Google creates AI to decode dolphin sounds

Developed with Georgia Tech and Wild Dolphin Project, the AI learns from real dolphin data.

Google has overtaken IBM to become the top filer of generative AI patents, also leading in agentic AI filings alongside Nvidia.

Google DeepMind has developed a groundbreaking AI model capable of interpreting and generating dolphin vocalisations.

Named DolphinGemma, the model was created in collaboration with researchers from Georgia Tech and the Wild Dolphin Project, a nonprofit organisation known for its extensive studies on Atlantic spotted dolphins.

Using an audio-in, audio-out architecture, the AI DolphinGemma analyses sequences of natural dolphin sounds to detect patterns and structures, ultimately predicting the most likely sounds that follow.

The approach is similar to how large language models predict the next word in a sentence. It was trained using a vast acoustic database collected by the Wild Dolphin Project, ensuring accuracy in modelling natural dolphin communication.

Lightweight and efficient, DolphinGemma is designed to run on smartphones, making it accessible for field researchers and conservationists.

Google DeepMind’s blog noted that the model could mark a major advance in understanding dolphin behaviour, potentially paving the way for more meaningful interactions between humans and marine mammals.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!