Early tests of Surya, an AI model developed by IBM, demonstrated the ability to predict solar flares up to two hours in advance. According to Juan Bernabe-Moreno, the project lead, Surya can assess not only the intensity and shape of the solar flares but also their location on the sun. While a two-hour warning might be insufficient to mitigate all potential threats from a significant flare, IBM claims this could potentially double the warning time compared to current methodologies. The exact lead times reported, however, may differ.
The underlying processes that lead to solar flares remain complex and not fully understood. Astrophysicist Harra noted that while the conditions for solar flare events are known, the exact timing and triggering factors are elusive. The effectiveness of Surya may hinge on its capacity to identify and analyze these destabilizing patterns more rapidly than existing approaches.
Beyond predicting solar flares, Bernabe-Moreno envisions Surya being used in conjunction with other models developed for IBM and NASA that analyze terrestrial weather patterns. He raises questions about the potential interconnections between solar activity and events on Earth, such as lightning. There is interest in understanding how solar phenomena may influence terrestrial weather systems.
Surya is considered a foundational model, meaning its design allows for versatile applications in various areas of solar physics. Similar to how generalized language models function, NASA and IBM anticipate that Surya could discover significant patterns pertaining not only to the sun but potentially to other celestial bodies as well. Bernabe-Moreno views the sun as an essential reference point for studying other stars in the universe.
Source: https://www.technologyreview.com/2025/08/20/1122163/nasa-ibm-ai-predict-solar-storm/