AI surpasses traditional weather forecast in accuracy for the first time – Kommersant

For the first time, artificial intelligence (AI) was able to predict the weather much more accurately than meteorologists using traditional information collection and analysis systems. According to a study published in the journal Sciencethe GraphCast neural network from Google DeepMind surpassed the forecast accuracy of the European Center for Medium-Range Weather Forecasts system.

In the study, GraphCast was able to produce a more accurate forecast for 90% of the 1,380 parameters tested, including temperature, pressure, wind speed and direction, and humidity, and, most importantly, significantly outperformed its competitor in predicting extreme natural events. In addition, the neural network does this much faster than a traditional forecasting system.

“It predicts hundreds of ten-day weather variables with a resolution of 0.25° worldwide in less than one minute,” the study says.

Matthew Chantry, machine learning coordinator at ECMWF, said in an interview Financial Timesthat artificial intelligence in meteorology has advanced “much faster and much stronger than we expected just two years ago.” In addition, according to him, the neural network also turned out to be “a thousand times” more efficient than traditional systems in terms of energy consumption required for calculations.

Kirill Sarkhanyants