Many people wonder what birds, dogs, or cats say to each other. Humans still struggle to fully grasp animal conversations. Artificial intelligence steps in as a bridge, translating animal chatter into something people can read. A team of researchers takes on this challenge, exploring what AI needs to translate across species boundaries.
Researchers Yossi Yovel and Oded Rechavi from the Faculty of Zoology and the Wise Faculty of Life Sciences at Tel Aviv University in Israel investigated what it takes for artificial intelligence to interpret animal signals. The study looks at both what is possible and what remains hard for AI in this field.
The study, titled Artificial Intelligence and Doctor Dolittle’s Struggle, outlines the potential of AI to understand animal signals and also the boundaries. The title nods to a beloved children’s book about a doctor who communicates with animals by listening to their language, including a West African parrot named Polynesian who helps Dr. Dolittle understand bird speech.
Trying to interpret another species’ communication goes beyond simply decoding sounds. It involves many factors that must be considered to read messages as a whole. Researchers identify three main obstacles in this work.
Context
The first obstacle is figuring out the context behind animal signals. Humans have long studied and even mimicked animal communication, drawing responses from various species without using AI.
For example, a female-looking robot frog attracted real male frogs to court, and a robotic fish interacted with live fish during learning, influencing their movement.
A robotic bee could recruit real bees to follow the waggle dance and fly to a forest location. The waggle dance describes how bees share food-location information through movement, essentially a nonverbal instruction system.
These demonstrations show how engineered stimuli can provoke responses and carry messages within a given situation. While AI can imitate animal-like sounds, deciding whether those sounds mean territory, courtship, danger, or something else is far more challenging.
After analyzing recordings, AI can produce a particular bird song, but more context is needed to decide if the song marks territory, signals a mate, or conveys another message altogether.
The same challenge exists for silent communication in insects, which often relies on chemical signals. Without observing insect behavior, it is hard to tell whether a chemical cue signals mating, danger, or a random release.
Today, AI needs human input and clear definitions to begin addressing animal communication, such as field recordings of bird songs. Human biases shape how signals are interpreted because context must be assigned to the songs to help the AI draw connections among vocalizations. Attribution Tel Aviv University researchers Yovel and Rechavi, 2024 study on AI and animal communication.
El eliciting a natural response
Animals display a wide range of behaviors, and responses can be shaped by factors like physiological state, social dynamics, and environmental conditions. Different species rely on diverse sensory channels for communication, including sound, chemical cues, and body language. Identifying specific responses tied to communication requires careful observation; experiments cannot train animals to produce responses, as that would distort natural communication and limit observations to laboratory settings.
Measuring responses can be tricky. External signs may be subtle or unclear, and observers may miss these nuances. AI systems built to interpret responses risk finding spurious correlations or overlooking subtle indicators.
Limited context range
The third obstacle centers on focusing mainly on narrow contexts such as general communication, alarm, and courtship. This narrow lens can limit understanding of animal communication across a broader set of topics, potentially hindering cross-species dialogue.
At minimum, expectations may fall short if an AI translator primarily reports that birds convey simple emotional cues like sadness or affection all day long.
Potential benefits
Advancing animal communication offers meaningful benefits. Understanding bees and other pollinators could support agriculture, while insights into domestic animals’ moods or tendencies can improve care and welfare. Detecting noise patterns in protected forests could alert conservationists to unseen threats, and researchers imagine applying these methods to farm settings to interpret animal states more clearly.
There is even speculative value in broader contexts, such as testing the possibility of first contact with nonhuman intelligences elsewhere. If clear communication with intelligent life on Earth remains elusive, it raises questions about what would be needed to communicate with life on other worlds.
References and related discussions contribute to this ongoing exploration of AI and animal language, highlighting progress and remaining questions. Endnotes and further reading are presented with attribution to the original researchers in the cited study. Attribution Tel Aviv University study on AI and animal communication.