Predict the progress of Covid-19 with Artificial Intelligence (AI).
This is the goal, partly already implemented, of the Paris Facebook Center for Research on Artificial Intelligence (FAIR), which today celebrates its sixth anniversary at the Facebook AI Innovation Summit.
Underlying the model is a mechanism that predicts future cases in a particular context, using data on recent cases in statistically similar areas.
The organization is making open-source all predictive models related to Covid-19 so that rescuers, governments and researchers can use them to continue helping their communities.
"Starting from the US counties, we found that social connection is an important aspect in this mechanism - explains Facebook - For our analysis, we use the Social Connectedness Index, a project released through Facebook's Data for Good initiative, which measures Strength of the connection between two geographic areas, as represented by the friendship ties on Facebook. Our results show that the social connection of statistically similar counties is two to eight times higher in our model than that of unrelated counties. this analysis may also provide new insights for other countries and illustrate the usefulness of this indicator at a general level ".
The work is based on the US scenario but also looks to Europe. Since the beginning of the pandemic, the Biocom-SC team of the Universitat Politècnica de Catalunya, which together with the University of Vienna collaborates with FAIR, has provided the European Commission with comprehensive reports and forecasts on the spread of the disease. "We want to allow governments and organizations to develop trends on Covid-19, to better address it today and, who knows, foresee a new one in the future," the team explained.
The Artificial Intelligence of the FAIR division of Facebook also takes into account vaccination rates, which means that it can dynamically adapt according to the situation of a specific country and, in this way, create valid models in real time, capable of mirroring the context. in a timely manner. "By opening the code, we hope to help advance epidemiological predictions by enabling researchers to reuse, extend and improve our methods. In addition, we continue to provide daily forecasts for the United States, which are available on the Data for Good archive of Facebook AI ".
