Google has unveiled a new and improved AI model (WeatherNext 2 ) for weather forecasts , which will be integrated into popular products like Search, Gemini, and Pixel phones . The company claims that the AI model is faster and more accurate than ever .
AI-powered weather forecasts have proven quite accurate so far. AI models can make predictions faster and more efficiently than traditional physics-based models . Until now, Google's work in this area has been mostly experimental. Now, these forecasts are becoming a selling point for the company's products and services.
“We’re taking it out of the lab and really putting it in the hands of users in more ways than before, and we’re essentially moving away from the experimental nature because we have confidence that our predictions are actually effective and useful,” said Peter Battaglia, senior director of research and sustainability at Google DeepMind.
See also: What Google Messages features are rolling out
WeatherNext 2: Google's effective AI model for weather forecasts
The new AI model, WeatherNext 2, can generate forecasts eight times faster than Google's previous model and is also more accurate at predicting 99.9% of variables like temperature or wind. WeatherNext 2 can produce hundreds of possible outcomes from a given starting point. It takes less than a minute using one of TPU Google's to make a forecast, something the company says would typically take several hours to achieve using physics-based models on a supercomputer.
Traditional models attempt to recreate the complex physics of the atmosphere to produce predictions. AI models, in contrast, attempt to discern patterns from historical weather data to predict future outcomes.
See also: Google Call Recording is coming to more Pixel devices
Google has managed to simplify its process by using a strategy it calls Functional Generative Network (FGN) in WeatherNext 2. Older AI weather models still required iterative processing to generate a forecast. FGN is more efficient because it incorporates noise — or targeted randomness — into the model each time it is given an input, so WeatherNext 2 can generate many different possible outcomes in a single step.
The developments allow WeatherNext 2 to make forecasts up to 15 days in advance and generate hourly forecasts. Google is counting on this to attract business customers as well as individual consumers.
“We found that energy, agriculture, transportation, logistics, and customers in many other industries are very interested in these hourly steps. It helps them make more accurate decisions about things that business their,” said Akib Uddin, product manager at Google Research.
See also: Google: Online shopping becomes easier with new AI features
In addition to adding WeatherNext 2 to Maps, Search, Gemini, and Pixel Weather, Google is also offering an early access for customers interested in custom modeling. Forecast data is also available in Google Earth Engine for geospatial analysis and BigQuery for large-scale data analysis.
