Using babies, Israeli study reveals why AI fails to truly understand animal calls

A Tel Aviv University study published in Current Biology reveals that AI fails to decode animal and toddler vocalizations because it analyzes acoustic properties rather than perceptual meaning. Researchers found that acoustically similar sounds can carry different meanings, while different sounds can convey the same message, depending on the receiver's perception. The study used human toddler vocalizations in distress, calling, and food-request contexts, testing classical acoustic methods and deep neural networks. None of the AI models successfully classified vocalizations by meaning or detected increasing urgency in sequences, which humans perceive naturally.

A groundbreaking Israeli study published in Current Biology reveals why artificial intelligence fails to truly understand animal and human toddler vocalizations. Led by Prof. Yosef Yovel of Tel Aviv University, the research demonstrates that AI analyzes acoustic properties like volume and pitch but misses the perceptual meaning that receivers actually use. The team recorded human toddlers vocalizing in three contexts: distress, calling to a specific parent, and asking for food. They tested classical acoustic methods and two state-of-the-art deep neural networks — one trained on animal vocalizations and another on adult human speech. None of the models successfully classified the vocalizations by their meaning. Acoustically similar sounds were grouped together despite carrying different messages, while different sounds conveying the same message were separated. The models also failed to detect increasing urgency in sequences, which humans perceive naturally. The study argues that communication relies on the receiver's perception and context, not just sound wave properties. Yovel emphasized that AI is a powerful tool but no substitute for understanding the animal's own perspective. The research involved scientists from Tel Aviv University, Hebrew University, the University of Edinburgh, and German institutions.