In brief: Though global economies have begun to open up, the Covid-19 virus is still spreading throughout the world, infecting thousands of new people every day. To help curb the spread of the disease, MIT researchers have developed an AI model that can detect the virus' presence in even asymptomatic individuals.
The potential good that such a model could do is probably pretty obvious. Suppose the model was refined and rolled out to the general public somehow, perhaps in the form of a free mobile app.
In that case, it could help people screen themselves for the infection and either get tested or avoid contact with others, if necessary.
Teachers, for example, could use it every day before heading into class, as could other individuals that work in close proximity to strangers – front-line retail employees are another key audience for such a tool.
Fortunately, porting the model to an app is precisely what researchers are working toward now. The team still needs to finish developing it and, of course, obtain FDA approval before it can be distributed widely, though.
However, early results are promising. To date, the model has been trained on "tens of thousands" of cough samples, as well as "spoken words." When researchers input new cough recordings into the model, it can accurately identify full-fledged Covid-19 infections in "98.5 percent of coughs" and 100 percent for asymptomatic individuals.
It seems that, even for people who aren't showing severe symptoms of Covid-19, the way they cough, breathe, and speak may contain small indicators that can point toward probable infection.
If this model truly is as accurate as researchers believe, we hope the FDA approves it sooner rather than later – the world could certainly use it right about now.
Image credit: Josep Suria