Developers
July 10, 2020

Facebook AI Partners with Academic Researchers to Tackle COVID-19

The use of AI models helps the medical field predict and analyze Covid-19 data. Together with universities, they help states and hospitals.
Source: Unsplash

Facebook recently is helping researchers that are working to know better and solve Covid-19. Facebook AI, the AI division of the company, is putting its expertise to the service of health experts.  To be specific to whom Facebook is helping, Facebook is working closely with the Faculty of Mathematics and the Data Science research platform at the University of Vienna.

AI is being used to generate projections and predictions of where and how Covid-19 is spreading in Austria. The predictions help the government and health-care providers know how the pandemic is occurring in terms of data. 

They are using public data that the Austrian government shares. The data is based on the cases that appear, and weekly forecasts are done on them. Models are built to respond to the changes in specific areas.

How the Models are Different

Multivariate Hawkes processes, deep relational autoregression, and neural jump stochastic differential equations are used. After working with the data and analyzing it, the results are shared with health officials. They later pass work on the information too, and finally, pass the results to health organizations and researchers.

Facebook counts with a local Covid-19 forecasting place. This place is New York and New Jersey. These forecasts are used to make decisions on resources such as ventilators and masks. In the future, they plan to evaluate more data, such as mobility maps and the relationship to previously described models.

Facebook is using natural language processing (NLP) to make the features more visible. This makes more people receive the necessary support. When any of the models detects a “get help” request in any of the news feed post, they recommend to publish it on the help community, so more people can receive the message.

More than a dozen languages are supported using the XLM pretraining method. More languages will be added. More than half of the posts are coming from the NLP model. This model is also used for blood donations. 

A translation initiative has taken place, it is called the TICO-19. It is a consortium that helps enable translation about the virus in multiple languages. Engineers and translators from many institutions contribute to TICO-19. Institutions like Carnegie Mellon University, Google, Appen, and others, contribute to this cause.

Facebook is providing translations of specific terms related to Covid-19. By facilitating this, professional translators can expedite their work with more accuracy. It's not the only role Facebook is taking on translation. They are contributing by curating data, and providing curated datasets related to the pandemic. This will help them build translation tools that can be used in moments of crisis.

Facebook Working with NYU to Create Advanced Forecasting Models of the Virus

Facebook is also working with the New York University's Courant Institute of Mathematical Sciences. They work together to create forecasting models of the virus. This forecastings help health workers know how to manage their resources. By knowing when to expect more patients, they know when to call more staffing.

Facebook is using public data from the State of New Jersey to create daily predictions for the state. NYU colleagues use the information obtained from the models to estimate how the virus increases and where.

Facebook is also collaborating with Cornell University using the same data to build models and help the state in the best way possible.

The company is cooperating with NYU for predictive analytics. They are building forecasts for hospitals. What do they use? Reinforcement learning, causal modeling, and supervised learning techniques. All of this is done to help experts manage better their resources within the clinic. The number of patients can be predicted and their situation too, how long they are going to stay. 

By joining with NYU, the focus is 100% on the development of models. The next step FB and NYU will take is to see how they can scale their findings and benefits to other organizations. Including open-source code for the models that can adapt to other data.

An interesting point and plan Facebook is doing are removing all hateful content from the network related to Covid-19 automatically. With the help of AI, hateful memes, images, videos, comments, and posts are deleted. Is this not giving freedom of speech to the users? Well, the policy is not only applied to Covid-19, it's a safe and healthy measure taken by the company as it promotes good speech and good relationships.  

In conclusion, Facebook AI, the AI division of Facebook, has partnered with many institutions to help them build models that can improve the management of resources within the medical field. They have partnered with universities and together they analyzed data provided by different states and countries. The data is received from the state and given back to the state with a whole process of digestion and analysis. 

