Smithsonian Magazine · People & society

AI Helps Decode Elephant Group Vocalizations

Scientists are using artificial intelligence to analyze the complex rumbles of African elephants, suggesting these joint vocalizations convey coordinated group messages.

A new study published in Scientific Reports used machine learning to analyze 902 recordings of wild African savanna elephant rumbles from Kenya, collected between 1986 and 2022. The analysis focused on rumbles made during six social contexts, including communication between mothers and calves, separated individuals, and during group movements and reunions.

The AI algorithm could predict elephant activities based on their rumbles better than chance, indicating that chorused vocalizations contain information about who is calling and what they are doing. The findings suggest choruses may offer a more detailed layer of information than individual rumbles alone.

The study observed a phenomenon called "vocal convergence," where individual rumbles within a chorus became acoustically similar in certain situations, similar to how humans adopt speech patterns in conversation. This convergence was noted during "cadenced," "contact," and "greeting" rumbles.

Researchers hypothesize that vocal convergence might aid in coordinating group behaviors, but they emphasize that further research is needed to confirm this. The study also notes that African savanna elephants may potentially call each other by name.

The application of AI in decoding animal communication is expanding, with similar tools being used to study mosquitoes, sperm whales, marmoset monkeys, and crows. Experts believe AI can significantly enhance the analysis of collected animal communication data if used responsibly.

AI-samenvatting op basis van de bron.