Using a vast neural network capable of making ultra-quick decisions based on images and predictive learning (rather than physics or meteorology), Google researchers may have finally figured out how to make weather forecasts hour by hour. Traditionally, weather forecasts have taken up to three hours to form, and thus are a little less accurate. As for Google’s, their neural network can formalize a report in less than 10 minutes. The faster technology finds real-life application when users want weather updates before popping out for a run, commuting, or planning an evening’s itinerary. Plus, because the neural network is examining rain-pattern imagery and not scientific data points, it can provide detailed forecasts for particular neighborhoods and stretches of remote land. Read more at Ars Technica.
Improved Weather Forecasts Made Using Neural Networks