Google announced Wednesday that its artificial intelligence (AI) agent has outperformed the world's best weather forecasters.
Google's DeepMind researchers Ilan Price and Matthew Willson wrote in a blog post: He said he made it recently. The “AI ensemble model,” dubbed GenCast, “provides better forecasts for both daily weather and extreme events up to 15 days in advance than the top-of-the-line operational system, the European Center for Medium-Range Forecasts (ECMWF) ENS. I will.”
In their post, Wilson and Price said they taught GenCast “on historical weather data up to 2018” and “tested it on data from 2019” when trying to analyze the model's skill.
“GenCast demonstrated better forecasting skill than ECMWF's ENS, the top operational ensemble forecasting system that many national and local decisions rely on every day,” the researchers said.
The researchers said they evaluated the capabilities of ECMWF's ENS and GenCast by examining “predictions of different variables at different lead times, a total of 1320 combinations.”
According to the American Weather Association, Predicted lead time This is the time from when a forecast is announced until the predicted phenomenon occurs.
The blog post says wind speed and temperature were also among the variables tested.
According to DeepMind researchers, their system outperformed ENS with 97.2% accuracy when it came to predicting different variables at different lead times.
They said that when lead times exceeded 36 hours, GenCast outperformed ENS in prediction accuracy for different variables at different lead times 99.8 percent of the time.
Wilson and Price note that despite GenCast's success, “traditional models remain essential for forecasting” because “traditional models provide the necessary training data and initial weather conditions for models such as GenCast.” said.
“This collaboration between AI and traditional meteorology highlights the power of combined approaches to improve forecasts and better serve society,” the researchers said.
The Hill has reached out to ECMWF for further comment.
