Independent AI Method Recovers M87* Black Hole Ring From EHT Data
A CSIRO-led team reanalyzed Event Horizon Telescope data with machine learning and found the same M87* black hole ring. The work appeared in Nature.

An independent team has rebuilt the image of the supermassive black hole M87* from Event Horizon Telescope (EHT) data and found a ring almost identical in size and shape to the one the EHT published in 2019, according to Space.com. The black hole has a mass around 6.5 billion times that of the sun and sits at the center of the galaxy Messier 87, 55 million light-years away.
The team, led by Samuel Lai of Australia’s Commonwealth Scientific and Industrial Research Organisation (CSIRO), published the work on Nov. 11 in the journal Nature.
How the team rebuilt the image
The EHT is a network of telescopes around the world that acts together as one Earth-sized instrument, called an interferometer. Because the telescopes are spread out and look through Earth’s atmosphere, the raw data have gaps and distortions, and the process of turning them into a picture can introduce biases of its own.
Lai and his small Australian group wrote a machine-learning method called GenDIReCT to handle both problems. It extracts the black hole’s structure, filters out atmospheric interference and accounts for the biases that come from combining many telescopes. Lai compared the task to photographing a doughnut on the moon from a backyard.
When the team ran its method on the original EHT dataset, it found a black hole almost identical in size and shape to the historic image. Lai said the model had no idea what a black hole looks like when it processed the M87 data. He called the result the first truly independent validation of the EHT image.
What general relativity predicts
Albert Einstein published general relativity in 1915. The theory says mass curves spacetime, the single entity formed by space and time, and gravity arises from that curvature. Karl Schwarzschild found a solution to the equations that describes a region of extreme curvature, a singularity, surrounded by a boundary beyond which not even light can escape. That boundary is the event horizon.
The theory also predicts what the surroundings look like: a ring of superheated, glowing plasma with the black hole’s shadow falling across it. According to Space.com, scientists were struck by how closely the 2019 image matched those predictions. For background on how black holes form, see this explainer on black holes and the first images.
How sure is this?
The work is peer-reviewed and published in Nature. Space.com does not give the number of observations used, but Laurent Loinard, director of the EHT Collaboration, said the method recovers the ring structure of M87* across several years, in excellent agreement with the collaboration’s published results. He called GenDIReCT “an elegant, entirely independent method.”
The source describes the result as likely to settle questions about how much atmospheric effects and processing methods shaped the final 2019 image. It does not report any disagreement between the new reconstruction and the EHT’s own.
Other uses for the technique
The CSIRO team’s method could also be applied to more down-to-Earth problems, according to Space.com, such as improving medical imaging and mapping earthquakes. The article does not say whether any such projects are planned.
Source: Space.com
Site editor: Nikita Sheikin



