AILSA CHANG, HOST:
Forest elephants, true to the name, spend their lives hidden in the rainforest, which is a problem if you study them.
PETER WREGE: We basically have no idea what they're doing, how they're using the landscape - all of those kinds of things.
ARI SHAPIRO, HOST:
Peter Wrege is a behavioral ecologist at Cornell University, and he says one way to solve the problem is to eavesdrop on the elephants instead.
(SOUNDBITE OF ELEPHANT TRUMPETING)
CHANG: Wrege leads Cornell's Elephant Listening Project, which uses an array of microphones in the rainforests of Central Africa to record the rumbling and trumpeting of elephants. They pick up other sounds, too, like the chest beats of gorillas. By now he estimates they've gathered a million hours of tape.
SHAPIRO: And he says analyzing that much tape is a beast.
WREGE: Very, very slow, very tedious.
SHAPIRO: Jonathan Gomes Selman agrees. He volunteered on the project as a teenager, hand-picking elephant calls, and he thought there had to be a better way.
CHANG: So he and fellow Stanford grad Nikita Demir trained artificial intelligence to do the job instead. Here's Gomes Selman.
JONATHAN GOMES SELMAN: We feed these models hundreds of examples of both audio clips with and without elephant calls, and then these deep learning models are, basically, over time able to learn specific features that the people training these models don't fully know ourselves.
CHANG: They'll present the model next week at a virtual meeting of the Ecological Society of America.
SHAPIRO: Although Wrege hasn't yet tried the new algorithm, he says it seems faster and more accurate than earlier AI attempts, which gives him and other scientists a better chance to decode the mysteries of elephants' rumbles.
WREGE: This is their language. If we can start understanding that better, we know more what's going on in the forest, where we can't see anything.
CHANG: Because to keep an eye on the forest, you've got to keep an ear on it, too.
(SOUNDBITE OF ELEPHANT TRUMPETING) Transcript provided by NPR, Copyright NPR.