Facebook says that it wants to use Artificial Intelligence tech and massive amounts of data it has at its disposal to predict the spread and spiking of coronavirus cases across the US and ultimately the world.
The system that Facebook proposes to develop would be implemented in all US counties (not at state-level), taking into account the features of the virus and social circumstances, and be capable of producing predictions about a spike in the number of cases two weeks before it happens.
Additional data that Facebook's new algorithm would require should come from what is said will be “anonymized” (personal) and time based case data from each county. Reports are saying that this will not only allow Facebook to forecast events around the spread of the epidemic, but also “influence” social distancing and limited mobility.
The tech giant said that its goal is to increase quality of predictions, and touts the model that is now being experimented on as performing strongly in forecasting the spread of the disease compared to other available tools, that are referred to as being state-of-the-art.
In addition to the algorithm-based system, Facebook is also working on what it calls a neural autoregressive model that would allow it to decouple regional elements like demographics, policies put in place locally, and mobility of people, from the characteristics of the disease itself thanks to which it spreads.
The company proposes a method that would predict coronavirus spikes in one area based on data collected and processed in one that is already affected, and then provide the healthcare system “with the right elements to limit the spread of the disease.”
The plan is to implement this in the EU as well, while Facebook said it would publish its predictions on a daily basis on UN's Humanitarian Data Exchange platform.