GraphCast: Google robot that will leave meteorologists out of work

In every newscast or television program, one of the passages that refuses to die is climate or weather condition. There, meteorologists provide their forecasts of conditions for the coming hours and days, usually supported by very popular applications such as Windy or AccuWeather.

The truth is that Google wants to dethrone meteorology as studies show its new AI robot called GraphCast has a 90% chance of giving a forecast.

The machine learning model promises 10-day forecasts that are better, faster, and more energy efficient than the tools that power your current weather app.

“We believe this marks a turning point in weather forecasting,” the researchers wrote. Google in a study published Tuesday.

The new GraphCast system essentially takes historical data and mixes it with specific climate information.

GraphCast starts with the current weather state Earth and weather data from six hours ago. It then makes a forecast about what the weather will be like in six hours. GraphCast then feeds these predictions back into the model, performs the same calculations, and returns long-term predictions.

The Google team compared the GraphCasts results with the current model used for medium-range weather forecasting, called HRES. According to the study, GraphCast “significantly” outperformed HRES on 90% of the targets used in the test.

GraphCast has also had surprising success predicting extreme weather events, including tropical cyclones and strange temperature changes, although it was not specifically trained to deal with them.

This is a relief for meteorologists:

“Our approach should not be considered a replacement for traditional weather forecasting methods,” the study authors write. “Rather, our work should be interpreted as evidence that [la predicción meteorológica de aprendizaje automático] “It is capable of solving real-world forecasting problems and has the potential to complement and improve current best-in-class methods.”