Throughout most of human history, animals have been signaling information to each other in all directions around us but without us knowing what they are trying to say. Birds chirp, whales make sophisticated sounds, elephants communicate over vast distances, and most species use combinations of sounds, gestures, and other signals.
AI now offers a way for scientists to study these systems more effectively.
Where previously scientists had to analyze recordings manually, machine learning models can process vast arrays of animal sounds and identify patterns which human analysis might overlook. These can be compared to behavior, environment, social relations, and other factors.
Project CETI is one such project, studying sperm whale communication by means of machine learning, robotics, and big-scale recordings of sound. The scientists have already identified some structured patterns in the whale “codas” that is, in the sequence of clicks they use in communication.
Nature
Other teams use AI in studying various species. Earth Species Project, for instance, created NatureLM-audio, which is an AI system that analyzes animal sounds in order to enable scientists to detect and categorize them.
Earth Species Project
However, finding patterns does not mean understanding the language.
This difference is likely the most crucial aspect of the entire topic.
An AI system can learn that a certain sound usually precedes a certain behavior. This is useful information, but it doesn’t necessarily tell us what the animal means consciously.
Thus, scientists still require behavioural tests and observations to confirm the predictions made using AI. For instance, Project CETI calls for playback tests as part of its planned validation approach.
Project CETI
However, there are serious problems. Animal communication is shaped by various factors such as social relations, environment, experience of the animal, and context. Thus, it is much more complicated than matching one sound to one translation.
Earth Species Project
In my opinion, AI will serve as a scientific tool for animals sooner than as their translators.
It will allow researchers to detect patterns hidden in millions of recordings and pose questions no human could even think about. However, finding the meanings of those patterns would require biology, observation, experiments, and validation.
Thus, is AI capable of decoding animal communication?
Yes, and it is doing it now.
However, we’re closer to learning the patterns of animal communication than talking to other species.
Earth Species Project
