Google’s DolphinGemma AI Listens, Learns, and Talks Dolphin

It was announced on National Dolphin Day, April 14, and is part of a long-term research effort to study dolphin sounds. The project is open-source and uses artificial intelligence to analyze the sounds dolphins make—like clicks, whistles, and bursts—and to create similar sounds.

The model was developed with help from Georgia Tech and the Wild Dolphin Project (WDP). WDP has been studying Atlantic Spotted Dolphins in the Bahamas since 1985. Over the years, they have collected a large amount of labeled audio and video from underwater, which helped train the AI system.

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe - We publish new crypto explainer videos every week!

What is Decentralized Crypto Gambling? (Animated Explainer)

SUBSCRIBE

ON YOUTUBE

DolphinGemma looks for repeated sound patterns and sequences. These might help show whether dolphin communication has a structure similar to language.

One feature of DolphinGemma is that it is small enough to run directly on Google Pixel phones. It uses a sound processing method that allows it to predict the next sound in a sequence, similar to how other AI tools predict the next word in a sentence. Field researchers currently use Pixel 6 phones to work with the model while at sea.

The model also works with another system called CHAT (Cetacean Hearing Augmentation Telemetry). CHAT assigns custom-made dolphin-like sounds to specific items that dolphins often interact with, like seagrass or scarves.

On April 5, David Sacks said the United States might be leading the global AI race, thanks to Meta’s latest release. What is it? Read the full story.