On a calm stretch of water in Shark Bay, Australia, researchers have spent thousands of hours listening to the whistles, clicks and other mechanical-sounding noises made by wild dolphins. Among them is Stephanie King of the University of Bristol, co-leader of Shark Bay Dolphin Research, one of the longest-running studies of wild dolphins, who has devoted much of her career to watching and cataloguing the signals of this population.
In recent years, animal behaviourists have begun to add machine learning and other forms of artificial intelligence to their toolkits. These methods can sift through vast archives of audio and movement data far more quickly than humans can, flagging patterns, clustering calls and correlating signals with behaviours or environmental conditions. Proponents say the techniques can reveal structure in animal communication that was previously invisible and scale up long-term monitoring in ways that could aid conservation and research.
But the rush to apply AI to non-human signalling has prompted hard questions about purpose and ethics. Many researchers caution against framing this work as a simple matter of "talking to animals": interpreting animal sounds through the lens of human language risks anthropomorphism, and automated classification can mask uncertainty about meaning and context. There are also practical concerns about how models are trained, whether datasets are representative, and how errors might affect animals if interventions are based on misread signals.
Those sceptical of a technology-first approach argue that the real measure of progress should be whether these tools improve animal welfare, conservation outcomes or scientific understanding — not whether they generate headline-grabbing claims about "decoding" animal language. That requires careful study design, cross-disciplinary collaboration between AI specialists and field biologists, transparent methods and sensitivity to the limits of inference from observational data.
Researchers also face questions about who benefits from the work. Public fascination with the idea of communicating with other species can attract funding and attention, but it can also skew priorities toward projects that are sensational rather than those that address pressing conservation needs. Responsible deployment, advocates say, means setting clear goals tied to ecological or welfare outcomes and engaging with local communities and stakeholders where research takes place.
Ultimately, AI offers powerful new ways to handle large, complex datasets in animal behaviour, but its value will depend on judicious use and rigorous validation. As teams working in places such as Shark Bay continue to combine decades of careful observation with computational tools, the field is wrestling with how to harness the technology so it serves both scientific understanding and the animals themselves, rather than merely amplifying human curiosity.