Google It: Search giant Says it's Better At Forecasts Than the Weatherman (and Woman)

Facebook Tweet LinkedIn

Google has announced that its DeepMind software, which uses AI for weather forecasts, has bested humans and their current forecasting technology.

In a press release, the company boasted that the “New AI model advances the prediction of weather uncertainties and risks, delivering faster, more accurate forecasts up to 15 days ahead.”

The claim is based on an article in the journal Nature titled “Probabilistic weather forecasting with machine learning.” The authors wrote that “Recent advances in machine learning (ML)-based weather prediction (MLWP) have produced ML-based models with less forecast error than single NWP simulation.” The weather prediction (NWP) is an older system centered on physics.

Google’s GenCast weather forecast system examines over 50 predictions. Each prediction is about the direction and severity of a weather system. This is part of the creation of the GenCast system, which has examined 40 years of weather data.

On target
Ultimately, the difference between the physics system and GenCast was substantial, the researchers said. “We comprehensively tested both systems, looking at forecasts of different variables at different lead times — 1320 combinations in total. GenCast was more accurate than ENS on 97.2% of these targets, and on 99.8% at lead times greater than 36 hours.”

Weather prediction has entered a new age. It will likely be safer for people because it gives them more time to prepare for disasters. It will also put a large number of weather people out of work.