Google's DolphinGemma AI Brings New Insights to Dolphin Communication

The AI model, developed with Georgia Tech and the Wild Dolphin Project, offers real-time analysis and aims to unlock interspecies communication potential.

Google unveiled DolphinGemma, a 400-million parameter AI model designed to decode and generate dolphin vocalizations using decades of research data from the Wild Dolphin Project.
The model is optimized to run on Google Pixel smartphones, enabling researchers to analyze dolphin sounds in real time during fieldwork in the Bahamas.
A companion system, CHAT, is being developed to facilitate rudimentary two-way communication by associating synthetic whistles with specific objects dolphins recognize.
DolphinGemma will be released as an open-source model this summer, allowing researchers to adapt it for studying other cetacean species like bottlenose dolphins.
The initiative builds on over 40 years of marine research and represents a significant step toward understanding the structure and potential meaning of dolphin communication.

Interesting Engineering