, kind of an economic engine in those countries and they say listen hey I have you, know tens. Of thousands of employees who cannot cannot, go to work unless you, state change the regulation, and work, with an open source nonprofit, it's, a lose-lose so that's those are good conversations, as well. Great. And. Do. You want to say anything about the, pros and cons of digital passports. And the key to adoption on those yeah. So there's a lot of controversy around. The. Serology tests, and I'm not a biologist so I don't want to get into that region there's, certain value to creating some kind of a passport on. Say. That I here it's Okuda for me to go back to work or I have recovered and so, on and, so yes we actually work with you, know credential, verification, companies who provide us tools for that and. We have some other partners, so we don't build the solutions ourselves, we use existing, technologies, to. Embed, them with us so there's a lot of value in creating some kind of a passport, we. Also have a team or. The South Africa. I called, Cobie ID and. They built QR code based solutions so they have integrated with us so, they're deploying it in many many scenarios, where you cannot expect the employees. To have latest, smartphones, so, even if you have a feature phone even. If you have no phone at all this. QR code based solutions actually work pretty well so. That's also part of our offering so I think I think there's a lot - it. Has to be multi-prong it's not one hammer that you, know Sol's on the problem and. I think there's a lot to learn we are sharing our learnings as we go.