Israeli study: does AI understand animal language
Scientists from Tel Aviv University tested whether AI can decipher animal "speech." The study showed that neural networks analyze the acoustic characteristics of sounds but do not understand their meaning. To decode animal communication, acoustic analysis alone is insufficient; AI must be combined with behavioral observation.
A group of scientists from Tel Aviv University, led by Professor Yossi Yovel from the School of Zoology and the Steinhardt Museum of Natural History, tested whether artificial intelligence can decipher animal language. The study, published in the journal Current Biology, used recordings of infant sounds as an intermediate model. The recordings included sounds in three situations: distress, addressing mom or dad, and requesting food. Scientists analyzed them using three methods: a traditional acoustic method and two deep neural networks. One network was trained on animal sounds, the other on adult human speech. Results showed that neural networks were better than the traditional method at finding acoustic patterns, but they often grouped similar sounds with different meanings and separated dissimilar signals conveying the same message. The neural networks also failed to detect increasing urgency in a sequence of signals. The main conclusion: AI analyzes how a signal sounds but does not understand what it means to the receiver. To decode animal communication, AI must be combined with behavioral observation, playback experiments, and brain activity studies.