Artificial intelligence will revolutionize weather forecasting, Google claims. A new technology created by Google Deepmind makes it possible to create a 10-day weather forecast in just one minute, with unprecedented accuracy, Deepmind representatives said. The predictions produced by GraphCast are not only more accurate, but also more efficient, which means they can be made with smaller data sources. GraphCast also makes it possible to detect extreme weather events, as it is able to predict the movement of cyclones and provide early warning of possible floods or extreme temperatures. Google therefore claims that it could help save lives by letting people know in time about impending extreme weather such as flash floods, cyclones, hurricanes and much more.
Currently, forecasts rely on a system called Numerical Weather Prediction, which combines physical equations and computer algorithms. These systems require supercomputers for their calculations with extreme demands on computing resources as well as energy resources. It is also necessary for them to be operated by experts from a number of metrologists, as equations often have to be entered manually. The new system, on the other hand, can make a weather forecast without requiring a user with any knowledge of metrology to operate it.
GraphCast does not use physical equations to determine the weather, but instead learns from past weather data and models the forecast based on how the weather changes over time and the current data it is fed. Building the model itself that GraphCas uses was challenging because the AI had to learn weather data from the entire planet over the past decades. However, the system is now able to make predictions with excellent accuracy. A standard desktop computer can produce a weather forecast 10 days ahead in a single minute with an accuracy greater than current systems that require many hours of supercomputer power to produce a forecast. GraphCast provides more accurate weather forecast information than currently used systems 90% of the time.
For example, in the heat of this year, the system was able to predict the exact path of Hurricane Lee nine days before its arrival, while the classic system could only do it six days before the arrival of the hurricane. Those three days for the preparation of residents can be quite crucial. Google will soon open up GraphCast to all users for weather forecasting and climate crisis research. If you're interested in this topic, you can join the GrapCast now on Github right here.
The endurance of iPhones is talked about every year and it seems to me that it is still the same.
So maybe they didn't contact our meteorologists. 😁