How Google is Using AI to Connect Humans and Animals
Google leverages its advanced large language models (LLMs), such as Gemini, to power enterprise-ready AI solutions across industries.
By integrating LLMs into platforms, Google empowers global businesses to build custom chatbots, automate information retrieval and deploy intelligent agents that enhance customer engagement and operational efficiency.
Now, the company has announced that it’s using LLMs for a very different and unique purpose: Dolphin translation.
What is DolphinGemma?
Google has developed a specialised large language model (LLM) designed to analyse dolphin communication patterns.
The model, DolphinGemma, processes the clicks, whistles and burst pulses that form dolphin vocalisations.
The project is a collaboration between Google, researchers at Georgia Institute of Technology (Georgia Tech) and the Wild Dolphin Project (WDP), which conducts field research on Atlantic spotted dolphins.
DolphinGemma functions as an audio-in, audio-out model that processes sequences of dolphin sounds to identify patterns and structure, similar to how language models (LMs) for human communication predict subsequent words in sentences.
The model utilises Google's SoundStream tokeniser to represent dolphin sounds efficiently. This 400 million parameter model runs directly on Pixel smartphones, which WDP researchers use in their field work.
“This model builds upon insights from Gemma, Google's collection of lightweight, state-of-the-art open models that are built from the same research and technology that powers our Gemini models,” Google says.
How Google Pixel smartphones enable real-time analysis with CHAT system
In addition to analysing natural communication, WDP has developed the Cetacean Hearing Augmentation Telemetry (CHAT) system in partnership with Georgia Tech.
This underwater computer system attempts to establish a shared vocabulary between humans and dolphins using synthetic whistles associated with objects of interest to the dolphins.
The system requires real-time sound processing capabilities to identify dolphin vocalisations amid ocean noise and provide feedback to researchers through bone-conducting headphones.
A Google Pixel 6 smartphone has handled the analysis of dolphin sounds in current research, with plans to deploy Google Pixel 9 devices in summer 2025 research.
The use of commercially-available smartphones also reduces the need for custom hardware.
“Using Pixel smartphones dramatically reduces the need for custom hardware, improves system maintainability, lowers power consumption and shrinks the device's cost and size – crucial advantages for field research in the open ocean,” Google says.
DolphinGemma model to be shared with wider research community this summer
Google plans to release DolphinGemma as an open model this summer, making it available to researchers studying other cetacean species.
While the model has been trained specifically on Atlantic spotted dolphin sounds, researchers studying bottlenose or spinner dolphins may be able to fine-tune it for their work.
The decision to share the model reflects an acknowledgement of the value of collaborative scientific research.
“By providing tools like DolphinGemma, we hope to give researchers worldwide the tools to mine their own acoustic datasets, accelerate the search for patterns and collectively deepen our understanding of these intelligent marine mammals,” says Google.
“We're not just listening anymore. We're beginning to understand the patterns within the sounds, paving the way for a future where the gap between human and dolphin communication might just get a little smaller.”
The foundation for DolphinGemma’s development
WDP has conducted underwater dolphin research since 1985, creating what it describes as the world's longest-running study of a specific community of wild Atlantic spotted dolphins in the Bahamas.
Its non-invasive approach has generated a dataset containing decades of underwater video and audio, paired with individual dolphin identities, life histories and observed behaviours.
WDP researchers have previously connected specific sound types to behavioural contexts, such as signature whistles used by mothers and calves to reunite, burst-pulse “squawks” during confrontations and click “buzzes” during courtship or shark-chasing behaviour.
The organisation's extensive, labelled dataset provided the training material for DolphinGemma, which aims to identify recurring sound patterns, clusters and sequences that might reveal structure within dolphin communication.
WDP has begun deploying DolphinGemma in field research this season.
Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Also sign up to our free weekly newsletter for the latest insights and stories straight into your inbox.
AI Magazine is a BizClik brand
