🔍 Summary:
Google has embarked on an innovative project with marine biologists and AI experts to develop DolphinGemma, a large language model (LLM) aimed at understanding and eventually replicating dolphin communication. This collaboration, involving the Wild Dolphin Project and Georgia Tech researchers, leverages a rich dataset from a 41-year study of Atlantic spotted dolphins in the Bahamas. The dataset includes various dolphin sounds linked to specific behaviors, which has enabled the training of DolphinGemma.
DolphinGemma operates by analyzing dolphin vocalizations and predicting subsequent sounds, effectively learning the structure of dolphin communication. This model is part of a broader trend where AI is used to interpret animal communications, such as identifying dog emotions from barks or deciphering whale communication patterns.
The ultimate goal of this project is to facilitate two-way communication between humans and dolphins using a system called Cetacean Hearing Augmentation Telemetry (CHAT). This system, implemented on modified smartphones, plays dolphin-like sounds associated with objects like toys or seagrass, encouraging dolphins to mimic these sounds to request the objects, thus creating a basic shared vocabulary.
Google plans to make DolphinGemma an open model to benefit broader research efforts across different dolphin species. This project not only highlights the potential of AI in wildlife research but also raises important ethical questions about interacting with wild animals as communication methods advance.
📌 Source: https://gizmodo.com/googles-new-ai-is-trying-to-talk-to-dolphins-seriously-2000589467