Google Launches DolphinGemma AI to Decode Dolphin Communication

The AI model, developed with decades of Wild Dolphin Project research, is now operational with real-time field testing on Pixel devices and plans for open-source release this summer.

DolphinGemma, a 400M parameter AI model, analyzes dolphin vocalizations to uncover patterns and potential meanings in their communication.
The model is trained on a unique dataset from the Wild Dolphin Project, which has documented Atlantic spotted dolphins' sounds and behaviors for nearly 40 years.
DolphinGemma runs on Google Pixel phones, enabling real-time audio processing in field research without the need for custom hardware.
In collaboration with Georgia Tech, researchers are testing the CHAT system to explore two-way communication by associating synthetic sounds with objects of interest to dolphins.
Google plans to release DolphinGemma as an open-source model this summer, aiming to support broader research on cetacean communication across species.

Interesting Engineering