An AI trained on human speech and birdsong can suddenly understand the secret language of elephants.
Scientists usually assume that every species has a unique vocal structure that requires its own specialized AI model to decode. This study used acoustic embeddings from humans and birds to classify elephant rumbles with incredible precision. It turns out that the fundamental rules of sound and frequency are universal across the animal kingdom. The AI did not need to know what an elephant was to recognize the patterns in its communication. This suggests there is a shared language of physics that all living things use to talk to each other.
From Birdsong to Rumbles: Classifying Elephant Calls with Out-of-Species Embeddings
arXiv · 2605.00225
We show that pretrained acoustic embeddings classify elephant vocalisations at a level approaching that of end-to-end supervised neural networks, without any fine-tuning of the embedding model. This result is of practical importance because annotated bioacoustic data are scarce and costly to obtain, leaving conventional supervised approaches prone to overfitting and to poor generalisation under domain shift. A broad range of embedding models drawn from general audio, speech, and bioacoustic doma