The Good Algorithm


It was quite an astonishing week for science and technology, in large part because we got to see the first photo of a black hole—something that has fascinated us humans for so long. The photo was a timely reminder of our potential.

From our politicians to our media, it often seems there is faint regard for the possibilities and opportunities presented by technology. We often forget the wonders it can enable. It is not the algorithm itself, but what it is tasked to do, that makes it good, bad, or dangerous. As used by Facebook, the algorithm seems like the work of a devil, but in the hands of Dr. Katie Bouman and her colleagues from MIT, it becomes the eye of an angel.

How did this image come about? Eight radio telescopes across five continents were stitched together to form The Event Horizon Telescope (EHT), an Earth-sized virtual telescope. The EHT uses a technique called very-long-baseline interferometry, or VLBI, which synchronizes telescope facilities around the world and exploits the rotation of Earth to form one huge telescope observing at a wavelength of 1.3mm. It was all done thanks to extremely precise atomic clocks at each of the telescopes.

Bouman’s algorithm then stitches all that data together into a photo. As someone explained on twitter, “First I thought “is the black hole photo really a ‘photo’? Or is it some collected wave frequency data interpreted into a color image” then I realized that’s every photo.” I wonder how and when the lessons from this achievement will impact the future of photography, computer vision, and augmented reality.

As an aside, data for this effort was collected over two weeks, amounting to roughly 350 terabytes per day, or about 5,000 trillion bytes of data on 1,000 high-performance helium-filled hard drives. The data drives were then taken to supercomputers known as correlators at the Max Planck Institute for Radio Astronomy in Bonn, Germany, and MIT’s Haystack Observatory in Westford, Massachusetts. The supercomputers used 140 Nvidia graphics-processing units. And the project also used 20 powerful virtual machines running in two Google data centers, one on the East Coast and one on the West Coast, to serve the EHT members’ computation needs across the globe.

To Dr. Bouman and the members of EHT consortium—more than 200 researchers from 13 stakeholder institutes located in Africa, Asia, Europe, North, and South America—thank you for reminding us of the possibilities.


This first appeared on my April 14, 2019, weekly newsletter. If you like to get this delivered to your inbox, just sign-up here, and I will take care of the rest.

A letter from Om

Sign up & get it delivered to your inbox