Google’s AI Weather Model Could End Getting Caught in the Rain

Google DeepMind’s GenCast, a cutting-edge AI-powered weather forecaster, is shaking up meteorology by delivering faster and more accurate predictions. Promising to outperform traditional supercomputers, GenCast boasts a success rate of 99.8% for forecasts up to 15 days ahead, setting a new benchmark for weather predictions.

GenCast is more than a tool for deciding whether to grab an umbrella. It’s a potential game-changer in spotting natural disasters, like typhoons, earlier than ever before. While traditional weather agencies rely on massive supercomputers crunching complex atmospheric equations, GenCast takes a radically different approach—using AI to analyze patterns in decades of historical weather data.

 

MORE: Google has revealed the Top Trending searches of 2023

DISCOVER: How a UK Couple’s Battle with Google Led to a £2.4 Billion Fine

 

How GenCast Works

GenCast generates an ensemble forecast—50 potential weather outcomes based on current data. Ilan Price, a Google DeepMind engineer, explains why this method is so impactful: “Such ensemble forecasts are more useful than relying on a single forecast, as they provide decision makers with a fuller picture of possible weather conditions… and how likely each scenario is.”

Unlike conventional models, which simulate intricate atmospheric physics, GenCast employs diffusion AI models typically used in video, image, and music creation. This AI predicts weather patterns in the same way it might create a video based on a text prompt. Specifically tailored for Earth’s spherical shape, it has been trained on 40 years of weather data.

 

MORE: Russia Fines Google Over YouTube Drama

DISCOVER: Google Unveils the Most Popular Gift Searches of 2023

Superior Speed and Accuracy

GenCast’s speed is one of its biggest advantages. While traditional forecasts, like the European Centre for Medium-Range Weather Forecasts (ECMWF)’s ENS model, take hours using thousands of processors, GenCast delivers a 15-day forecast in just eight minutes using a single processor.

When tested against 2019 data, GenCast outperformed ENS in 97.2% of cases. In extreme scenarios, like predicting the devastating Typhoon Hagibis, GenCast provided a warning 12 hours earlier than ENS. As Price points out, “Getting better and more advanced warnings of where they’ll strike land is invaluable.”

 

MORE: Google Searches for ‘Marketing Degree’ Soar Thanks To Emily in Paris

DISCOVER: Everything You Need To Know About Google’s European Black Founders Fund

AI vs. Traditional Models

Despite its successes, GenCast isn’t replacing traditional models anytime soon. Systems like ENS still deliver higher-resolution forecasts, dividing the globe into finer grids of 0.1 degrees compared to GenCast’s 0.25 degrees. Additionally, traditional models generate the foundational data needed to train AI.

However, the sheer efficiency of GenCast opens new possibilities for AI-powered forecasts, especially in renewable energy planning and extreme weather event preparation. As climate change drives more volatile weather, accurate and swift forecasts are more critical than ever.

 

MORE: What is ‘The Google Effect’?

DISCOVER: Google Pixel 9 Just Solved the One Problem Every Photographer Faces

With GenCast’s ability to predict weather faster and potentially save lives, the future of meteorology looks brighter—and drier—than ever before.