Pandemic Resilience

Ramesh Raskar is an Associate Professor at MIT Media Lab and founder of the PathCheck Foundation. His focus is on AI and Imaging for health and sustainability. They span research in physical (e.g., sensors, health-tech), digital (e.g., automated and privacy-aware machine learning) and global (e.g., geomaps, autonomous mobility) domains. He received the Lemelson Award (2016), ACM SIGGRAPH Achievement Award (2017), DARPA Young Faculty Award (2009), Alfred P. Sloan Research Fellowship (2009), TR100 Award from MIT Technology Review (2004) and Global Indus Technovator Award (2003). He has worked on special research projects at Google [X],  Facebook and Apple and co-founded/advised several companies.  

MIT SafePaths develops free, privacy-preserving tools for COVID-19, and prevent a surveillance-state response. PathCheck Foundation, spun out of MIT SafePaths, is the world's largest nonprofit opensource project for Covid-19. PathCheck helps nations build their own solutions. PathCheck plans to build pandemic resilience solutions.

National readiness is hampered by data silos, unproven policies and the chaos of the pandemic. Unlike natural disasters, pandemics are about innocent people inflecting other innocent people. We need real time bottom-up participation from people, hospitals and public heath stake holders. 

The following approach does not work: 'Let us all contribute to the social good because it is your responsibility': We must provide a solution that works in presence of the need to maintain trade secrets between organizations and privacy of individuals.

We aim to create tools each nation (or state) can deploy for readiness. This public-private partnership involves many govt organizations, large companies (with access to data and tools), citizen groups and public health entities. Pandemics cost billions of dollars, we expect readiness with digital tools will cost less than $100K in capex plus $1 per 1,000 people per year. ($50K for a country of 50M people).

Consider the Google/Apple exposure notification API that alert individuals if they may have been exposed by an infected person using smartphone proximity. Despite the good intention of this technology, the ability of public health to deploy such solutions and integrate them with existing health infrastructure is daunting. In addition, we need a similar coordination with hospitals, pharma companies, health insurance providers and policy makers.

Pandemics readiness are about two main data related challenges: (i) unknown social risks: people being at risk because of proximity to other innocent people and (ii) unknown treatment efficacy/test reliability/hospital availability: public health and hospitals are unaware of the best approach for a rapidly evolving dynamics of the pandemic such as test relaib. We need real time participation from people and that is why smartphone apps can play a big role. We need a new AI that relies on the information stored in people’s smartphones. Via our work on privacy preserving AI at MIT, that we can indeed create such a decentralized AI and orchestrate the socio-economic interaction between the government, businesses, individuals and their communities without creating a surveillance state.

The readiness plan creates a new digital coordination tool. If we can create a new AI that has (i) a bird’s eye view of the evolving epidemic in real time (all social contacts, all treatments/tests, all hospital activities) and (ii) if we have models of prior epidemics, we can support data-driven and adaptive policies, selective actions and provide meaningful guidance for the population, businesses and government.

Via MIT SafePaths + PathCheck Foundation, If plans go well, we may have 100's of millions of users using our smartphone solutions. However, this is insufficient for a comprehensive pandemic response and inadequate for preparedness. 

In Phase1, (i) We will build on the success of PathCheck smartphone as also expand privacy preserving and trade-secret complaint to hospitals and public health entities. (ii) I am engaged with WHO, CDC, and US Congress. We will build on this momentum for policy discussions. 

In Phase2, (i) We will work with technology companies like Apple/Google, telecom players, pharma companies and insurance provider to help create a real time bird's eye view where we can aggregate data without revealing trade secrets.

In Phase3, we pilot with nations and states using a a low stakes scenario of seasonal flu.

The project serves the residents of the nation, public health decision makers as well as the government policymakers.

However, there are several stakeholders and we are engaged with many of them. We are already in the process of bringing together two dominant smartphone operating systems for the pandemic resilience (Apple and Google). I have worked at both companies, and thanks to Covid-19 efforts of PathCheck foundation, both companies are impressively engaged.

We have ongoing pilots for manual contact tracing as well as smartphone based solutions for digital contact tracing. We also engage with pharma companies and health insurance players. As faculty at MIT Media Lab, we have 80 large companies and we have held several brainstorming sessions with them.

The following approach does not work: 'Let us contribute to social good because it is your responsibility': We must provide a solution that works in presence of the need to maintain trade secrets.

Let us compare how we use technology for a hurricane versus a pandemic. National weather service can analyze and predict by collecting sensor data every day. However, a hurricane is different from the Pandemics as they are about people being at risk because of other innocent people. We need real time participation from people and that is why smartphone apps and public health compute clients can play a big role. We can indeed create such a decentralized AI and orchestrate the socio-economic interaction between the government, businesses, individuals and their communities without creating a surveillance state.

