Archive — US — techcrunch
Nvidia’s new AI weather models probably saw this storm coming weeks ago
Nvidia has introduced its Earth-2 weather forecasting models, leveraging AI to improve accuracy and speed. Announced at the American Meteorological Society meeting, the medium-range model reportedly outperforms Google’s GenCast. The new models democratize weather forecasting, aiming for accessibility across nations and industries, addressing both climate impacts and national security concerns.
Key points
- Nvidia's Earth-2 models were released amid a winter storm affecting the U.S.
- The Earth-2 Medium Range model surpasses Google's GenCast in over 70 variables.
- Models utilize a new Nvidia architecture called Atlas.
- Nowcasting provides short-term predictions, aiding meteorologists in storm impacts.
- Global Data Assimilation uses diverse sources for continuous weather snapshots.
- Nvidia's models use significantly less computing power than traditional methods, enabling quicker processing.
- Aim to make advanced weather forecasting accessible to less wealthy nations and smaller organizations.
Why it matters
- Government agencies and meteorological services in various countries are increasingly reliant on AI advancements for accurate weather predictions.
- The rise of AI in weather forecasting reflects broader trends in IT, where AI democratizes access to critical technologies, particularly regarding climate resilience.
- Recent weather extremes have heightened the urgency for reliable forecasting models that balance accuracy and accessibility.
Nvidia's Earth-2 models signify a leap in AI-driven weather forecasting, crucial for enhancing global preparedness against climate change, making advanced tools available to a wider audience and intertwining national security with weather prediction capabilities.