The sun is the most Sophisticated secrets can soon be solved thanks to artificial intelligence. August 20 IBM and NASA announced the launch of Surya, a Foundation model to the sun. This AI tool, trained in the field of enormous sets of solar activity data, aims to deepen the understanding of clear weather by humanity and to accurately predict clear flashes – explosions of electromagnetic radiation emitted by our star, which threatens both astronauts on orbit infrastructure and communication on earth.
Surya was trained from nine years of data collected by the NASA Solar Dynamics Observatory (SDO), an instrument that has circulated in the sun since 2010, taking high -resolution images every 12 seconds. SDO gives observations of the sun with different lengths of the electromagnetic wave to estimate the temperature of the star layers. He also undertakes precise measurements of the Sun magnetic field – necessary data to understand how energy moves through the star and to predict clear storms.
Historically, the interpretation of this huge amount of diverse and complicated data was a challenge for Heliophysics. To solve this challenge, IBM says The fact that Surya programmers used SDO data to create a digital twin of the sun – a vigorous virtual replica of the star, which is updated after capturing fresh data that can be manipulated and easier to explore.
The process began with the unification of various data formats transferred to the model, enabling them to consistently process. Then a long-range vision transformer was used, which enables detailed analysis of images with very high resolution and identification of the relationship between their components, regardless of their distance.
The performance of the model has been optimized using a mechanism called spectral gate, which reduces memory consumption by up to 5 percent by filtering noise in data, thus increasing the quality of processed information.
More correct forecasts in a shorter time
His programmers say that this project gives Surya a significant advantage: unlike other algorithms that require wide determination of data that they are fed, Surya can learn directly on the basis of raw data. This allows you to quickly adapt to various tasks and provide reliable results in a shorter time.
While testing, Surya showed its versatility in the integration of data from other instruments, such as the Parker Solar probe and the Solar and Heliosphere Observatory (SOHO), two other spacecraft observing the sun. Surya also proved to be effective in various predictive functions, including predicting flame activity and solar wind speed.
According to IBM, time-honored forecasts can only predict a hollow based on signals detected in specific regions of the sun. In contrast, “Surya provided a two-hour advantage using visual information. It is believed that the model is the first to provide this type of warning. In the early test model the team said that it has achieved a 16 % improvement in the accuracy of the classification of sunlight, a significant improvement in relation to existing methods,” said Wawa statement.
NASA emphasizes that although the model has been designed to study heliophysics, its architecture is adapted to various fields, from planetary science to land observation. “When developing a model of foundations, trained in the field of NASA heliophysical data, we facilitate the analysis of the complexity of sunlight with unprecedented speed and precision,” said Kevin Murphy, Director of NASA, WA data statement. “This model enables a broader understanding of how solar activity affects critical systems and technologies on which we all rely here on Earth.”
The risk posed by incorrect solar activity is not low. The main solar storm can directly affect global telecommunications, collapse electrical nets and interfere with GPS navigation, satellite operations, internet connections and radio transmission.
Andrés Muñoz-Jaramillo, a sun physicist at the Southwest Research Institute in San Antonio, Texas and the main scientist of the project, emphasized that Surya’s goal is to maximize the time of the implementation of these possible scenarios. “We want to give the earth the longest possible delivery time. We hope that the model has learned all the critical processes of our star’s evolution in time so that we can bring out useful observations.”
This story originally appeared Wired in Spanish and was translated from Spanish.