Image of Sagittarius A*, black hole in the center of the Milky Way galaxy
eht
The heart of our Galaxy may include an exceptional space rotating top. This is a black hole that appears to be spinning as fast as possible.
Michael Jansen Ladboud University in the Netherlands and his colleagues were studying black holes in the Milky Way center, Sagittarius A*, using data collected by a network of collectively known as the Event Horizon Telescope (EHT). To address the complexity of data, they turned to artificial intelligence.
First, they simulated about a million black holes using well-known mathematical models. This was a calculation feat that in itself required millions of hours of supercomputer time. These simulations were then used to train a type of AI called neural networks, allowing us to determine the properties of black holes based on observed data. Finally, they gave AI the data on Sagittarius A* that EHT collected throughout 2017.
The AI showed that Sagittarius A* was spinning at 80-90% of the highest possible speed. He also warned the researchers that none of the magnetic field models are particularly suitable for black holes. That’s why more mathematical work is needed. Janssen says past research narrowed down the range of traits Sagittarius A* could have. For example, how fast it spins, what magnetic fields surround it, but this new approach has fixed them more accurately.
Dimitrios Psortis Georgia Tech in Atlanta says they found some of these findings counterintuitive. Previous analysis did not even clarify whether black hole spins could be accurately determined from EHT data, he says.
Some previous works have shown that Sagittarius A* could be spinning very quickly, he says. Mizuno Yuishi At Ziaoton University, Shanghai, China. However, he points out that there is room for improvement in the computer models used in new research. “Our theoretical model is not perfect yet,” he says.
But both Mizuno and the Poetry say that they are becoming an integral part of how AI learns about exotic space objects like black holes. “We have a lot of data, a lot of models and we need a modern way of combining the two,” Psaltis says. “This is where machine learning makes a huge difference.”
At the same time, this brings unique challenges. This is because AI work needs to be double-checked and proven analysis for subsequent hallucinations.
Janssen and his team have already performed many such checks, including testing AI using specially designed simulation data. Analyzing data from other years’ EHT operations and ultimately analyzing data from new observatory results in more testing, he says.
topic:
