Google Deep Mind and Google Research started a public preview of the Weather Lab on Thursday. It is an interactive website where the company will share its artificial intelligence (AI) seasonal models and share weather forecasts based on production. Mountain View -based Tech Dev has also released a model of his modern experimental AI -based tropical storm. It is said that this model is capable of forming a storm, track, intensity, size and shape. In particular, the company says the scientific verification of the AI model is currently pending.
Google releases a new AI model for storm forecast
In a blog post, the Deep Mind announced the launch of the new Weather Lab website and detailed its new -based AI model. The website shows direct and historical storms, using both the EI seasonal models and physics -based models of the European Center for Medium -sized weather (ECMWF).
Google Deep Mind highlights that on the website, several AI models, such as a veterinary graph, Visner text general, and a new storm model, run in real time to analyze and predict weather data. In addition, the Weather Lab also includes more than two years of historic AI-generation predictions that researchers can download to evaluate the performance of the model.
Weather Lab also allows users to compare the predictions of various AI and physics -based models. In particular, the company emphasizes that the website is a research device and is not intended to provide official alerts.
Coming into the new AI -based storm model, Google has published the pre -print version of its dissertation. However, this has not yet been reviewed. Google, Google, has contributed with the US National Hurricane Center (NHC) for the Scientific Confirmation of the Research Community.
Deep Mind says that in the prediction of the traditional storm, two different physics -based models are used. A global low resolution model predicts storm tracks, which require environmental steering streams to analyze, while the regional high resolution model is used to track the severity of the storm, which requires a complicated process inside and around its compact core.
It is said that the new AI model will solve this dual approach by uniting both the storm track and the intensity forecast. According to the post, this model is trained on both “Realis Dataste”, which creates past weather over millions of observations, and a special database that has the key to about 5,000 observed hurricanes over the past 45 years about the track, intensity, size and air Ready. “
Highlighting an example, Deep Mind said that the model was deployed in the North Atlantic and the Eastern Pacific, between 2023-24, for examination, and during this time, its five-day storm track predicted, on average, the ECMWF’s ESSS model was close to 140 km. In addition, the company has claimed that the results of the internal testing storm model are at least equal to the physics -based model.
