Friends, this is another cross-post of an article that was originally published on the Google Cloud blog on February 24, 2025 by Will Grannis, VP and CTO of Google Cloud. I wrote one response to a question from the LinkedIn community, and several of my OCTO colleagues contributed other answers as well. Note that my response had very little to do with Gemini 2.0 itself! But hey, they decided to publish it anyway. Enjoy, and please submit questions for future columns here.
In addition to language, video, code, and other modalities, AI is now transforming science in significant ways. This includes advancements in protein folding, materials science, and weather forecasting that are accelerating as we head into 2025.
In the AI-driven weather space, Google DeepMind’s WeatherNext Graph and Gen AI models predict global weather variables with state-of-the-art accuracy up to 15 days in advance. Unlike traditional numerical weather prediction systems, WeatherNext models learn from historical weather observations rather than predicting the weather with hand-crafted physics formulas. The probabilistic WeatherNext Gen model also uses diffusion, similar to the architectures used by Imagen and Veo. WeatherNext models are faster and more efficient than traditional physics-based weather models and yield superior forecast reliability.
What does this mean for you and your business? Well, almost every industry and community is affected by the weather. If you’re working in industries like agriculture, manufacturing, logistics, energy, and financial services, it’s worth checking out what these new AI weather forecasting models can do to improve your day to day operations. Imagine knowing about severe weather events days earlier and with better accuracy – powerful stuff.
Jeff Sternberg 