Introducing WeatherNext 2: Google’s Revolutionary AI Weather Forecasting Model
Google DeepMind and Google Research have unveiled WeatherNext 2, their most sophisticated AI weather forecasting model to date. This groundbreaking model generates predictions at a speed eight times faster than previous systems, capable of producing hundreds of weather scenarios in under a minute using a single TPU chip. This is a stark contrast to the hours required by traditional supercomputers.
WeatherNext 2 outperforms its predecessor in 99.9% of weather variables. This includes factors such as temperature, wind, and humidity across all forecast periods ranging from 0-15 days. The model employs a novel Functional Generative Network (FGN) approach, injecting sophisticated noise to create physically realistic forecasts with hourly resolution capabilities.
Google has already integrated WeatherNext 2 into a variety of platforms and services:
- Search
- Gemini
- Pixel Weather
- Google Maps Platform’s Weather API
A broader rollout is planned for Google Maps. Additionally, the technology is accessible through Google Earth Engine, BigQuery, and an early access program on Vertex AI for enterprise customers seeking customized forecasting solutions.
Source: https://blog.google/technology/google-deepmind/weathernext-2/