- Scientists hope that Surya extracts ideas from complex magnetic processes of the sun
- Researchers have treated nine years of imaging from the solar dynamic observatory
- Surya has carried out a reported improvement of 16% of the precision of the pushing classification
IBM and NASA presented Surya, the first open source foundation model for solar physics.
IBM says that the AI model, whose name comes from the word Sanskrit for the sun, is formed to predict solar activity such as eruptions and storms that can disrupt satellites, navigation systems and electrical networks.
It has been made available through the front, GitHub and the IBM Terrache library, alongside a collection of data sets called Suryabench.
From land to solar forecasts
The project is recovering from space technology, aviation and communication to future deep space missions.
The prediction of solar time remains a difficult task, since these events come from millions of kilometers on a body whose physics is still partially understood.
“We are on this trip to push the limits of technology with NASA since 2023, offering pioneering pioneering foundation models to acquire an unprecedented understanding of our planet Earth,” said Juan Bernabé-Moreno, director of IBM in charge of scientific collaboration with NASA.
“With Surya, we created the first foundation model to look the sun in the eyes and predict its moods.”
This collaboration follows the previous work of IBM and NASA on models focused on AI for the prediction of land and weather, which led to the development of the Prithvi model which analyzed satellite data to help studies on climate and atmospheric systems.
With Surya, they try something similar for the sun, transforming years of high resolution solar imaging of the NASA solar dynamic observatory into a sort of digital twin.
Scientists hope that the model will allow forecasts that go beyond the question of whether a rocket will occur.
The first reports suggest that Surya can generate high -resolution visual forecasts of rockets up to two hours before occurring, doubling the time of traditional methods.
This would mean additional preparation time for astronauts and critical infrastructure operators on earth.
To build Surya, the researchers treated nine years of images of the solar dynamic observatory, which captures the sun every 12 seconds to several wavelengths.
They used a short short vision transformer with spectral trigger to manage the huge data load.
The model was formed not only to analyze the current conditions, but to deduce what future observations would look like, by testing its precision against real data.
“We want to give the land as long as possible,” said Andrés Muñoz-Jaramillo, solar physicist at the Southwest Research Institute and Main Scientific on the project.
“Our hope is that the model has learned all the critical processes behind the evolution of our star in time so that we can extract usable information.”
Like other models of great language and AI tools, Surya raises questions about the question of whether its results should be dealt with as a discovery or an increase in human expertise.
However, its donors emphasize automation and efficiency, highlighting a claimed improvement of 16% of the precision of the pushing classification.
However, forecasts remain far from certain, because the activity of the sun involves many processes that remain poorly understood.
While Surya is described as a step towards better anticipation of solar threats, researchers take care not to present it as a final response.
Instead, they train it as a bridge that can help scientists work more effectively with massive data.
As with any writer or LLM of AI, its predictions are limited by the data on which it has been formed and the hypotheses integrated into its design.
