Google says its new AI model outperforms the top weather forecasting system

Google’s DeepMind team unveiled an AI weather forecasting model this week called GenCast.

In a paper published in Nature, DeepMind researchers say they found that GenCast outperforms the European Center for Medium-Range Weather Forecasts’ ENS – apparently the world’s leading forecasting system.

And in a blog post, the DeepMind team gave a more accessible explanation of the technology: While its previous climate model was “objective, and provided a single, very good estimate of the future climate,” GenCast “combines a collection of 50 or more forecasts, each representing a trajectory of possible climate,” creates a “complex probability distribution of future climates.”

As for how it stacks up against the ENS, the team said they trained GenCast with weather data up to 2018, then compared its predictions for 2019, finding that GenCast was more accurate 97.2 percent of the time.

Google says GenCast is part of its AI-based weather models, which are starting to feature in Google Search and Maps. It also plans to release real-time and historical forecasts from GenCast, which anyone can use in their research and models.