FOR EACH OF the past three years South Africa has lost more than 1,000 rhinos to poachers, despite intensive efforts to protect them using armed rangers, drones and specially trained tracker dogs. Guarding rhinos is particularly difficult because they roam across vast areas of veld where poachers can hide easily. But two novel approaches using artificial intelligence may help rangers catch their hunters. The first was developed by a group of computer scientists who had previously used artificial intelligence to detect roadside bombers in Iraq and insurgents in Afghanistan. In South Africa they used machine learning to predict where rhinos were most likely to be feeding the following day. The computers also crunched historical data on poaching incidents to identify areas where they were likely to happen. Rangers and drones could then be sent to patrol in areas most likely to have both poachers and rhinos, says V.S. Subrahmanian, who worked on the project at the University of Maryland.

Scientists from IBM used a different approach to protect rhinos in the Welgevonden game reserve near Johannesburg in South Africa, using radio collars and a data network. The collars they have developed can beam back the position and the speed at which the animals wearing them are moving. Instead of attaching these to the rhinos, they are collaring other animals such as impala or zebra that normally move alongside rhinos. The data being sent back are constantly monitored by computers for specific patterns of movement. If a natural predator gets into a herd of zebras, they will scatter. But if a man with a gun approaches, the animals will all start to move in the opposite direction, providing the rangers with an early warning of intruders. So instead of relying on just a dozen rangers to keep an eye on things, the park benefits from hundreds of four-legged sentinels, each one connected to a computer in the cloud.