Can Artificial Intelligence (AI) help us understand and protect elephants? Speech processing, facial recognition and many forms of forecasting have been transformed as computers learn to scour incoming streams of data for significant patterns. Now, it is showing promise for improving how we study elephant lives, as well as the state of their populations and ecosystems. The ability to identify individual elephants has been key to understanding their rich social world, and to shedding some light on their complex consciousness. But the technique of using ears and tusks to recognise individuals is still only used systematically in a handful of locations, and AI could change that.
Working with WildMe, a group that have created algorithms to recognise individual zebras and humpback whales, we’ve taken the first steps towards training computers to help in this work. We’re also working to transform the way elephants and other wildlife are counted during aerial surveys, alongside our partners Dr Richard Lamprey, Vulcan and the Kenya Wildlife Service. A landmark paper that STE co-authored in Biological Conservation showed just how much observers in hot, cramped cockpits have been missing.
When AIs are able to help analyse the hundreds of thousands of photos, aerial survey looks set to become cheaper as well as more accurate. Movement data analysis in real-time offers another opportunity. With the volume of data streaming in from tracking devices across the continent steadily rising, we have developed algorithms that can remotely detect signs of stress such as increasing nocturnal activity and streaking in near-real time, giving protected area managers a valuable head start. We have no time to lose.
Photo by Robbie Labanowski/Save the Elephants