Project Ceti Trains AI on Dominican Sperm Whale Codas for Translation and Conservation
Updated
Updated · The Guardian · Aug 27
Project Ceti Trains AI on Dominican Sperm Whale Codas for Translation and Conservation
2 articles · Updated · The Guardian · Aug 27
Summary
Project Ceti is training machine-learning models on sperm whale codas off the Dominican coast to detect patterns that could eventually support translation and sharpen conservation tools.
Its latest findings suggest whales use different codas in different contexts, while some click sequences resemble vowel-like sounds once gaps are removed and shift tempo and frequency around shipping noise.
Those signals could help researchers tie communication to behavior and disturbance, much as earlier acoustic work led Canada in 2018 to close key beluga grounds to boats after calf calls were shown to be easily masked.
Researchers still dispute whether such parallels amount to language in the human sense, but broadly agree that decoding whale communication could guide more targeted protection in a noisier, warming ocean.