Researchers using artificial intelligence have made strides in identifying structured communication patterns in species including mice, dolphins, apes, birds, whales, and cuttlefish, raising prospects for a deeper understanding of interspecies signals.
A study published in Current Biology marked the first decoding of hidden vocalizations in wild African striped mice, revealing seven distinct call types that convey individual identity and location. Playback experiments confirmed that mice respond with heightened vigilance to familiar calls.
The findings, reported Wednesday by CNN, align with broader efforts employing machine learning to analyze animal vocalizations at scales previously impossible. Projects such as the Earth Species Project and Project CETI (Cetacean Translation Initiative) apply AI models, including large language models, to bioacoustic data.
Recent Progress and Key Projects
In the mouse study, led by researchers including Nicolas Mathevon, over 122,000 squeaks were recorded and processed through artificial neural networks. The AI identified vocal signatures unique to nests and individuals, representing “static” information about identity that remains consistent over time, per CNN.
Other species show comparable complexity. Chimpanzees combine up to 12 call types into pairs to convey messages such as nest-building instructions. Zebra finches link specific sounds to contexts like hunger or danger. Sperm whales produce codas with phonetic-like structures, including vowel-resembling elements, as documented by Project CETI.
The Earth Species Project collaborates with biologists on species including crows, beluga whales, and elephants, developing tools like NatureLM-audio to classify vocalizations by species, sex, and life stage.
The Coller-Dolittle Prize has recognized efforts in dolphin whistles, cuttlefish arm-waving patterns, and nightingale songs, offering up to $10 million for verifiable two-way communication.
Timeline of Developments
2020: Project CETI launches, focusing on sperm whales.
2023–2025: Advances in AI analysis of whale codas and dolphin models, including Google’s Dolphin Gemma.
2025: Multiple teams advance in the Coller-Dolittle Challenge; mouse and primate studies gain traction.
May 2026: Earth Species Project releases updates on individual animal recognition via AI.
Potential Impacts
Improved decoding could enhance conservation by revealing how animals respond to environmental changes, such as noise pollution affecting whale communication. Legal scholars, including those in collaborations between Project CETI and NYU Law’s More Than Human Life Project, are examining the implications for animal rights and environmental policy.
Experts caution that full translation remains distant. Current AI excels at pattern recognition and correlation with behaviors but does not yet capture meaning or enable conversational exchange. Ethical questions include potential disruption of natural behaviors and privacy-like concerns for wild animals, per the Wild Animal Initiative.
No two-way human-animal dialogue has been achieved. Researchers emphasize listening and understanding before attempting interaction. Claims of imminent “direct communication” with animals exceed current verified capabilities but reflect measurable progress in pattern decoding.
