What does a black hole really look like?
For the first time, scientists might find out because of a machine-learning algorithm — one that pieces together image data collected from multiple radio telescopes scattered around the globe. If it works, artificial intelligence will have contributed to our first non-artist rendering of a captivating scientific mystery.
No single telescope is up for the job, because none is big enough to produce a high-quality image. And we will never build one, said Katie Bouman, a graduate student at the Massachusetts Institute of Technology working with Harvard researchers on what's called the Event Horizon Telescope.
"A black hole is very, very far away and very compact," Bouman said in an MIT blog post. "It's equivalent to taking an image of a grapefruit on the moon, but with a radio telescope. To image something this small means that we would need a telescope with a 10,000-kilometer diameter, which is not practical, because the diameter of the Earth is not even 13,000 kilometers."
The solution is to use multiple telescopes around the world, joining them together to form a huge single telescope, then merging the results into one picture. But that comes with its own problems: Each participating telescope is a little different from the others, and based on its geographic location, is affected by Earth's atmosphere in different ways. That produces interference that scientists must account for if they want to produce a successful composite image.
The machine-learning algorithm being used by the Event Horizon Telescope — which is really almost a dozen individual telescopes — is designed to do just that. It crunches all the incoming data and tries to narrow or eliminate the differences between readings by comparing them to one another.
The result could give us the first true images of a celestial phenomenon that, for decades, we've left to artists to imagine and describe with pictures.