The first picture of a black hole was captured in 2019. The image of black hole M87 was taken by the Event Horizon Telescope and at the time was an epic moment in astronomy. But now using machine learning AI, researchers just made that very image even sharper.
Category
📚
LearningTranscript
00:00This is the first image of a black hole ever taken, captured only back in 2019.
00:09The image of black hole M87 was taken by the Event Horizon Telescope, and at the time was
00:14an epic moment in astronomy.
00:16But now, using machine learning AI, researchers just made that very image even sharper.
00:21This is the new and improved version of the 2019 picture, upscaled by an algorithm called
00:25PRIMO, or the Principal Component Interferomatic Modeling Program.
00:29Finding images of black holes is difficult for a multitude of reasons, not least of which
00:33is that they gobble up anything that passes over their event horizons, even light.
00:37But even the closest black holes are over a thousand light years away.
00:40In fact, this one, housed in the Messier 87 galaxy, is 55 million light years away.
00:46Though the one advantage astronomers have in finding black holes is their size.
00:49In the case of M87, it's 6.5 billion times the mass of our sun, which is not only how
00:55experts were able to detect it, but also how the machine learning was able to fill in the
00:58gaps and transform the image.
01:00The AI was trained using some 30,000 images of simulated black holes, allowing it to work
01:05out different patterns to find the one that would fit just right for M87.
01:09With one of the astrophysicists working on the project saying about it, quote, if a picture
01:13is worth a thousand words, the data underlying that image have many more stories to tell.
01:17PRIMO will continue to be a critical tool in extracting such insights.