IBM and NASA built an artificial intelligence model to predict solar flames

Using the Surya model, the first open source of artificial intelligence in the field of solar physics and introduced by IBM and NASA’s Space Agency, scientists are seeking information on complex sun magnetic trends. IBM has announced that this model of artificial intelligence, derived from the word Suncrete, is trained to predict solar activities such as storms and storms, as these can disrupt satellite, navigation and electrical networks.

The model is now available to the public through the Haging Face and Gate Habe platforms, as well as the IBM Terratorch Library, and a set of data called Suryabench. The project is introduced in a situation where space technologies are expanding in areas such as aviation, communications and future missions in the depths of the space. The forecast of solar climate is a difficult challenge due to the origin of these phenomena, which is millions of kilometers away from the Earth and their physical nature is not yet fully known.

“We’ve been on this route for the past year to move the borders of technology with NASA and present a pioneering fundamental artificial intelligence models for an unprecedented understanding of the planet’s planet,” said Juan Brent. “With Surya, we have built the first fundamental model to look directly on the sun and can predict its moods,” he said. This collaboration is the continuation of IBM and NASA’s previous activities in the field of artificial intelligence -based models to predict the state and weather status that led to the development of the Perito model. “This model helped study climate and climate systems by analyzing satellite data.”

With the Surya model, they are trying to do the same thing for the sun, namely turning years of high -resolution images from NASA solar dynamics to a digital simulation model. Scientists hope that this model will allow predictions beyond or not to occur. According to preliminary reports, Syria is able to produce high -resolution visual predictions of the phenomena up to 4 hours before their occurrence, which doubles the alert time compared to traditional methods. This means more time to prepare for astronauts and vital infrastructure on earth.

To build Syria, researchers processed the solar dynamics observatory for five years, which depicts the sun every 2 seconds at several wavelengths. They used a visual converter and spectral analysis to manage the huge volume of data. This model was evaluated not only for analyzing the current conditions but also for the inference of how the future was observed and accurately measured by real data.

“We want to give the longest possible warning time to the ground,” said Anders Moniuses-Khararamilo, a solar physicist at the Southest Research Institute and the main scientist of the project. “We hope that this model will learn all the critical trends behind our star transformation over time so that we can extract usable insights.”

Like other large language models and artificial intelligence tools, Syria raises the question of whether its outputs should be considered as a new exploration or as a complement to human expertise. However, its supporters emphasize automation and productivity, pointing to a 2 % allegation in accuracy of the classification of the cornerstone. However, forecasts are not yet certain, because the activity of the sun contains numerous processes that are not yet well understood.

While Syria is described as a step towards a better prediction of solar threats, researchers are careful not to introduce it as a final response. Instead, they consider it a bridge that can help scientists work more effectively with bulk data. Like any author of artificial intelligence or a large linguistic model, its predictions are limited by the data trained and the assumptions included in its design.