AI Searches for Patterns in Animal Sounds
Thursday, 2026/09/10171 words3 minutes746 reads
Researchers are using AI to look for structure in animal communication, not to claim that they can already translate it. Traditional research often requires people to sort recordings by hand. AI can process much larger collections and identify repeated calls, individual animals and links between sounds and behaviour.
Earth Species Project has built NatureLM-audio, an audio-language model for bioacoustics. It was trained with animal recordings, human speech and music. The project says the model can support species classification and transfer knowledge to species it did not meet during training.
Other studies show why the question is interesting. Project CETI analysed 8,719 sperm-whale click sequences and described a richer combinational structure than previously recognised. Features were compared with rhythm, tempo, ornamentation and rubato. That does not reveal what the whales are saying.
The limits matter. Animal systems work differently from human language. Before sending artificial calls to wild animals, researchers need to consider possible effects on fear, feeding, mating or movement. For now, AI is mainly a tool for listening more carefully.
