Google Cloud AI and the Harvard Global Health Institute on Nov. 16 released new versions of their jointly developed COVID-19 Public Forecasts dashboard, which provides projections for hospitalizations and death rates across the U.S., according to VentureBeat.
Four details:
1. Google Cloud and Harvard launched the dashboard in August using data from sources including Johns Hopkins University, Descartes Lab and the U.S. Census Bureau to create projections at the state and county levels.
2. The newly updated models now expand beyond the U.S., and instead of offering regional predictions for 14 days in the future, it now provides projections for 28 days in advance, with confidence interval estimates to account for any uncertainty in the data. The updated model is about 50 percent more accurate than the first version.
3. Google will introduce the COVID-19 Public Forecasts for Japan and said it is looking into support for other countries as well.
4. In addition to the 28-day forecast, Google also has added customizable features to the dashboard to tackle new problems and datasets. The company also is creating an AI-powered "what-if" model for decision-making around COVID-19 and other infectious diseases.