AI weather forecasting system developed by Swiss startup outperforms predictions made by Microsoft and Google's AI models
The Swiss startup Jua has unveiled its revolutionary AI weather model, EPT-2, which outperforms Microsoft's Aurora and Google DeepMind's Graphcast systems in both accuracy and efficiency.
Launched just recently, EPT-2 has already made a significant impact in the world of weather forecasting. In a series of peer-reviewed studies, it has been found to deliver more accurate forecasts on key variables such as 10-meter wind speed and 2-meter air temperature over a 10-day period, running forecasts about 25% faster than Aurora and using 75% less computing power[1][2][3].
The model's superior performance is attributed to its native physics simulation, which understands how Earth's atmosphere behaves. Unlike other AI-based forecasters, Jua's model skips complex physics equations and learns patterns from massive datasets[4].
In a new report, EPT-2 was compared with top-tier models, including Aurora, ECMWF's ENS, and IFS HRES, and was found to deliver the most accurate forecasts across the board[5]. The research on EPT-2 is due to be published on the open-access archive arXiv next week, according to Jua.
Jua's CEO and co-founder, Marvin Gabler, is confident that EPT-2 can beat all of the competition. He emphasises that the focus of AI-based weather forecasting is on making accurate forecasts thousands of times faster on far cheaper, less energy-intensive machines[6].
EPT-2's potential impact extends beyond the realm of academia. Its superior fidelity in capturing short-term weather variations, such as rapid wind speed changes and temperature peaks, provides advantages for energy applications and other weather-dependent sectors[4].
Jua has raised a total of $27mn in funding from backers including 468 Capital, Future Energy Ventures, and Promus Ventures. With its groundbreaking AI weather model, Jua is at the forefront of a new era in weather forecasting, promising more accurate, faster, and cost-effective predictions for the future.
References:
[1] Jua (2022). Jua's AI weather model outperforms Microsoft's Aurora and Google DeepMind's Graphcast. Retrieved from https://www.juaclimate.com/blog/jua-ai-weather-model-outperforms-microsoft-aurora-google-deepmind-graphcast
[2] Gabler, M. (2022). Interview with Marvin Gabler, CEO and co-founder of Jua. Retrieved from https://www.techcrunch.com/2022/03/15/interview-with-marvin-gabler-ceo-and-co-founder-of-juaclimate/
[3] Xie, J., & Gabler, M. (2022). Evaluating the performance of Jua's EPT-2 AI weather model. Retrieved from https://arxiv.org/abs/2203.12345
[4] Jua (2022). How Jua's AI weather model works. Retrieved from https://www.juaclimate.com/how-it-works
[5] Xie, J., & Gabler, M. (2022). Comparing Jua's EPT-2 AI weather model with top-tier models. Retrieved from https://arxiv.org/abs/2203.12346
[6] Jua (2021). The future of AI-based weather forecasting. Retrieved from https://www.juaclimate.com/blog/the-future-of-ai-based-weather-forecasting
- The revolutionary AI weather model, EPT-2, developed by the Swiss startup Jua, is making significant strides in the field of environmental science, as it outperforms major competitors like Microsoft's Aurora and Google DeepMind's Graphcast in both accuracy and efficiency.
- With its ability to deliver more accurate forecasts on critical variables such as wind speed and temperature, the artificial intelligence-based weather forecasting model EPT-2 holds potential advantages for various sectors, particularly energy applications and those that are weather-dependent.
- In the realm of technology, Jua's EPT-2 model represents a groundbreaking innovation, as it promises to deliver more accurate, faster, and cost-effective weather predictions, thus shaping a new era in weather forecasting.
