Google working to decode dolphin communication using AI

Cracking the dolphin code.

Dolphins are one of the smartest animals on Earth and have been revered for thousands of years for their intelligence, emotions and social interaction with humans.

Now Google is using artificial intelligence (AI) to try and understand how they communicate with one another – with the hope that one day humans could use the technology to chat with the friendly finned mammals.

Google has teamed up with researchers at Georgia Institute of Technology and the Wild Dolphin Project (WDP), a Florida-based non-profit which has been studying and recording dolphin sounds for 40 years, to build the new AI model called DolphinGemma.

Google is using artificial intelligence (AI) to try and understand how dolphins communicate with one another – with the hope that one day humans could use the technology to chat with them. (iStock)

‘SUPER POD’ OF MORE THAN 1,500 DOLPHINS OFF CALIFORNIA COAST CAPTURED ON DRONE VIDEO

For decades, WDP has correlated sound types with behavioral contexts. For instance, signature whistles have been used by mothers and calves to reunite, while burst pulse "squawks" are often observed during dolphin fights, researchers said, according to a Google blog on the project.

Click "buzzes" are often used during courtship or chasing sharks.

Now, using the vast data gathered by WDP, Google has built DolphinGemma, building upon Google’s own AI lightweight open model, known as Gemma.

DolphinGemma has been trained to analyze the vast library of recordings to detect patterns, structures, and even potential "meanings" behind the dolphin communications or vocalizations. 

Over time, DolphinGemma will try to organize the dolphin sounds into categories — almost like words, sentences, or expressions in human language.

Google has teamed up with researchers at Georgia Institute of Technology and the Wild Dolphin Project (WDP), a Florida-based non-profit which has been studying and recording dolphin sounds for 40 years, to build the new AI model called DolphinGemma.

CHINESE HUMANOID ROBOT WITH EAGLE-EYE VISION AND POWERFUL AI

"By identifying recurring sound patterns, clusters and reliable sequences, the model can help researchers uncover hidden structures and potential meanings within the dolphins' natural communication — a task previously requiring immense human effort," a post on Google about the project reads.

"Eventually, these patterns, augmented with synthetic sounds created by the researchers to refer to objects with which the dolphins like to play, may establish a shared vocabulary with the dolphins for interactive communication."

DolphinGemma uses Google’s Pixel phone technology, specifically the audio recording technology used in Pixel devices, to make clean, high-quality sound recordings of dolphin vocalizations.

The Pixel phone technology can separate out dolphin clicks and whistles from background noise like waves, boat engines, or underwater static. That clean audio is critical for AI models like DolphinGemma, because messy, noisy data would confuse the AI, researchers said.

A zookeeper wearing a protective mask looks inside the mouth of a dolphin at the zoologic park "Planete Sauvage" in Saint-Pere-en-Retz, outside Nantes, on May 6, 2020. Dolphins are one of the smartest animals on Earth and have been revered for thousands of years for their intelligence, emotions and social interaction with humans. (LOIC VENANCE/AFP via Getty Images)

GET FOX BUSINESS ON THE GO BY CLICKING HERE

Google says it plans to release DolphinGemma as an open model this summer, allowing researchers around the world to use and adapt it.

Although it’s trained on Atlantic spotted dolphins, the model could also help study other species like bottlenose or spinner dolphins with some fine-tuning, researchers said. 

"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," the blog post reads.