Leading researchers have unveiled DolphinGemma, the first major language model designed to decipher the language of dolphins. The project is the result of a collaboration between research organization The Wild Dolphin Project and Google DeepMind.

Dr. Denise Herzing, who heads The Wild Dolphin Project, explains that the main goal is to learn how to “speak” dolphin language in the future. The project focuses on studying a pack of Atlantic spotted dolphins off the coast of the Bahamas.

DolphinGemma is based on Google's Gemini technology. It can predict sequences of sounds, similar to how auto-selection of text works in search engines. The project also aims to create two-way communication with dolphins.

Herzing has amassed an extensive library of dolphin sounds over the past 40 years, working with Dr. Ted Starner of Google DeepMind. DolphinGemma uses sophisticated algorithms to analyze acoustic data and identify patterns in dolphin sounds.

DolphinGemma is expected to be used extensively to study dolphin behavior. Google plans to make the model open to other researchers in the middle of the year