Open your phone and scroll back to the least interesting photo in your camera roll. A blurry parking lot, maybe, or the inside of a pocket.<\/p>\n
Now picture an entire folder of nothing but that: thousands of near-identical frames of grass and wind, with the occasional tail disappearing out of the corner of the shot. Your job is to look at every single one and say honestly what you saw. Would you make it past the first hundred before giving up?<\/p>\n
A very large number of strangers didn\u2019t give up. Since 2010, a network of motion-triggered cameras strapped to trees and posts across Serengeti National Park in Tanzania has been firing at anything that moves, day or night, and the flood of photographs it produced has been sorted, frame by frame, by people who signed up for free from wherever they happened to be sitting that evening. The project is called Snapshot Serengeti<\/a>, and the numbers behind it are worth sitting with for a minute.<\/p>\n Two hundred twenty-five cameras and a question nobody could answer fast enough<\/p>\n Researchers from the University of Minnesota<\/a> began setting camera traps across the park in June 2010, starting with 200 of them and expanding the grid to 225 by 2012, the count that held for most of the years since and the one in the title above.<\/p>\n Spread across roughly 1,125 square kilometers of the park, home to more than 40 mammal species according to the study\u2019s own published dataset<\/a>, the cameras were built to answer a question that sounds simple and isn\u2019t: how do lions, hyenas, cheetahs, and a long list of grazing animals actually share that ground without one group crowding out the rest? By the time the first three years of results were formally published, the cameras had captured 1.2 million distinct image sets, and roughly a third of a million of those actually had an animal in frame.<\/p>\n The project outlived the study that made it famous. More than a decade after the first camera clicked on, the Snapshot Serengeti team<\/a> has kept running new rounds of the same survey, adding well over a dozen additional seasons of photos on top of those original three years.<\/p>\n The part no computer could do in 2010<\/p>\n How do you get through a million photographs before the grant runs out? In 2010, the honest answer was that you don\u2019t, not with a research team of a handful of people, and not with the image recognition technology available at the time. \u201cWe needed to rely on the human eye,\u201d said Margaret Kosmala<\/a>, who was a graduate student at the University of Minnesota at the time and is now a postdoctoral fellow at Harvard.<\/p>\n So the team partnered with Zooniverse<\/a>, a platform built for exactly this kind of crowdsourced pattern-matching, and opened the entire photo archive to anyone willing to click through it. No biology degree required, no application process, no plane ticket. Just patience, and a working pair of eyes.<\/p>\n \t\t\t\t\t\tWhat changed in the universe since you were born?<\/p>\n Six sourced measurements. No predictions. No astrology.<\/p>\n What all that sorting actually turned up<\/p>\n At the University of Oxford, Alexandra Swanson<\/a> now works as a postdoctoral fellow, but when this camera grid first went into the ground, she was the University of Minnesota doctoral student running it day to day. Her own read on the project\u2019s size doesn\u2019t undersell it: \u201cThis was the largest camera tracking survey conducted in science to date.\u201d<\/p>\n That scale is what made a specific discovery possible a year later. A 2016 study led by Swanson, published in the journal Ecology and Evolution<\/a>, drawing on the same camera network, expected to find lions, hyenas, and cheetahs carving the landscape into separate territories, the way ecologists usually assume weaker predators avoid stronger ones out of fear. That isn\u2019t quite what the data showed. Instead of avoiding lion territory altogether, the study found that cheetahs are able to use patches of preferred habitat \u201cby avoiding lions on a moment-to-moment basis,\u201d slipping into prime hunting ground for a few hours at a time and clearing out before trouble arrived. It\u2019s a pattern drawn from years of photographs, not a lab experiment, but it held up consistently enough across the dataset to publish.<\/p>\n None of that shows up in a single photograph. It only shows up once thousands of labeled images, tagged with species, time, and location, get stacked against each other across years of data. One photo tells you almost nothing. A few million of them, sorted patiently, tell you something real.<\/p>\n