Overview

• DolphinGemma, developed by Google with Georgia Tech and the Wild Dolphin Project, uses AI to decode dolphin vocalizations and identify potential language-like structures.
• The model runs on Pixel phones, enabling real-time field analysis and reducing the need for costly custom hardware.
• Researchers are integrating DolphinGemma with the CHAT system, which plays synthetic whistles linked to objects, aiming to create a rudimentary shared vocabulary with dolphins.
• Google plans to release DolphinGemma as an open-source model this summer, allowing adaptation for other cetacean species and broader research collaboration.
• The AI is trained on over 40 years of meticulously labeled data, including sound-behavior pairings of Atlantic spotted dolphins, collected by the Wild Dolphin Project in the Bahamas.