Understanding whales on their own terms, without inserting ourselves
For decades, researchers have strained to hear what animals are truly saying to one another — not metaphorically, but literally. Now, artificial intelligence is lending an ear vast enough to process what human attention cannot: hundreds of thousands of vocalizations from carrion crows in Spain and orcas off the Canadian coast, surfacing patterns that reframe our understanding of how other species coordinate their lives. The work is less about bridging the gap between human and animal minds than about learning, at last, to listen without imposing.
- Researchers studying carrion crows found themselves drowning in their own data — six to seven days of continuous audio per recorder, far more than any human team could ever manually review.
- AI models developed by the Earth Species Project have now categorized over 127,000 crow vocalizations, distinguishing species, age, and individual callers even when multiple birds vocalize simultaneously.
- A surprising finding has already emerged: most crow communication is not the loud cawing we recognize, but quiet, close-range murmurs — suggesting the real social fabric of crow life is woven in whispers.
- The same tools are being applied to orca research in Canada, where ocean noise pollution threatens whale communication and where AI can compress months of data alignment into a single hour.
- Scientists are urging caution against the temptation to 'talk back' — the technology is designed to understand animals on their own terms, not to translate them into human conversation.
Vittorio Baglione and Daniela Canestrari have spent decades watching carrion crows in northern Spain, drawn by the birds' unusual social structure: extended family units where siblings and cousins cooperate to raise young and defend nests. Understanding how such coordination works — not just why — became the driving question. They deployed audio recorders and tracking devices across the countryside, only to find themselves overwhelmed. Each recorder captured days of continuous sound, generating far more data than two researchers could ever process by hand.
Since 2024, they have partnered with the Earth Species Project, a U.S. nonprofit building machine learning tools for animal communication research. The AI models can sort crow calls into categories, distinguish adults from juveniles, separate species, and synchronize audio from multiple recorders to identify which bird was calling while another responded. More than 127,000 individual vocalizations have already been catalogued. One early revelation: most crow communication is not the loud cawing familiar to human ears, but soft, low-amplitude murmurs exchanged at close range — suggesting that the real work of social coordination happens quietly, between individuals who know each other well.
The researchers are now building a semantic map that layers audio with video and accelerometer data, aiming to connect specific calls to specific behaviors and begin decoding what crow utterances actually mean.
The same technology is reaching into the ocean. At the Raincoast Conservation Foundation in Canada, researchers are applying AI to orca vocalizations — work with direct conservation stakes, since ocean noise pollution from ships and industry disrupts the acoustic bonds that hold whale families together. Tasks that once took months — isolating vocalizations from continuous recordings, synchronizing field notes with audio and video — now take an hour.
Yet Raincoast's Valeria Vergara offered a measured reminder as public excitement grows around the idea of 'talking to animals.' The goal, she said, is not to insert human meaning into animal exchanges, but to understand other species on their own terms. The technology is a tool for listening — one that may finally make it possible to hear what has always been there, just beyond the reach of human patience.
Vittorio Baglione and Daniela Canestrari have spent decades watching carrion crows in northern Spain, trying to understand what they're saying to each other. The birds are worth the attention. Unlike most species, carrion crows raise their young as extended family units—siblings and cousins pitch in alongside parents to feed chicks and defend nests. This kind of cooperation demands something more than instinct. It requires coordination, negotiation, the constant exchange of information between individuals who know each other well.
"They have a very complex society, and they do very complex things together," Baglione, a professor at the University of León, explained. "It's really coordinated behavior and we have answered why they do it, but we want to know how they coordinate and exchange information." To find out, the researchers deployed audio recorders and small tracking devices across the Spanish countryside, capturing every sound the birds made and monitoring what they did in response. The plan was straightforward. The execution became overwhelming. Each microphone ran for six to seven days before its battery died, and in that time it collected far more audio than two researchers could ever manually review. "We realized we actually have too much data to analyze," Canestrari said.
