Google has developed an AI model to understand the language of dolphins

Google DeepMind research lab has created a large language model (LLM) called DolphinGemma to help decipher dolphin vocalizations.

The model was trained using data from the Wild Dolphin Project, a non-profit organization that studies the behavior of Atlantic bottlenose dolphins. According to Google, DolphinGemma can create sound sequences similar to those produced by dolphins, while being powerful enough to run on smartphones.

By identifying recurring sound patterns, clusters, and reliable sequences, the model can help researchers uncover hidden structures and potential meanings in natural dolphin communication.

WDP researchers will use special underwater systems CHAT (Cetacean Hearing Augmentation Telemetry). The system will be able to recognize, record and reproduce individual sound patterns. Inside CHAT will be a Google Pixel 9 smartphone with DolphinGemma installed.