The artificial intelligence of Google, DeepMind, he learned a predict the rain before it comes. The experiments carried out with the UK Meteorological Service allow you to predict if it will rain in the next 90 minutes with great precision. “Hey Google, do I have to bring an umbrella?”
Google, DeepMind’s artificial intelligence to predict rain
The research team led by Shakir Mohamed applied the DeepMind supercomputer AI to weather forecasts. Google’s algorithm has put itself to the test in what is called “nowcasting“, Ie the forecasts in a given place in the next two hours. An interesting challenge because supercomputers usually aggregate a lot of data to make predictions over the very long term. When instead surgical precision is needed and the data is less, the situation becomes complex. Especially for thunderstorms and storms, conditioned by many variations.
To solve the problem, Google researchers applied a deep learning network called Deep Generative Model of Rainfal (DGMR). As for macro forecasts, he analyzed the data over time to build a model, then verifying it with the forecasts made with more “classical” methods. This allowed the computer to learn from its mistakes and refine its predictions.
The result achieved greatly pleased the researchers. They predicted the weather in the next 90 minutes of the forecast. A team of 56 meteorologists then compared the data with traditional methods. The result is that 89% of them preferred DGMR data, because more reliable.
Google’s intelligence could therefore take meteorology one step further, not only for calculations of seasonal or epoch-making changes, but even for the most immediate time turns. In short, the AI can tell us with great precision if it is convenient to bring an umbrella when we leave the house. Hoping it doesn’t start raining after 91 minutes.