Shane Gero, A cetacean biologist from Carleton University in Canada, has spent over 20 years studying the communication of whales. His research reveals that “sperm whales produce specific sounds to identify their families” and even use distinct “dialects” based on their geographical location. While other species, such as dolphins, elephants, and birds, also rely on patterned sounds to communicate, deciphering these intricate nuances poses a significant challenge for humans. This is where artificial intelligence proves invaluable, with its ability to detect and analyze these subtle variations.
A recent article in Nature highlights that, over the past year, several AI-powered studies have made discoveries about animal communication. African elephants and common marmosets, for example, have been found to assign names to their companions. Some machine learning tools are decoding the sound of crows, offering insights into animal self-awareness. So the results is translating animal sounds remains an immense challenge. However, translating animal sounds remains a highly complex task. “It’s a significant assumption to believe we can apply this technology to different species and achieve accurate translations,” explains David Gruber, the founder of the CETI Project (Cetacean Translation Initiative), which is dedicated to studying the acoustic communication of sperm whales.
With support from the CETI Project, Gero has has compiled data on more than 30 sperm whale families in the Caribbean. These whales, which dive as deep as 2,000 meters to forage, use sound clicks known as “codas” to communicate. Each group, led by females, possesses unique diets, behaviors, and dialects. According to Gero, variations in the rhythm of codas create cultural boundaries between clans. Thanks to AI, It has significantly enhanced the analysis of these sounds, enabling researchers to identify entire conversations and subtle variations like “rubatos” and “ornaments,” borrowing terminology from music.
Mickey Pardo, a behavioral ecologist, has shown that African elephants produce distinctive sounds that may function as names. By applying AI models, he determined that these sounds could identify individual elephants with 27.5% accuracy. Similarly, neuroscientist David Omer discovered through experiments that tamarins within the same family share specific acoustic traits when “naming” each other.
The goal of these studies extends beyond understanding animal languages to actively expanding their vocabularies, such as teaching elephants location-related terms. AI tools are not only decoding communication patterns but also revealing the intricate world of animal interactions, offering significant opportunities to advance species conservation.
