What it is
Project CETI (Cetacean Translation Initiative) is a multidisciplinary research project using machine learning and natural language processing to decode sperm whale communication. Based in Dominica, the team has deployed underwater microphone arrays and bio-logging tags to build the largest dataset of sperm whale vocalizations ever assembled. The project brings together marine biologists, AI researchers, linguists, and roboticists from institutions including MIT, Harvard, and the Weizmann Institute.
Why we picked this
This is the most ambitious interspecies communication project in history, and the implications extend far beyond whales. If CETI can demonstrate that sperm whale codas contain structured, meaningful communication (early results suggest they do), it would fundamentally change how we understand animal cognition and our ethical obligations to other species. The project also advances AI: training models on non-human communication pushes the boundaries of language understanding.
Key takeaways
- Sperm whales produce 'codas': patterns of clicks that vary by clan, region, and social context, suggesting cultural transmission similar to human dialects.
- CETI has recorded over 100,000 whale codas so far, using deep learning to identify patterns that human researchers missed in decades of study.
- The project's AI models have identified at least 150 distinct coda types, with evidence of combinatorial structure (whales combining elements to create new meanings).