That's where artificial intelligence entered the picture. Since 2024, Baglione and Canestrari have worked with the Earth Species Project, a U.S.-based nonprofit that builds machine learning tools specifically designed to parse animal communication. The collaboration has yielded something remarkable: AI models that can listen to crow calls and sort them into categories, identifying patterns that would take humans months to spot. The technology has already detected more than 127,000 individual vocalizations. It can tell the difference between adult crows and juveniles. It can separate the calls of one species from another. When multiple birds vocalize at once—which happens constantly in a crow colony—the AI synchronizes data from different recorders so researchers can see exactly which bird was calling while another responded.
What's emerged from this analysis is a portrait of crow communication that contradicts some assumptions. Most of what carrion crows say to each other is quiet—soft, low-amplitude murmurs rather than the loud caws people associate with the species. This suggests that the real work of coordination happens at close range, between individuals who are near enough to hear whispers. Baglione and Canestrari are now building something more ambitious: a semantic map that combines audio recordings with video footage and data from accelerometers, devices that track movement and speed. The goal is to match specific calls to specific behaviors, to begin understanding not just what crows say but what those utterances actually mean.
The same technology is being applied to an entirely different world. At the Raincoast Conservation Foundation in Canada, researchers are using AI to decode the vocalizations of orcas, whales that live in tight family groups and coordinate their hunting and movement through sound. The work carries immediate conservation implications. Ocean noise—from ships, from industrial activity, from human expansion into marine space—interferes with whale communication. Understanding how orcas use their calls to stay connected and coordinate behavior could help scientists measure the damage that noise pollution causes. As part of a pilot project, the Raincoast team and the Earth Species Project are combining drone footage, acoustic recordings, and behavioral observations to link specific whale calls to specific actions and environmental conditions.
The efficiency gains are striking. Valeria Vergara, co-director of Raincoast's cetacean research program, described a task that once consumed months of human labor. "Instead of having to listen to every minute of a recording, this tool can just pick out the vocalizations so that we can then look at the vocalizations and label them," she said. The AI can also synchronize field notes recorded by hand with corresponding audio and video, a process that previously required painstaking manual alignment. "Synthesizing all these data sets would have taken us months in the past, but the tool can do it in an hour," Vergara said. "What it does is it gets the data ready to analyze and quantifies our very qualitative voice notes."
But Vergara offered a note of caution as the public imagination increasingly turns toward the possibility of AI-enabled communication with animals. The goal, she emphasized, is not to speak to whales or to insert human meaning into their exchanges. "I would say let's just be happy understanding whales on their own terms," she said, "without inserting ourselves in the conversation." The technology is a tool for listening, not for talking back. It accelerates the slow work of learning how other species experience their own lives, on their own terms, in their own languages.
Notable Quotes
They have a very complex society, and they do very complex things together. It's really coordinated behavior and we have answered why they do it, but we want to know how they coordinate and exchange information.— Vittorio Baglione, University of León
Let's just be happy understanding whales on their own terms, without inserting ourselves in the conversation.— Valeria Vergara, Raincoast Conservation Foundation
The Hearth Conversation Another angle on the story
Why does it matter that we understand how crows coordinate? They're raising chicks either way.
That's true, but the question is how. Cooperation this complex—where unrelated birds help raise young—requires constant negotiation. Understanding the signals they use tells us something about the sophistication of their social world. It's not just about survival; it's about how they organize themselves.
And the AI part—why couldn't researchers just listen to the recordings themselves?
Volume. A single microphone running for a week generates more audio than a person could listen to in months. The researchers had thousands of recordings. AI doesn't replace listening; it finds the patterns humans would miss because there's simply too much data.
The crow calls are mostly quiet murmurs. Does that change how we should think about them?
It suggests that the real communication happens in intimate spaces, between birds that are close to each other. We tend to think of animal calls as loud broadcasts, but these crows are having conversations. That's a different picture entirely.
What about the orcas? Ocean noise is a known problem. How does understanding their calls help?
If we know which calls coordinate hunting, which maintain family bonds, which signal danger, we can measure what noise pollution actually disrupts. Right now we know noise is bad. Understanding the calls tells us exactly what's being lost.
Vergara warned against trying to communicate with animals using AI. Why the caution?
Because there's a temptation to project human meaning onto animal sounds. The real work is understanding them as they are, not as we want them to be. The technology should serve that humility, not replace it.