By Googling, today’s forecast could soon be even more reliable.

Just like your friendly TV meteorologist, current weather models may be a thing of the past. Google has unveiled AI meteorology technology that is far faster and more accurate than traditional forecasts, according to a study published in the journal Nature.

The “GenCast” model developed by DeepMind, the search engine company’s AI division, can detect whether it will rain 15 days in advance with a higher accuracy than the European Center for Medium-Range Weather Forecasts’ ENS (ECMWF) – according to Google Deepmind -Blog it is the world’s leading operational forecasting system.

This discrepancy has to do with a completely new method of monsoon divination. While current iterations “are deterministic and provided a single, best estimate of future weather,” GenCast includes “an ensemble of 50 or more forecasts, each representing a possible weather trajectory,” the blog’s authors write.

This large amount of data allows forecasters to predict the weather with far greater accuracy, specifically with an accuracy between 97.2% and 99.8% depending on the circumstances.

“Such ensemble forecasts are more useful than relying on a single forecast because they provide decision makers with a more comprehensive picture of possible weather conditions in the coming days and weeks and the likelihood of each scenario,” the blog says.

GenCast also uses a type of AI called a diffusion model, which is typically found in video, image and music generators.

Unlike most versions, this cybernetic weatherman has been adapted to Earth’s spherical geometry and is based on four decades of historical weather data (up to 2018) from the ECMWF, ranging from temperature to pressure at various altitudes.

To evaluate the effectiveness of the typhoon detector, researchers compared GenCast’s forecasts with real weather data from 2019 and ENS forecasts for this year.

They specifically examined 1,320 combinations of different variables at different lead times.

The AI ​​hardness test found that GenCast was more accurate than ENS on 97.2% of these targets, and 99.8% when lead times were greater than 36 hours.

This true Deep Blue of weather detection was also far more efficient.

A single Google Cloud processor reportedly takes just eight minutes to produce a 15-day forecast in the GenCast ensemble, compared to the hours required to produce physics-based ensemble forecasts – such as those produced by ENS to create a supercomputer with tens of thousands of processors.

Additionally, GenCast could better predict extreme weather events—extreme heat and cold, as well as high wind speeds—which could help meteorologists better keep track of hurricanes and typhoons.

The only downside is that the current ENS system can generate significantly higher resolution forecasts than its AI counterpart, the Daily Mail reported.

DeepMind representatives also admitted that the current meteorology machines are irreplaceable for now, as they provide the data used to train models like GenCast.