Compared to traditional models, SEEDS can predict severe weather faster and more efficiently. Image Source: Fun Science Website
Tech Daily News (Reporter: Liu Xia) - Google recently unveiled a new generative artificial intelligence (AI) model called Scalable Ensemble Enveloping Diffusion Sampler (SEEDS). The company claims that SEEDS can provide more accurate weather forecasts at a lower cost than traditional methods, and can detect extreme weather events that are difficult to discover. The related paper was published in the latest issue of "Science Advances" magazine.
Google stated that the SEEDS model is similar to popular large language models like ChatGPT and generative AI tools like Sora. Compared to traditional forecasting models, SEEDS can generate more weather scenarios faster.
Current meteorological forecasting methods often struggle to anticipate the consequences of extreme weather events and currently use probabilistic forecasting models, treating random variables as initial conditions, leading to a rapid increase in error rates, making it difficult to accurately predict future extreme weather.
The research team pointed out that SEEDS can infer up to 31 weather scenarios based on only 1-2 input data. Researchers modeled the 2022 European heatwave to test the system's performance. The results showed that seven days before the heatwave outbreak, the forecast data from the US meteorological agency failed to predict the event, while SEEDS successfully did.
Google stated that the computational cost of using SEEDS is extremely low compared to existing methods. In a sample of Google Cloud architecture, the system can provide 256 prediction results every 3 minutes.
