A team at NYU Abu Dhabi has built an AI model that can forecast solar wind speeds up to four days in advance, and do it far more accurately than the tools in use today. Take the occurrence in 2022, for instance, when a burst of solar wind dragged 40 Starlink satellites out of orbit. Instead of analysing text like a chatbot, the system ingests high-resolution UV images from NASA’s Solar Dynamics Observatory and pairs them with historical solar wind data. In the researchers' tests, the model delivered a 45% improvement over current operational forecasts, and a 20% gain on earlier AI attempts. Reference of the news:A Multimodal Encoder–Decoder Neural Network for Forecasting Solar Wind Speed at L1, published in The Astrophysical Journal Supplement Series, September 2025.