Competitor Swiss AI outperforms Microsoft and Google's weather prediction systems, claims local technology firm.
Swiss AI Weather Model Outperforms Google DeepMind and Microsoft's Forecasting Models
A new AI weather model, EPT-2, developed by the Swiss startup Jua, has made waves in the industry by outperforming Microsoft's Aurora and Google DeepMind's Graphcast in terms of accuracy and efficiency.
According to separate peer-reviewed studies, EPT-2 has demonstrated the most accurate forecasts for 10-meter wind speed and 2-meter air temperature out to 10 days. It outperforms Aurora especially beyond 6 days in wind speed and across almost all lead times for temperature. The model also surpasses Google DeepMind's Graphcast, according to the latest industry reports.
In terms of efficiency, EPT-2 processes forecasts about 25% faster than Microsoft Aurora while using 75% less computing power. This efficiency stems from Jua developing a physics simulation model from scratch, enabling a more fundamental understanding of atmospheric behavior without relying on the billion-dollar supercomputers typical of traditional systems.
Jua's CEO and co-founder, Marvin Gabler, is confident that EPT-2 can beat all of the competition. He believes that the model's native physics simulation, which understands how Earth's atmosphere behaves, and its ability to skip complex physics equations and learn patterns from massive datasets, set it apart from other AI-based forecasters that retrofit AI onto legacy systems.
The development of EPT-2 is a significant advancement in AI-driven weather forecasting technology. Traditional weather models, like those from ECMWF or NOAA, use billion-dollar supercomputers and complex physics equations, making them slower and less energy-efficient. In contrast, EPT-2 is potentially thousands of times faster and less energy-intensive than these models.
AI-based weather forecasting has been making waves in recent years due to the demand for more accurate and cheaper ways to predict the Earth's climate. Jua's first global AI weather model was released three years ago, and the company has since raised a total of $27mn in funding from backers including 468 Capital, Future Energy Ventures, and Promus Ventures.
A new report, published today, puts EPT-2 head-to-head with top-tier models, including Aurora and two of ECMWF's best: ENS and IFS HRES. According to the paper, EPT-2 came out on top, delivering the most accurate forecasts across the board. The research is due to be published on the open-access archive arXiv next week, according to Jua.
As AI-based weather forecasting continues to evolve, Jua's EPT-2 is leading the way in accuracy and efficiency, paving the way for a more predictable and sustainable future.
The Swiss AI weather model, EPT-2, has shown superiority in both accuracy and efficiency compared to Google DeepMind's Graphcast and Microsoft's Aurora, marking a significant advancement in environmental-science and weather-forecasting. This innovation in artificial-intelligence technology processes forecasts faster and uses less computing power than traditional methods, potentially revolutionizing the field. The development of EPT-2 is not only a victory for Jua but also a step towards a more predictable and sustainable future in the realm of science and technology.
