AI Cannot Yet 'Translate' Animal Languages, Tel Aviv University Researchers Find

Researchers from Tel Aviv University and other institutions have identified a fundamental limitation in using artificial intelligence to decipher animal communication: the similarity of sounds does not equate to similarity in meaning. The study, published in Current Biology, was announced by Tel Aviv University.

To test their hypothesis, the scientists employed a unique model: the vocalizations of young children who had not yet learned to speak. Unlike animal communication, human adults can generally infer the meaning of these sounds. The recordings captured three scenarios: a child experiencing discomfort, calling for a specific parent (mother or father), or requesting food.

The recordings were analyzed using a traditional acoustic method and two advanced deep neural networks. One network was trained on animal sounds, the other on human speech. While the neural networks excelled at identifying acoustic patterns compared to the traditional method, they failed to reliably categorize the sounds by their meaning. The AI models sometimes grouped signals with different meanings and, conversely, separated sounds used by a child to convey the same message.

The authors concluded that effectively deciphering animal communication requires more than just collecting vast amounts of recordings for AI analysis. It necessitates simultaneous study of animal behavior, their responses to played-back signals, and in some cases, their neurological activity. Essentially, an algorithm may accurately recognize sound structure but does not necessarily comprehend the meaning behind the animal's vocalization.