An MIT machine learning model predicts a plateau in the US, and an "exponential explosion" if lockdown ends early
What works: staying indoors and keeping businesses closed. (Valentine Flauraud/Getty Images)
Using COVID-19 data from earlier this year, MIT has trained an AI to predict what will happen in the United States and elsewhere if governments relax quarantine rules early. The paper’s estimates are both promising and worrisome.
First, the positive: the paper “predicts a halting of infection spread by 20 April 2020,” which indicates that lockdown measures in the United States have worked and the country will soon plateau.
“Our results unequivocally indicate that the countries in which rapid government interventions and strict public health measures for quarantine and isolation were implemented were successful in halting the spread of infection and prevent it from exploding exponentially,” the paper writes.
While evidence shows that lockdowns are working, the authors are clear that their model's estimates don't suggest that quarantines can be lifted without severe consequences. George Barbastathis, one of the model's developers, told TechCrunch that if the United States and others open soon, they may see a devastating second-wave similar in nature to Singapore's.
MIT believes that its model, built using COVID-19 data, may be more accurate than others relying on data from past epidemics, like SARS or MERS. And by retroactively testing their predictions against infection counts from earlier in March, MIT is confident in the model’s ability to estimate what spikes may occur following a societal reopening.
From the paper’s abstract:
Relaxing or reversing quarantine measures right now will lead to an exponential explosion in the infected case count, thus nullifying the role played by all measures implemented in the US since mid March 2020.