Breakthrough in the weather forecast. New AI will save years of work is a thousand times more powerful

« This is a completely different approach than people used before. It is quite clear that it will change it all, that it will be a new way of predicting, » he said according The Guardian Richard Turner, professor of machine learning at the University of Cambridge.

He said that the model will eventually be able to create accurate eight -day predictions compared to the current five -day as well as hyperlocal predictions. The only desktop researcher will provide accurate weather forecasts using new artificial intelligence, which is many times faster and uses a thousand times less computing performance.

The new system will be used to quickly provide tailor -made forecasts for a specific industry or places, such as predicting temperatures for African agriculture or wind speed for a company engaged in renewable energy sources in Europe.

This is in contrast to traditional weather forecast systems, where the creation of a system for customized will take years of work of large teams of researchers. Supercomputers take to process the real world measurement to create prediction models for hours.

The artificial intelligence Aardvark Weather is trained on data from meteorological stations, satellites, balloons, boats and aircraft from around the world, allowing real predictions to create.

Weather forecasts are currently being created through complex phases, each of which lasts several hours of custom supercomputers and requires large teams of experts who develop, maintain and deploy them.

According to the research published in Nature, the University of Cambridge, the Alan Turing Institute, Microsoft Research and the European Center for Medium -term weather forecasts (ECMWF), offer the possibility to significantly increase speed and accuracy.

Scott Hosking, director of science and innovation in the field of environment and sustainability at the Alan Turing Institute, said that this breakthrough discovery could « democratize predictions » by making power technology available to developing countries around the world.

It would also be assistance for legislators, crisis experts during emergencies and industries that are dependent on accurate weather forecasts.

Anna Allen, the main author of the article from the University of Cambridge, noted that these findings open the way to better predictions of natural disasters such as hurricanes, fires and tornadoes, as well as other climatic questions such as air quality, ocean dynamics and seabed predictions.

Aardvark is based on recent research by Huawei, Google and Microsoft, which showed that one step of the weather forecasting process known as a numerical solver that calculates weather development, can be replaced by artificial intelligence and gain more and more accurate predictions. ECMWF already uses this approach.

Researchers said that using only 10 % of the input data required by existing systems, the Aardvark may already overcome the US National GFS forecast system in some respects and is competitive with the United States Meteorological Services.