DeepMind has developed an artificial intelligence model that will make it possible to refine weather predictions in the future.
Artificial intelligence has been talked about a lot during the year 2023 because of the emergence of generative AI. But the technology goes far beyond mainstream tools like ChatGPT or Bard, and can be applied to many areas of activity. Specialists in cutting-edge fields are seeing their daily lives turned upside down by the interference of AI, as could soon be the case for meteorologists.
AI 90% more precise than usual models
DeepMind, the Google subsidiary specializing in AI, can be proud. An article published on November 15 in the Science series states that its weather forecast program using Graphcast artificial intelligence could be very useful for refining estimates of the weather.
By comparing this AI model to the best current weather forecasting techniques over several months, scientists were able to establish that Graphcast performed better than the latter in 90% of the cases studied. In certain cases, such as that of Hurricane Lee last September, it can thus know the characteristics such as the speed or trajectory of weather phenomena several days before the instruments used by meteorologists.
More economical technology
An efficiency which is all the more impressive as it requires fewer resources. Indeed, AI, based on data from the last 40 years, gives a result in one minute, using a single computer. Classic weather tools require enormous computing power that only supercomputers can provide.
With artificial intelligence which is, for example, also used to detect icebergs, should we expect in the coming years to see AI become the main technology in all areas of thought where events must be anticipated?
Source : France Info
6