Meteorology as a science is advancing thanks to technological development. Currently, several computer programs exist that can directly predict when and where it will rain. DeepMind has developed an artificial intelligence capable of predicting almost exactly when and where it will rain. This company has worked with meteorologists in the United Kingdom to create a model that is better at making short-term predictions than existing systems.
In this article, we'll tell you everything you need to know about the Robleda stock market and DeepMind's weather forecasting technology.
weather forecast
DeepMind, a London-based artificial intelligence company , continues its mission to apply deep learning to challenging scientific problems. DeepMind has developed a deep learning tool called DGMR in cooperation with the UK Met Office, which can accurately predict the probability of rain within the next 90 minutes—a significant challenge in weather forecasting.
In a comparison with existing tools, dozens of experts believe that DGMR's predictions are the best on several factors, including its predictions of the location, range, movement and intensity of rain, 89% of the time. DeepMind's new tool opens a new key in biology that scientists have been trying to solve for decades.
However, even small improvements in forecasting are important. Forecasting rainfall, especially heavy rainfall, is crucial for many industries, from outdoor activities to aviation and emergency services. But getting it right is difficult. Determining how much water is in the sky and when and where it will fall depends on many weather processes, such as temperature changes, cloud formation, and wind. All of these factors are complex enough on their own, but they become even more complex when combined. To learn more about this phenomenon, you can consult the article on the definition of meteorology.
The best available prediction technology uses a large number of computer simulations of atmospheric physics. These are suitable for long-term forecasts, but they are not very good at predicting what will happen in the next hour. This is called an immediate forecast.
DeepMind development
Deep learning techniques have been developed before, but these techniques typically perform well in one area, such as predicting location, at the expense of another, such as predicting strength. Radar data for heavy rains that helps predict immediate rainfall remains a major challenge for meteorologists.
The DeepMind team used radar data to train its AI. Many countries and regions frequently publish snapshots of radar measurements that track cloud formation and movement throughout the day. For example, in the UK, new readings are published every five minutes. By piecing these snapshots together, you can create an updated stop-motion video showing how a country's rainfall pattern changes.
The researchers send this data to a deep generation network similar to GAN, which is a trained AI that can generate new data samples that are very similar to the actual data used in training. GAN has been used to generate fake faces, including the fake Rembrandt. In this case, DGMR (which stands for "Generative Deep Rain Model") has learned to generate false radar snapshots that continue the actual measurement sequence.
DeepMind AI Experiments
Shakir Mohamed, who led the research at DeepMind, said this is the same as watching a few frames from a movie and guessing what will happen next. To test this method, the team asked 56 meteorologists from the Bureau of Meteorology (who were not involved in the work) to delve into the more advanced physical simulations and a set of opponents.
89% of people said they prefer the results provided by DGMR. Machine learning algorithms generally attempt to optimize a simple measure of how good their predictions are. However, weather forecasting has many different aspects. Perhaps one prediction got the wrong rainfall intensity in the right place , or another prediction got the right combination of intensities but in the wrong place, and so on. To learn more about how weather systems work, we recommend reading about the difference between anticyclones and cyclones.
DeepMind said it will release the structure of all proteins known to science. The company has used its AlphaFold protein folding artificial intelligence to generate structures for the human proteome, as well as for yeast, fruit flies and mice.
The collaboration between DeepMind and the Met Office is a good example of how working with end users can significantly improve AI development. While this is a good idea, it's not always the case. The team worked on the project for several years, and input from Met Office experts shaped its development. Suman Ravuri, a research scientist at DeepMind, said, "It fosters the development of our model in a way that differs from our own implementation. Otherwise, we might have created a model that wouldn't be particularly useful in the end." For more information on the various applications of meteorology, you can read the article on drones in meteorology.
DeepMind is also eager to show that its AI has practical applications. For Shakir, DGMR and AlphaFold are part of the same story: the company uses their years of experience solving puzzles. Perhaps the most important conclusion here is that DeepMind has finally started listing real-world scientific problems.
Advances in weather forecasting
Weather forecasting must be supported by the development of technology as we are getting closer and closer to fully understanding how our atmosphere works. Many times the human being and his calculations can be subject to common mistakes that can be avoided with the development of artificial intelligence.
Weather forecasting is crucial for humanity because it allows us to use water resources much more efficiently and prevent catastrophes during storms and heavy rainfall . For this reason, meteorologists are increasingly focused on developing artificial intelligence projects for rainfall prediction.
I hope that with this information you can learn more about the DeepMind project and its characteristics.