I come to these questions having spent much of the last decade researching at MIT in AI, digital health, and algorithms for privacy. We build privacy preserving global AI: a distributed machine learning method called Split Learning that can build AI without accessing any raw identifiable data from individuals. Other teams have built wonderful techniques such as federated learning and differential privacy that is now used for US Census 2020.

In March, our team at MIT created one of the first privacy preserving smartphone app called MIT SafePaths. That research led to the creation of PathCheck Foundation, a charitable nonprofit organization dedicated to building free open software and industry standards that assist US states, nations and private sector organizations with their pandemic response. Our non-profit is already building exposure notification and case management apps, backend servers and dashboards for various US states and nations.

It has been humbling experience to work on these complex challenges and based on our learnings so far, I and my team has launched the larger effort of pandemic resilience. 

Friction in data sharing is a large challenge for large scale public health solutions that use analysis, AI and machine learning. Recently techniques such as Federated Learning, Differential Privacy and Split Learning aim to address siloed and unstructured data, privacy and regulation of data sharing and incentive models for data transparent ecosystems. Split learning is a new technique developed at the MIT Media Lab’s Camera Culture group that allows for participating entities to train machine learning models without sharing any raw data.

While sharing photos on social media is so easy, the current pandemic reminds us how far behind we have fallen in keeping up with challenges that really matter. The only way to solve it by creating a privacy preserving and trade secret complaint solution.

Congressional hearing: https://pathcheck.org/event/us-congressional-hearing-exposure-notification-contact-tracing-ai/

Davos talk: https://splitlearning.github.io/

MIT research:

PathCheck Foundation: https://pathcheck.org/

The Lemelson foundation award summarized my background at the intersection of humanitarian causes and technological advances, especially in health.

This video is a good summary: https://youtu.be/9c3PxR3s72k

https://lemelson.mit.edu/winners/ramesh-raskar

http://news.mit.edu/2016/ramesh-raskar-awarded-lemelson-mit-prize-0913

I and my team conducted a tech driven intervention for the 30M people Kumbh Mela

http://kumbha.org

We need to be ready for the next epidemic or pandemic. National Weather Service that can use physical sensors to analyze and predict the next hurricane. Pandemic response requires real time co-operation from people who maybe sick or exposed. 

Instead of creating a surveillance state and a top-down system, let us build a bottom-up smartphone based solution. We need a new AI that relies on the information stored in people’s smartphones. We have learned through Split Learning, our work on privacy preserving AI at MIT, that we can indeed create such a decentralized AI and orchestrate the socio-economic interaction between the government, businesses, individuals and their communities without creating a surveillance state.

We need a National Pandemic Response Service that will allow this micro and macro aggregation for analysis and prediction. The service can have three parts: research, readiness and response. And our analysis indicates a modest budget with public-private partnership can provide a data-ready national service. And same the national service can help us be ready with a nasty flu season in the future, boost digital solution for public health and be prepared for the next pandemic.

The end goal is citizen centric digital solutions that allows citizens and public health to self-coordinate socio-economic activities, get alerts and stay safe. The necessary ingredients to achieve this goal are well studied in our work so far. 

“There’s an app for that” is a common sentiment but we need to go well beyond that. And focus on outcomes in coordination with public health.

Philanthropic and grant funding

Membership fees from countries

We have raised funds for previous phase of the project for contact tracing.

MIT Safe Paths began as a movement to develop free, open-source, privacy-first tools for COVID-19, and prevent a surveillance-state response to the pandemic. We spun off PathCheck Foundation, a 501(c)(3) nonprofit, to help nations, states and employers build their own custom solutions (app, server and dashboard). We are here to augment and simplify manual contract tracing (exposure notification, case interviews and contact followup) using privacy preserving solutions for the users. We provide a modular framework that can be customized into two separate citizen apps (i) PathCheck Bluetooth based on Google-Apple Exposure Notification and (ii) PathCheck GPS (currently known as CovidSafePaths) based on location data. Follow updates at our GitHub repo, the largest non-profit Covid19 open-source repository. Team is now made up of ~50 full time volunteers and ~20 software engineers. We have engagements in dozens of countries and US states. Team has raised philanthropic funding for software development.

The non-profit has two arms: (i) Solutions division: delivers software products and engages with public health and (ii) CoLab division: a think-tank that advances the cause of privacy first Covid19 solutions. MIT SafePaths continues cutting edge research and innovation in health, privacy and machine learning for Covid19.

Video update: (YouTube)
Slides : (Google Slides)
Website: https://covidsafepaths.org/  
Blog: https://covidsafepaths.org/blog/
Webinars: https://covidsafepaths.org/webinar/ 
Github: https://github.com/Path-Check/covid-safe-paths
Papers: https://github.com/PrivateKit/PrivacyDocuments/

News Stories: https://covidsafepaths.org/news/

We expect $800K costs in next one year for engineers and volunteers.

WHO, African CDC, Pharma Companies, Big tech companies, Citizen groups

We are in touch with most of them and work closely with them