For AI company DeepMind, being present in the media is nothing new. They have frequently demonstrated unique developments in artificial intelligence, but with their latest development they aim to revolutionize an area that has been derided for decades for its inaccuracy – the weather.

Working with the Met Office, the UK’s weather service, they have developed the Deep Generative Model Of Rainfall (DGMR). DGMR tackles one of the biggest challenges in weather forecasting, determining the probability of rain. According to DeepMind and Met Office, the new AI tool is able to determine the probability of rain in the next 90 minutes with extreme accuracy.

 

DGMR by DeepMind in action (Source: DeepMind)
DGMR by DeepMind in action (Source: DeepMind)

Compared with previously used tools, the experts said, the prediction of the location, extent and intensity of rain is correct nearly 90 percent of the time.

Looking at the current news of floods and chaos, a tool like DeepMind’s is incredibly helpful. In particular, reliable forecasts of heavy rainfall can make it possible to take early precautions and reschedule. But it’s not just in disasters that the AI-powered solution offers many benefits. From aviation to outdoor event planning, reliability can save a lot of money and energy.

DeepMind’s team of experts trained their artificial intelligence using radar data, which is published regularly by many countries. Here, the movement of clouds and similar factors can be seen, which helped train the model. The AI then generated its own simulations from the training data, which simulatively continued the sequence of actual measurements.

The collaboration between DeepMind and the Met Office is a great example of how useful it can be to develop new technologies, like AI, directly with the end user and create a solution together. This way, each party can benefit from the expertise of the other and a truly useful and long-term tool, like DGMR, is created. Other industries and companies can therefore take a lot from this example and, after an initial proof-of-technology, as the check whether AI is really currently relevant for their own area, quickly develop new concepts that can both relieve employees and bring real benefits to the company.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.