IBM and NASA create first-of-its-kind AI that can accurately predict violent solar flares
IBM and NASA scientists have unveiled a revolutionary artificial intelligence (AI) model that can predict the fierce sun explosions more precisely than ever, giving us a chance to react to a dangerous and disruptive solar activity.
The new model of AI, known as “Surya” (Sanskrit for the Sun), absorbs raw images which are captured by the satellite of the dynamic solar observatory (SDO) – which has been looking directly in the sun for 15 years – and treats them more quickly than humans.
Using this raw data, which IBM representatives said in a statement Researchers have barely scratched the surface of, the fundamental model can predict violent explosions before they occur.
In this way, we can protect astronauts and equipment in space, and even plan disturbances of electrical networks and communication systems on earth.
“We are in this journey to push the limits of technology with NASA since 2023, offering basic models of fundamental pioneers to acquire an unprecedented understanding of our planet Earth”. Juan Bernabé-MorenoThe director of IBM Research Europe for the United Kingdom and Ireland which is in charge of scientific collaboration with NASA, in the press release.
“With Surya, we created the first foundation model to look the sun in the eyes and predict its moods.”
Solar activity has an increasing impact on our lives, the more we ventured into space, and the more we count on technology on earth.
Sunscreen and coronal mass ejections can eliminate satellitesDisturbing the navigation of the airlines, triggering power outages and putting a risk of radiation for astronauts, which makes the prediction precise of solar times.
In relation: AI enters an “unprecedented diet”. Should we stop it – and can we – before it destroys us?
The forecast of storms on earth is notoriously difficult, said scientists, and predicting solar storms is even more difficult. When sunscreen burst in the magnetic field of the sun, it takes eight minutes to this light to reach our eyes – this gap (the eight minutes in which we have no visibility on what happened) means that scientists must be even more advanced.
The Soraya AI model is comparable to separation “Prithvi“The family of models of AI. These models treat gigantic volumes of satellite data to create a more precise representation of the earth in order to better predict its climate and time, as well as the realization of other tasks such as the mapping of deforestation, measuring the impact of the flood and the projection of the effect of extreme heat.
The Soraya model is an open-source system of 360 million parameters designed to learn solar representation at eight Atmospheric imaging assembly (AIA) Canals and five Helose and magnetic imaging (HMI) Products.
AIA is designed to provide different views at the top of the sunshine atmosphere, known as solar corona – taking images that cover 1.3 solar diameters in several wavelengths to improve the understanding of physics behind what we can observe in the atmosphere of the sun. HMI, meanwhile, is an instrument that studies oscillations and the magnetic field on the surface of the sun.
The system can precisely foresee solar dynamics, solar wind and solar eruptions and detect extreme ultraviolet spectra (EUV). Scientists say that Soraya’s new architecture means that it can learn underlying physics behind solar evolution. They described their results in a study downloaded on August 18 arxiv Pre -printed database, which means that it has not yet been evaluated by peers.
“It is a great way to achieve the potential of this data”, ” Kathy ReevesA solar physicist at Harvard – Smithsonian Center for Astrophysics, who was not involved in the study, said in the statement. “Pulling features and events from data petacts is a laborious process and now we can automate it.”
Look directly in the sun
The data of the model come from the SDO, which orbit around the earth, taking images every 12 seconds. These images capture the sun in different wavelength strips to take the temperature of its layers, which vary from 5,500 degrees Celsius to the surface up to 2 million degrees Celsius to the crown.
SDO also captures magnetic activity, with emerging sun spots revealed in white light while other imaging tools check the speed of the bubbles on the surface and follow the torsion of the magnetic lines of the sun.
The researchers trained Soraya by taking a nine -year extract from this data, first harmonizing the different layers – which means that the different types of data are merged to create a more holistic image – then experiment with different AI architectures to treat it.
With Soraya, they challenged sequential images, then predict what SDO would see an hour in the future – by checking these predictions in relation to the real observed images.
The sun also has various oddities that scientists have tried to cocoded in the model – including the fact that the sun turns more quickly to its equator than its poles. However, remarkably, they found that Soraya was more effective in learning these quirks in itself, from data, than by any human entry.
During the tests, the AI model could predict whether an active region was likely to trigger a solar rash an hour before it occurs, and in certain experiences, they obtained predictions in two hours (when led by visual information). This represents an improvement of 16% compared to existing prediction methods, IBM representatives said in the press release.
The team has made the AI opening model, and it is now available on Github And Face – An open source platform that hosts AI models and sets of data. SorayabenchAn organized set of data and benchmarks aimed at helping researchers better understand the behavior of the sun, is also available to freely access.
