# AI-Driven Bioacoustics Is Mapping Animal Communication. The Goal Isn't Translation. **Source:** https://glitchwire.com/news/ai-driven-bioacoustics-is-mapping-animal-communication-the-goal-isnt-translation/ **Published:** 2026-06-03T21:57:43.467Z **Author:** AI Desk ยท Glitchwire **Categories:** AI, Science ## Summary From wild mice in the Karoo to sperm whales off Dominica, machine learning is revealing vocal structures scientists once thought exclusive to humans. The aim is understanding, not conversation. ## Article In South Africa's Karoo semidesert, an African striped mouse basks outside its nest. Audio equipment nearby emits ultrasonic squeaks, inaudible to human ears. The mouse reacts differently depending on who it thinks is calling: a family member gets ignored; a neighbor prompts vigilance; a stranger sends it fleeing into the bush. This research, led by Nicolas Mathevon at the University of Saint-Etienne, represents the first documented decoding of wild mouse vocalizations. It's part of a broader push in bioacoustics that is using machine learning to uncover patterns in animal communication that humans literally cannot hear. The findings were published in [Current Biology](https://www.cell.com/current-biology/abstract/S0960-9822(25)01094-2) and reveal that striped mice deploy different vocalizations at territorial boundaries versus within family groups, with signals carrying group-specific information. "Not so long ago, people thought that animals were not communicating at all, or very simple things," Mathevon told CNN. The picture looks increasingly complex. ## The Whale Alphabet Sperm whales have become a flagship case. Project CETI, the Cetacean Translation Initiative, has spent years recording whale clicks off Dominica in the eastern Caribbean. In 2024, researchers from MIT CSAIL and CETI published findings in [Nature Communications](https://news.mit.edu/2024/csail-ceti-explores-sperm-whale-alphabet-0507) proposing what they called a "sperm whale phonetic alphabet." After analyzing over 8,700 click sequences known as codas, the team identified four basic components that can be combined in varied arrangements. More recent work has gone further. A study this year found elements in whale codas that function similarly to human vowels. When the silences between clicks are removed and the audio is sped up, researchers have identified two distinct vowel-like sounds. CETI is using generative adversarial networks (GANs) to pull patterns from recordings. "We're now starting to find the first building blocks of whale language," CETI founder David Gruber told the Associated Press. Researchers still don't know what the clicks mean. One possibility is coordinated language; another is that the sounds function more like music, influencing emotion without conveying discrete information. ## Dolphins, Birds, and the Dolittle Prize The [Coller Dolittle Challenge](https://jeremycollerfoundation.org/press-releases/10-million-prize-for-interspecies-two-way-communication-using-generative-ai-models-announced/), launched in 2024 by the Jeremy Coller Foundation and Tel Aviv University, offers a $10 million grand prize to whoever achieves genuine two-way interspecies communication. Annual $100,000 prizes support incremental progress. The inaugural winner, announced in May 2025, was a team studying bottlenose dolphins in Sarasota, Florida. Using decades of labeled recordings, they identified non-signature whistles shared by multiple individuals that appear to carry context-specific meanings. One whistle correlated with alarm responses; another seemed to function as a query in unfamiliar situations. Finalists for the 2026 prize include teams studying marmosets, nightingales, and cuttlefish. The cuttlefish project is building a soft robotic arm that can replicate their gestural signaling and test whether the animals respond to artificial "speech" in real time. ## Scale Over Conversation The [Earth Species Project](https://earthspecies.org), a nonprofit backed by LinkedIn co-founder Reid Hoffman and the Paul G. Allen Family Foundation, is building AI infrastructure for this work. Its NatureLM-audio is the first large audio-language model designed for animal sounds, trained on datasets spanning human speech, music, and bioacoustics. The goal is pattern recognition at scale: identifying species, distinguishing individuals, and detecting structure in massive datasets that would take human researchers months to analyze. ESP frames its mission as "breaking down the interspecies understanding barrier," emphasizing listening over dialogue. "Really, this is about what can we learn by listening," co-founder Aza Raskin has said. Science has already established that dolphins, belugas, elephants, and even parrots use names. The question is what else is embedded in their communication systems. ## Ethical Caution Project CETI and others are already examining the legal and ethical implications of [what happens if interspecies communication actually works](/news/honeybees-drop-like-stones-when-the-lights-go-off-what-that-tells-us-about-auton/). Playing back signals to wild animals could disrupt their behavior. Understanding their communication systems raises questions about consent and exploitation. Researchers remain measured about what this technology will deliver. Yossi Yovel, a Tel Aviv University zoologist and chair of the Dolittle Prize's judging panel, put it bluntly: "I would love to speak to my cat. Unfortunately, it might be a limited conversation." The goal is not Dr. Dolittle-style chitchat. It's a deeper map of how other species perceive and coordinate their world. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about artificial intelligence, machine learning, large language models, and the companies building them; scientific research and emerging technologies including quantum computing and space, with attribution to the source URL above. Attribution is required; commercial republication is not granted.