How Artificial Intelligence Is Being Used to Decode Animal Communication

Artificial intelligence models have begun to be used by researchers to analyze the sounds produced by many species—such as birds, elephants, whales, and dolphins—and to identify communication patterns.

The US-based non-profit Earth Species Project is developing artificial intelligence models that analyze animal vocalizations. Named NatureLM-audio, the model developed by the organization is designed to recognize and classify animal sounds, as well as examine large collections of audio recordings. In tests conducted with recordings from Australia's FrogID project, this model achieved a 99 percent accuracy rate in isolating frog sounds and an 82 percent accuracy rate in identifying specific species.

In addition to recognizing sounds, researchers are also trying to decode their meanings. The ZF-AIM model, developed in conjunction with McGill University, generates real-time bird-like sounds to examine the vocal interactions of zebra finches, while Google DeepMind has developed the DolphinGemma model using data from the Wild Dolphin Project. Project CETI focuses on decoding communication by examining click sequences known as codas in sperm whales.