TagsCOVID-19Facebook AIAINYU
Lucas Bonder
Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles

Back
DevelopersJuly 10, 2020
Facebook AI Partners with Academic Researchers to Tackle COVID-19
The use of AI models helps the medical field predict and analyze Covid-19 data. Together with universities, they help states and hospitals.

Facebook recently is helping researchers that are working to know better and solve Covid-19. Facebook AI, the AI division of the company, is putting its expertise to the service of health experts.  To be specific to whom Facebook is helping, Facebook is working closely with the Faculty of Mathematics and the Data Science research platform at the University of Vienna.

AI is being used to generate projections and predictions of where and how Covid-19 is spreading in Austria. The predictions help the government and health-care providers know how the pandemic is occurring in terms of data. 

They are using public data that the Austrian government shares. The data is based on the cases that appear, and weekly forecasts are done on them. Models are built to respond to the changes in specific areas.

How the Models are Different

Multivariate Hawkes processes, deep relational autoregression, and neural jump stochastic differential equations are used. After working with the data and analyzing it, the results are shared with health officials. They later pass work on the information too, and finally, pass the results to health organizations and researchers.

Facebook counts with a local Covid-19 forecasting place. This place is New York and New Jersey. These forecasts are used to make decisions on resources such as ventilators and masks. In the future, they plan to evaluate more data, such as mobility maps and the relationship to previously described models.

Facebook is using natural language processing (NLP) to make the features more visible. This makes more people receive the necessary support. When any of the models detects a “get help” request in any of the news feed post, they recommend to publish it on the help community, so more people can receive the message.

More than a dozen languages are supported using the XLM pretraining method. More languages will be added. More than half of the posts are coming from the NLP model. This model is also used for blood donations. 

A translation initiative has taken place, it is called the TICO-19. It is a consortium that helps enable translation about the virus in multiple languages. Engineers and translators from many institutions contribute to TICO-19. Institutions like Carnegie Mellon University, Google, Appen, and others, contribute to this cause.

Facebook is providing translations of specific terms related to Covid-19. By facilitating this, professional translators can expedite their work with more accuracy. It's not the only role Facebook is taking on translation. They are contributing by curating data, and providing curated datasets related to the pandemic. This will help them build translation tools that can be used in moments of crisis.

Facebook Working with NYU to Create Advanced Forecasting Models of the Virus

Facebook is also working with the New York University's Courant Institute of Mathematical Sciences. They work together to create forecasting models of the virus. This forecastings help health workers know how to manage their resources. By knowing when to expect more patients, they know when to call more staffing.

Facebook is using public data from the State of New Jersey to create daily predictions for the state. NYU colleagues use the information obtained from the models to estimate how the virus increases and where.

Facebook is also collaborating with Cornell University using the same data to build models and help the state in the best way possible.

The company is cooperating with NYU for predictive analytics. They are building forecasts for hospitals. What do they use? Reinforcement learning, causal modeling, and supervised learning techniques. All of this is done to help experts manage better their resources within the clinic. The number of patients can be predicted and their situation too, how long they are going to stay. 

By joining with NYU, the focus is 100% on the development of models. The next step FB and NYU will take is to see how they can scale their findings and benefits to other organizations. Including open-source code for the models that can adapt to other data.

An interesting point and plan Facebook is doing are removing all hateful content from the network related to Covid-19 automatically. With the help of AI, hateful memes, images, videos, comments, and posts are deleted. Is this not giving freedom of speech to the users? Well, the policy is not only applied to Covid-19, it's a safe and healthy measure taken by the company as it promotes good speech and good relationships.  

In conclusion, Facebook AI, the AI division of Facebook, has partnered with many institutions to help them build models that can improve the management of resources within the medical field. They have partnered with universities and together they analyzed data provided by different states and countries. The data is received from the state and given back to the state with a whole process of digestion and analysis. 

COVID-19
Facebook AI
AI
NYU
About the author
Lucas Bonder -Technical Writer
Lucas is an Entrepreneur, Web Developer, and Article Writer about Technology.

Related Articles