As we have already informed you, Google Cloud and Harvard Global Health Institute have partnered to develop the COVID-19 Public Forecasts for predicting COVID-19 cases, deaths, and other statistics in the US. Recently, the two entities announced an update to this Artificial Intelligence operated (AI) model. This model aims to be of great help to healthcare organizations and first responders in tackling the pandemic. According to a blog post by Google, the model has been improved significantly in five ways:
Longer Forecast: Before, the model could make predictions for the next 14 days but now it can make more accurate predictions for 28 days while involving confidence intervals in the mix to account for uncertainty.
Improved Model — Google claims that the AI quality has also improved through cutting-edge research done by a dedicated AI team working tirelessly for weeks. Under the new model, the predictions have improved by approximately 50% even after the longer forecast prediction horizon.
Expansion to Other Countries: Initially, the model could only make predictions in the US. The new and improved model can now also make forecasts for Japan. The model uses public COVID-19 Situation Reports in the respective countries to provide the forecasts. Moreover, the model will also be able to deliver prefecture-wise (subdivision-wise) predictions in Japan.
Customized Forecasts: In this rapidly evolving pandemic, the forecasts are used by first responders, public sector, and various other organizations and not all of them have the same needs regarding the predictions. Since the organizations wanted to include their data in the model as inputs, Google has introduced the option to customize the forecasts making the model into a customizable system that takes further data into the mix to provide more precise results.
What-if Analysis: The enterprises are also developing a What-if Model that can be used in decision making and providing acumen for better-informed policy decisions regarding COVID-19.
Google Cloud AI research head Tomas Pfister said, “on the one hand, the potential impact of more accurate and robust forecasts on the COVID-19 response was large, so launching quickly was important. However, we also needed to make sure that the quality was high enough to help inform decision-makers as well as ensuring that it did not further any existing disparities through model bias.”
Further, Google has also partnered with a few early-testers, including HCA Healthcare, to test the early version of the predictions before they were available to the public to improve the forecasts. Lastly, Google also held a Fairness Analysis to analyze the disproportionate impact of COVID-19 on communities of color to make the data more precise and accurate.