Google GenCast The high-resolution artificial intelligence model, announced as , focuses on weather forecasts. Nowadays, it is of great importance to make weather forecasts with accuracy. Google also developed its new artificial intelligence for this. The first information about Google GenCast was published in the journal Nature.
Google GenCast is similar to the rendering AI models we know, but it is tailored to the Earth’s geometry. Google GenCast, trained with the last 40 years of data from the European weather forecasting center, managed to surpass ENS, the most sensitive forecasting tool of the same center, with a difference of 25 days.
In tests, Google GenCast was trained with weather forecast data up to 2018 and made 1320 different predictions for 2019. These predictions were compared with ENS data. Google GenCast, which is 97.2 percent more accurate in all data, increased this success to 99.8 percent in the last 36-hour predictions.
Aiming to give local authorities more time before a disaster, Google GenCast can also make special forecasts, such as wind speed near wind farms or weather conditions above solar farms.
Google GenCast, running on Google Cloud TPU v5, can produce more than fifty predictions for different possibilities. It is stated that a forecast spread over 15 days is produced in 8 minutes. Normally, this can take hours on supercomputers.
Google GenCast is a public model and Google has started sharing its codes. The company will ensure that the algorithm produces much better forecasts by partnering with weather forecasting agencies and scientists.
This news our mobile application Download using
You can read it whenever you want (even offline):