On April 10, 2019, the Event Horizon Telescope (EHT) collaboration released the first image of a black hole: the supermassive object at the center of the galaxy M87, shown as a bright ring around a dark shadow. The EHT is not a single dish but a network of radio telescopes scattered across the planet, combined through a technique called very long baseline interferometry (VLBI) to act as a single Earth-sized instrument. The results were published the same day as a series of six papers in The Astrophysical Journal Letters.
The image was a feat of computation, not photography. VLBI data is extremely sparse and noisy, and a vast number of different images could fit the same measurements. The fourth EHT paper, “Imaging the Central Supermassive Black Hole,” describes how teams used image-reconstruction algorithms to find the most plausible picture consistent with the data. To guard against fooling themselves, separate teams reconstructed the image independently using different methods, and the group ran tests to make sure their pipelines did not simply impose a ring where none existed.
It is worth being precise about the methods: the headline reconstruction relied on regularized maximum-likelihood and other computational-imaging techniques rather than the deep neural networks most people associate with modern AI. The work built on years of algorithm development, including the CHIRP method by Katherine Bouman and colleagues, and stands as a landmark example of extracting an image from data that no telescope could capture directly.