Holiday Ayo - Google introduced GenCast, a new AI model for weather prediction. This AI is claimed to be much more accurate than the best current systems, including ENS from the European Center for Medium-Range Weather Forecasts (ECMWF).

Accurate weather forecasts are important for both carrying out daily activities or preparing for disasters, even renewable energy.

Quoting GSM Arena, GenCast surpasses the best current system, ECMWF by ENS, regarding weather forecasts up to 25 days in the future.

GenCast is a diffusion model, similar to the supporting model for AI image-generating tools that are widely found in the community, but GenCast will focus more on the geometry of the Earth.

This AI model is trained with historical data for four decades from the ECMWF archive.

To test it, Google trained GenCast with historical weather data up to 2018 and ran 1,320 different forecasts for 2019, then compared them with ENS and actual weather.

GenCast was reported to be more accurate than ENS in 97.2 percent of cases, rising to 99.8 percent more accurate for forecasts 36 hours or longer in the future.

Giving local governments more time to prepare for severe weather is one use case for this AI model.

In addition, GenCast is also reported to be able to predict wind speeds around sources, weather around solar panels, and more.

GenCast is an ensemble model, meaning it can produce more than 50 forecasts with varying probabilities.

Google explained that one prediction covering a 15-day weather forecast can be made in 8 minutes using Google Cloud TPU v5. Some forecasts can also be done in parallel.

Meanwhile, traditional weather forecast models take a long time to do their work on supercomputers. Google released GenCast as an open model and shared its code and weights.

Google plans to continue working with weather forecasting companies and scientists to improve the ability of its model to predict future weather.