Will Artificial Intelligence allow us to talk to animals?

By Isaac Williams

In the last couple of years, AI has begun to permeate almost all spheres of our lives. From recommending songs on Spotify to helping us finish our history essays after pulling an all-nighter, there are few students’ lives that have not been touched by Artificial Intelligence. One of the areas that has not been touched by AI is the natural world, at least until recently, where there are now many scientific research groups who are developing new AI processes to try and develop their understanding of the whole animal kingdom.

One of the main fields of this research that AI is rapidly transforming is animal communication. This involves the noises, gestures, and signals animals emit and receive from one another to share information. A key research group that has been pioneering this is Project CETI (Cetacean Translation Initiative). Cetaceans are a group of animals, including whales, porpoises, and dolphins, which are known for their high intelligence and complex social structure, making them a key target of interest for researchers. The scientific study of their songs began with the first recording being taken in the late 1960s.

A recent breakthrough has been the identification of individual “vowels” in whale calls and that they can be intentionally transitioned between

Progress in developing a sense of understanding of the “language” of whales has been slow, with the hums, clicks, and shrieks of a whale being too alien for any human to analyse or identify complex patterns within the songs. This is where AI becomes useful, with a greater ability to identify patterns in unknown systems as its processes of pattern recognition are more abstract and less affected by the subconscious biases exhibited by humans when trying to identify patterns. A recent breakthrough has been the identification of individual “vowels” in whale calls and that they can be intentionally transitioned between; this can even happen mid-call like an equivalent of diphthongs in human speech. There has also been the discovery of dialects in whale calls, with different pods having distinctive features in their calls. With pods within close proximity to one another having stronger variation, indicating a call’s characteristics are intentionally changed to highlight a whale’s group identity. There have also been discoveries regarding “linguistic rules” used by whales and whether certain combinations of calls and clicks would be present in a whale song.

These discoveries with cetaceans represent just the beginning, as researchers are finding that AI can decode communication patterns in species far beyond the ocean’s depths. A recent tool developed by the Earth Species Project is a large language model (like ChatGPT) called NatureLM-audio, a model that has been tailored specifically for animal sounds. When given animal sounds, it can perform a range of tasks such as classifying the species and life stage of the caller, identifying the call type, captioning bioacoustic data, and it even has some success at identifying the caller when a species is present that it has not yet been trained on. Another model has also been developed that, when presented with an unknown set of calls, can identify which features of the call are most likely to contain meaning within the set. Knowing this is a key step on the route to fully translating an animal’s call into English or another human language.

The implications of this research extend far beyond scientific curiosity. If we can truly decode animal communication, it could revolutionise conservation efforts by allowing us to understand what animals need to thrive in their environments. It might also force us to reconsider fundamental questions about animal consciousness, intelligence, and rights. While nobody knows exactly where this technology will lead us, the rapid development of machine-learning techniques suggests we may be closer than ever to bridging the communication gap between humans and other species. Perhaps in the future, buying a pet will involve having a conversation first, not just to see if you like them, but to understand what kind of life they actually want to live.

Image: Charles J. Sharp via Wikimedia Commons
HTTPS://CREATIVECOMMONS.ORG/LICENSES/BY-SA/4.0/