“We've seen through Facebook, through Twitter, that you can glean incredible volumes of data to reach really powerful insights about human behavior,” says lead author Michael Gil, of the University of Colorado Boulder, UC Santa Cruz, and the National Oceanic and Atmospheric Administration. “Technology is now allowing us to bring that sort of renaissance of big data into nature.”
Gil and his colleagues have amassed that data by essentially turning coral reefs into video games. From above, the camera array watched the fish move around the reef. The researchers placed a waterproofed iPad on the seafloor and played a video of something called a “looming stimulus,” basically a black circle that expands to mimic an approaching predator. “This has been shown to scare the bejesus out of all sorts of critters in lab experiments,” says Gil. As you can see in the GIF above, fish in the wild are no different.
The researchers trained an algorithm to recognize individual fish and then tracked them as they swam about. They borrowed a technique called “ray casting” from early first-person shooters like Wolfenstein, which cast rays from the perspective of the player’s character to render what should be in their view at a given moment. “You can map out what that player ought to be seeing in a three-dimensional virtual environment,” says Gil. “We can repurpose the same technology, the same idea—but we can do it with fish.”