Google’s DolphinGemma AI model brings humans closer to understanding dolphin language

Google’s new DolphinGemma AI model represents a significant breakthrough in decoding animal communication, potentially enabling humans to understand and interact with dolphins in their natural environment. By applying machine learning to decades of dolphin vocalization research, Google has created a system that not only analyzes dolphin sounds but can generate realistic dolphin-like responses, marking a pivotal advancement in interspecies communication technology.

The big picture: Google has developed DolphinGemma, a foundational AI model designed to learn and generate dolphin vocalization patterns, in collaboration with Georgia Tech researchers and the Wild Dolphin Project.

• The ~400M parameter model uses the SoundStream tokenizer to represent dolphin sounds and can run directly on Pixel phones for field research.
• Released on National Dolphin Day, the technology builds upon decades of dolphin communication research to identify patterns in natural dolphin sounds and predict likely subsequent vocalizations.

Behind the research: The Wild Dolphin Project has conducted the world’s longest-running underwater dolphin study since 1985, documenting communication patterns of Atlantic spotted dolphins in the Bahamas.

• Researchers have identified different types of dolphin communications, including signature whistles that function as unique names, burst-pulse “squawks” used during confrontations, and click “buzzes” employed during courtship or shark encounters.
• This extensive multi-generational research provided the foundation for training the DolphinGemma AI model.

Practical applications: The research team is developing the CHAT (Cetacean Hearing Augmentation Telemetry) system that leverages Pixel phones to enable real-time dolphin communication analysis.

• The system can accurately identify dolphin mimicry amid ocean noise, determine which specific whistle was mimicked, and inform researchers about which object a dolphin “requested.”
• This technology allows researchers to respond quickly to dolphin communications, potentially enabling more sophisticated interspecies interactions.

What’s next: Google plans to share DolphinGemma as an open model this summer, making the technology widely available to the scientific community.

• The open release may accelerate research across various cetacean species beyond the Atlantic spotted dolphins currently studied.
• This approach reflects Google’s commitment to advancing scientific understanding through AI while making powerful research tools accessible to specialists worldwide.