Google DeepMind researchers have developed a new AI weather forecasting tool called GenCast, which is more accurate and faster than the current best system, ENS. GenCast outperformed ENS 97% of the time for predictions up to 15 days in advance. This model was tested on over 1,320 weather scenarios. GenCast is a diffusion machine learning model, trained on four decades of data from the European Centre for Medium-Range Weather Forecasts (ECMWF). During experiments, GenCast generated a forecast for 2019 and compared the results to the actual weather and ENS’ predictions.
The model creates an ensemble of over 50 different predictions to help authorities prepare for extreme weather events in advance. GenCast can generate predictions in 8 minutes, compared to hours for traditional models. GenCast improves upon DeepMind’s olderGraphCast model from last year. AI is not likely to replace traditional forecasting soon, but it can enhance current methods by combining human assessment, traditional models, and AI-based weather forecasting.
The ainewsarticles.com article you just read is a brief synopsis; the original article can be found here: Read the Full Article…
