AI Projects Decoding Animal Communication Achieve 72% Accuracy in Whale Click Translation

Key Points

• Project CETI achieves 72% accuracy in decoding sperm whale clicks using machine learning
• Earth Species Project's NatureLM-Audio identifies animal species from vocalizations
• Dolphin communication contains word-like functions verified by $100,000 research prize
• $500,000 prize still unclaimed for AI enabling undetected human-animal communication

Full Details

A wave of AI projects is transforming our ability to understand animal communication. Project CETI (Cetacean Translation Initiative) is using machine learning to decode sperm whale codas—the rhythmic clicks that function like underwater Morse code—with 72% accuracy in behavioral prediction. The Earth Species Project's NatureLM-Audio model operates like a universal wildlife translator, training on human language and music to identify animal species from vocalizations. Marine biologist Dr. Sayigh received $100,000 from the Coller-Dolittle Prize in May 2025 for analyzing dolphin whistles that reveal word-like functions in their communication. A $500,000 challenge remains unclaimed for any AI system that can enable undetected human-animal communication. Scientists have also established ethical guidelines to prevent AI research from disrupting animal social structures.