The Swiss Startup Jua has launched an AI-driven weather forecast button, which is that striking models of tech giants are beating, which may make the world's most accurate weather forecast system.
Jua claims that his model-called EPT-2-SEI faster and more precisely than both Aurora from Microsoft and the Google Deepmind Graphcast. In separate, examined by experts StudiesIt was shown that both models are more precise than the European Center for Middle Weather forecasts (ECMWF) of the ESE forecast with a medium area, which is widely considered worldwide.
Jua supports his bold demands with a new report that was published today and sets with top animal models, including Aurora and two of the best ECMWs, from EPT-2 from Head-to Head: EnS and IFS-Gres.
According to the newspaper, EPT-2 came out and provided the most precise forecasts across the board. Aurora hit important variables such as 10 meters wind speed and 2-meter air temperature over a period of 10 days, predicted 25% faster and recorded the lowest error values of all tested models. Jua says that this has achieved all of this, while she used 75% less computing power than Aurora, the second most efficient system was tested.
According to Jua, research is to be published in the open access archivariv-arxiv in the next week.
The 💜 the EU technology
The latest rumors from the EU -Tech scene, a story of our wise old founder Boris and some questionable KI art. It is free of charge every week in your inbox. Register now!
The Graphcast model from Deepmind was not included in the study. Nevertheless, Marvin Gabler, CEO and co -founder of Jua, is confident that it can surpass the entire competition.
“We respect players like Microsoft Aurora, Graphcast and tomorrow.
The AI-based weather forecast has led waves in recent years, which are due to the demand for more precise and cheaper opportunities to predict the climate of the earth.
Traditional weather models, such as those of ECMWF or NOAA, use complex physics equations that are operated on billions in dollars supercomputers. AI models skip the equations, learning patterns from massive data records, which makes thousands of forecasts faster on much cheaper, less energy-intensive machines.
According to Gabler, however, Jua goes one step further than previous AI-based forecast. “While other AI retrofit in Legacy systems, we have built up a native physics simulation that understands how the atmosphere of the earth actually behaves,” he said.
Jua released his first global AI weather model three years ago. The startup has now introduced a total of $ 27 million of supporters from supporters, including 468 capital, future energy ventures and promus ventures.
