Before Mark Zuckerberg can hook the entire world on Facebook, he first has to find all the people.
In a Facebook post on Monday night, Zuckerberg drew attention to a company project that does exactly that. Facebook’s Connectivity Lab, which focuses on expanding internet connectivity to remote locations, has developed artificial intelligence to accurately map where in the world people are.
Knowing precisely where people live in remote locations has been a surprisingly difficult task in the past. A fuzzy yellow glow of a general location is all that previous maps have been able to deliver, while Facebook’s new A.I. maps 21.6 million square kilometers (around 8.3 million square miles).
“We will share these maps openly with the community so other organizations can use them too,” Zuckerberg wrote in his post. “This should help with planning energy, health and transport infrastructure, as well as assisting people who need help in disasters.”
And, of course, get more people on the internet so they too can sign up for Facebook.
This new A.I. mapping is a crucial first step for Facebook to take before it sends solar-powered internet drones into the stratosphere to drop wifi signals on us all. Especially when you consider how much of a money drag it would be to deliver a wireless signal to empty desert.
To scour where the people are, Facebook engineer Tobias Tiecke designed a system that automatically recognizes where people are living using existing satellite images — 15.6 billion satellite images to be exact.
The A.I. lab used neural networks to create the maps. In simplified terms, Facebook employees analyzed 8,000 satellite images of India and marked where the people are. Then the A.I. technology took those indicators that showed where people live, and applied it to the 15.6 billion satellite images down to a resolution of 5 meters (around 16 feet). It can do this, Facebook Connectivity Lab engineering director Yael Maguire told Wired, with less than a 10 percent error rate.
Facebook’s new map covers around 20 countries. That, combined with census data, gives Facebook a good idea of population density.
The entire world is now one step closer to being a Facebook click away from each other — which could really put Facebook’s 3.5 degrees of separation theory to the test.