Google's DolphinGemma AI Advances Dolphin Communication Study

Google has developed DolphinGemma, an AI model in collaboration with the Wild Dolphin Project (WDP) and Georgia Tech, to decode dolphin communication by analyzing their vocalizations. DolphinGemma is based on Google's Gemini models and is trained on a vast dataset of dolphin sounds, aiming to understand whether these vocalizations constitute a language. WDP has been collecting data on dolphin sounds and behavior for over 40 years, providing an extensive dataset for the AI. The model, capable of running on Google Pixel phones, will be deployed in the field to help researchers uncover hidden structures and potential meanings in dolphin communication. This initiative marks a significant step towards potentially 'speaking dolphin' by identifying patterns and predicting subsequent sounds in a sequence, similar to how language models predict human text. The project also incorporates the CHAT system to facilitate interactive communication with dolphins.