We thought we had already observed Sagittarius a*, the supermassive black hole in the center of our galaxy. In 2022, the Event Horizon (EHT) telescope delivered a first image – or rather a reconstruction from radio waves captured by antennas distributed all over the earth. But these shots were far from having said everything.
A black hole better surrounded by AI
An international group of astronomers wondered: what if we plunged into the data? Except that instead of resuming the old methods, they used a well -trained AI. Thanks to millions of simulations created by Superordinators, the algorithm learned to identify details that humans had not seen. And surprise: the black hole turns almost at its maximum, and its axis is (roughly speaking) directed towards the earth.
That’s not all. The AI also made it possible to review what we thought about the environment of Sagittarius A*. For example, light emissions around him would not come from a jet of matter, as we thought, but rather from ultra-dirty electrons present in the accretion disc. And the magnetic fields in this area seem to behave differently from what the classic models provide.
So far, EHT researchers worked with a few hundred simulations. This time, thanks to the power of the Calculation Center of the University of Wisconsin (CHTC), they were able to push up to several million. It has nourished a neural network called “Bayesian”, capable of estimating uncertainties and comparing tons of data in the blink of a computer eye.
« We are clearly shaking up the received ideas “Explains Michael Janssen, researcher at Radboud University, in the Netherlands. But for him, this is only the beginning: ” Our approach with AI is just the first step. We will continue to improve it. The project requested a hell of a organization: more than 12 million calculation tasks were distributed over servers for three years. “” This kind of job is right in our strings », Had Miron Livny, director of the center.
What to remember? Artificial intelligence will not replace astronomers, but it can help them dig where no one thought they can find anything. Even in data old years of several years, it remains obviously full of secrets to discover.
🟣 To not miss any news on the Geek newspaper, subscribe to Google News and on our WhatsApp. And if you love us, .
