Photo credit: www.entrepreneur.com
What do dolphins communicate about among themselves? Google’s latest AI model, aimed at deciphering the hidden language of these marine mammals, seeks to bridge the gap between human and dolphin interaction.
Recently, Google, in collaboration with the Georgia Institute of Technology and the Wild Dolphin Project, unveiled an innovative AI model named DolphinGemma.
DolphinGemma marks a groundbreaking effort to interpret the vocalizations of dolphins. This AI model has been trained on an extensive collection of audio and video recordings spanning 40 years, focusing on the Atlantic spotted dolphins found in the Bahamas. The objective is to uncover recurring patterns, structural elements, and potential meanings within dolphin communication.
Similar to how AI models predict the next word in a text, DolphinGemma endeavors to forecast the subsequent sound a dolphin will produce based on established patterns. Additionally, it possesses the ability to generate new, artificially created dolphin sounds.
The Wild Dolphin Project plans to utilize DolphinGemma in their field research this season, marking its inaugural application in real-world settings to enhance insights into dolphin communication.
One significant advantage of AI is its capacity to detect patterns that may elude human researchers, allowing for quicker and more comprehensive analysis of dolphin vocalizations. Dr. Denise Herzing, the founder and research director of the Wild Dolphin Project, emphasized that manually sifting through the data would take humans over 150 years to uncover the same insights that DolphinGemma can achieve in a fraction of the time.
“By analyzing dolphin vocalizations through an AI model like DolphinGemma, we can gain profound insights into subtle patterns that might go unnoticed by human observers,” Herzing noted in a recent announcement video. “Ultimately, our goal is to achieve a form of communication with dolphins.”
DolphinGemma will also generate novel, synthetic dolphin-like sounds, which researchers will introduce into the water to observe dolphins’ responses to these new vocalizations.
The experimental process involves researchers swimming alongside dolphins, playing the AI-generated sounds while engaging in food exchange, using items that dolphins favor, such as seagrass or sargassum. If a dolphin mimics the AI-generated sounds, the researchers will reward it with a treat.
However, the study is currently confined to one specific population of dolphins, as communication methods may differ among dolphin groups. Google aims to release DolphinGemma as an open-source AI model this summer, enabling academic researchers to expand its application to different dolphin species, such as bottlenose or spinner dolphins.
“By providing tools like DolphinGemma, we aspire to empower researchers worldwide to analyze their acoustic datasets, hasten the discovery of patterns, and collectively enhance our understanding of these intelligent marine creatures,” Google stated in a recent blog post.
AI applications extend beyond dolphins, with ongoing projects such as the Earth Species Project unveiling an AI model called NatureLM, designed to assess an animal’s species, age, and emotional state from audio. Additionally, Project CETI (Cetacean Translation Initiative) utilizes AI techniques to investigate communication among sperm whales.
Source
www.entrepreneur.com