Thanks to artificial intelligence, a new age in weather forecasting has begun. Google DeepMind, the company’s AI research department, is working on a machine learning model that can predict weather properly in seconds rather than hours. This AI model, called GraphCast, claims to surpass 90% of the benchmarks used by the most advanced weather prediction systems in the world.
The process of AI weather forecasting begins with training the model on nearly 40 years’ worth of historical weather data. It took the AI four weeks and thirty-two computers to complete the training. However, once trained, the algorithm can forecast weather up to ten days ahead of time in less than a minute on a single desktop computer. GraphCast outperforms existing weather systems on 90% of 1,380 parameters, showing much better accuracy. It is particularly effective in forecasting catastrophic weather occurrences, such as excessive temperatures and the tracking of tropical storms.
In contrast, traditional weather forecasting relies on supercomputers to perform complex calculations based on data from buoys, satellites, and weather stations. This process is expensive and time-consuming, with the most accurate forecasts taking six hours to produce. Supercomputers use a method known as numerical weather prediction, which requires cracking weather regulating equations. This is a laborious task that can be improved with the use of AI.
AI has superior weather prediction skills because it can find patterns in the data that are difficult to perceive in equations. By utilizing previous meteorological data, AI can make more accurate and affordable weather predictions. In fact, GraphCast’s energy efficiency is around 1,000 times less expensive than traditional weather forecasting techniques, according to the Financial Times.
AI weather forecasting can also aid in the fight against global warming. Google is a partner in the World Economic Forum’s new effort, the AI Governance Alliance, to ensure the appropriate development and implementation of the technology. GraphCast can be used to predict issues related to climate and ecology, energy, agriculture, and human and biological activity. This is crucial in predicting extreme weather events brought on by global warming, which is recognized as one of the top three global dangers for the next two to ten years.
The advancements made by Google DeepMind’s AI model offer a promising future for weather forecasting. With its ability to provide accurate predictions in seconds, AI has the potential to revolutionize the field. As technology continues to evolve, it brings us one step closer to better understanding and preparing for the Earth’s ever-changing weather patterns.
