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

An Israeli study has uncovered a critical flaw in artificial intelligence's ability to comprehend animal communication. The research, published in Current Biology, reveals that AI is missing the point when decoding animal and human toddler language. For years, scientists have turned to AI to understand bird calls, bat chatter, and whale songs.

However, the new study, led by Prof. Yosef Yovel from Tel Aviv University, suggests AI needs a significant paradigm shift. Instead of focusing solely on raw audio analysis, AI should aim to understand neural and perceptual meaning.

The study, conducted by a team of researchers from various institutions, including the Hebrew University of Jerusalem and the University of Edinburgh, highlights the importance of considering the receiver's perception in animal communication. A spectrogram, a visual representation of sound frequencies, shows that acoustically similar sounds might not always convey the same message, while different sounds could carry the same meaning depending on context.

This finding emphasizes that communication relies heavily on the receiver's perception rather than just the physical properties of the sound wave.

The researchers discovered that even though AI models performed better than traditional acoustic methods in classifying sounds, they still failed to capture the nuances of meaning. For instance, they grouped together vocalizations with different messages and separated vocalizations with the same meaning. Moreover, the AI models failed to identify how sequences of vocalizations expressed increasing urgency, a distinction that humans perceive naturally.

To overcome this challenge, the team suggests combining AI tools with behavioral observations, playback experiments, and research into the nervous system. While AI is a powerful tool, it cannot replace the unique perspective of the animal itself.