For decades, though, the radio-collar was the state of the art in conservation tech. All over the world scientists followed the beeps, painstakingly piecing together the stories of individual animals through their daily comings and goings. It was a process that required first darting and drugging an animal, then taking samples of blood, hair and whatever else, then cinching a heavy leather collar with a small metal box around its neck. There were always risks that the trauma of the experience could prove fatal. Even when everything went according to plan the battery that powered the electronics would eventually run down and the animal would need to be darted again in order to replace it.
Some field biologists, most famously Jane Goodall, would instead embed themselves within a wildlife community to better observe family dynamics up close. Ginger Kathrens (whom I met while filming a National Geographic segment on 9/11) has been recording the lives of the wild horses on the Wyoming / Montana border for well over a decade. The latest Sir David Attenborough series for the BBC, Dynasties, tells the stories of five species (lions, tigers, chimps, penguins and painted wolves) filmed over a two-year span. Sometimes collars are used as part of these efforts.
Both approaches have generated significant insights, but require sustained funding over a long period of time and a team of very special people willing to devote their lives to the work. They simply aren't scalable, nor are they practical for most species. Also, extrapolating data from such limited population samples can be misleading.
FORGET THE BEEPS. FOLLOW THE PIXELS...
So was with great interest that I crashed a lecture at Google's Chicago office seven years ago to see Tanya Berger-Wolf, a computer science professor at the University of Illinois at Chicago, talk about using photos and algorithms for conservation. In a roomful of techies focused on the digital details of the how the newly open-sourced Stripe Spotter software worked, I sat dazzled by the implications. Stripe Spotter was the winning solution to a challenge that Berger-Wolf and her colleague, Princeton evolutionary biologist Dan Rubenstein, had given to a group of students on a trip to Kenya: to develop an algorithm to identify an individual zebra in a photograph by its unique pattern of stripes. For the first time it was now possible to see the zebras and the herd, to use visual data to better understand the dynamics of a wildlife population.
Since then Stripe Spotter has gone from a rarefied moment of fame (a question on NPR's news quiz, Wait Wait, Don't Tell Me) to a robust, global platform called Wildbook that has already had a profound impact on conservation research and policy. A vast and exponentially-expanding wildlife database has been created from hundreds of thousands of photographs and videos submitted by citizen-scientists, along with those captured by in-the-field camera-traps, and also scraped from websites such as YouTube and Flickr. The entire process has been automated, from uploading the digital imagery, to determining an animal's species and whether it is already in the database, to noting where and when a photo or video was taken. It's not just about spotting stripes either. An animal's spots, wrinkles and notches are as unique as a fingerprint.
With funding from Microsoft under the umbrella of AI for Earth, the Wildbook platform will grow from a database of dozens to thousands of species with a few years. It is now possible to take data gathered in near real-time and use it to model the impacts on wildlife of weather, climate, the proximity of predators, the availably of prey, the spread of diseases, pollution, urban sprawl and other variables.
In 2016 and again in 2018 as a field test proof-of-concept, the platform was used to take a population count of Grevy's zebras in Kenya. Instead of the typical guesstimate with a ± margin of error of thousands of animals, the margin of error for the Wildbook count was just a few dozen zebras. Teams of citizen scientists—including school kids from Kibera, one of the poorest neighborhoods in Nairobi—fanned out to cover a 25,000 km area, using GPS-enabled cameras to take 48,000 photos in 2016 and 70,000 in 2018.
The data were such a significant improvement that the IUCN is now using it for its Red List. The data are also being used to help shape conservation policy. To keep a better balance between predators and prey in one section of the range, wildlife managers in Kenya are considering using birth control on lions.
Meanwhile far from the African savanna, Wildlbook has also been used to identify nearly 8,000 individual whale sharks swimming the world's oceans. Before Wildbook the tally was 70. Not only has the platform provided a way to aggregate and analyze data, and also for researchers to connect and collaborate.
BLESSING AND CURSE
Long before the Cambridge Analytica / Facebook scandal brought the issue of data privacy into the headlines, the team at Wildbook began worrying about the privacy of wildlife, particularly species that have been hunted to near extinction. When such a vast trove of data can be broken down to the specificity of an individual in the blink of an AI's (artificial) eye, it can become a powerful tool to fight wildlife crime, or a poacher's handbook. No tiger ever put a paw-print to an electronic release form. So far the platform is secure, but privacy, along with digital fakery, is an ongoing concern.
According to a recent IPCC report, there is only about a decade left to get global carbon emissions under control before what is now called "extreme weather" becomes simply weather and seas seep into every coastal city. Against that stark backdrop, both the mission and the urgency of conservation have changed. The baseline for normal is shifting, as are entire habitats. The chemical composition of the air and water have been dramatically altered, while agriculture and urbanization have almost completely reshaped the landscape.
At the same time that populations of mammals, birds, fish and reptiles have declined by an estimated 60% over the last half century, the human population has doubled to nearly eight billion. A staggering 26,000 species are at risk of extinction, according to the IUCN, with insects speeding toward an "insect apocalypse," — and the sobering possibility that they could take us with them.
Wildbook has come along in the nick of time. The combination of artificial intelligence, machine-learning and crowd-sourced citizen science have given conservation a powerful and nimble new tool. Time will tell whether it will be used to develop strategies that help protect species and restore ecosystems, or ends up giving us a deeper insight into all that has been lost.
That part of the story depends on the human algorithm.