Искусственный интеллект делает прогноз осадков точнее существующих технологий
Weather forecasts are not only important for the average person leaving home. Predicting precipitation for the next hour or two is especially important for outdoor and aviation events, as well as for emergency planning. The DeepMind development team uses high-precision radar data that tracks precipitation every five minutes at a resolution of one kilometer. The researchers used an approach known as generative modeling. Such a model analyzes the last 20 minutes of radar observations and creates a forecast for the next 90 minutes ahead. This tool is used to predict moderate to heavy rainfall, which has the greatest impact on people's lives and the economy. More than 50 meteorologists from the U.K. Bureau of Meteorology evaluated the new artificial intelligence. DeepMind's "Deep Generative Rain Model" was rated higher in 89 percent of cases than current weather forecasting methods commonly used today.
Translated with www.DeepL.com/Translator (free version)