Google just launched a new AI model called DolphinGemma that might help humans understand how dolphins “talk” and it’s all happening with the help of Pixel phones and decades of underwater research!
- DolphinGemma is Google’s AI model built to decode dolphin vocal patterns using underwater recordings.
- It uses audio tech called SoundStream to process dolphin sounds into language-like tokens.
- The model will run on Pixel smartphones during a real ocean test in summer 2025.
What Is DolphinGemma?
DolphinGemma isn’t just another chatbot or AI tool. It’s a language model trained not for humans but for dolphins. It uses real dolphin sounds, recorded over decades, to try and spot patterns that might reveal whether dolphins have a language of their own.
🔹 Built on Gemma Models:
DolphinGemma comes from the same family of AI models as Google’s Gemini, but this version is fine-tuned to understand marine mammal communication.
🔹 How It Works:
Dolphins make lots of different sounds clicks, whistles, and even “squawks.” Using a special AI process, these sounds are broken into tokens (think of them like words or parts of words).
DolphinGemma then tries to predict what sound comes next, just like how chatbots predict what you’re going to type.
Watche video below:
🐬 How Dolphins Talk
Dolphins are super smart. They teach each other, work together, and even recognize themselves in mirrors. That’s something only a few animals can do.
Scientists already discovered that dolphins:
- Use signature whistles like names
- Make angry squawks during fights
- Communicate differently depending on their activity
But here’s the big question:
Is it really a language?
That’s where DolphinGemma might make history.
“We do not know if animals have words,”
said Dr. Denise Herzing of the Wild Dolphin Project (WDP).
“But if they do, we want to find out.”
🤖 How Google’s AI and Pixel Phones Are Helping Scientists Talk to Dolphins
🔸 Pixel-Powered Ocean Research
This isn’t just happening in labs. Researchers take Pixel phones underwater using a tool called CHAT (Cetacean Hearing Augmentation Telemetry), which lets them record dolphin sounds and even try synthetic replies.
The next upgrade?
In summer 2025, they’ll be using a Pixel 9-powered CHAT device, strong enough to run AI models in real-time while swimming with dolphins.
🔸 Small but Mighty
DolphinGemma is made up of 400 million parameters that’s small for AI but powerful enough to run on a phone. This is a big deal for researchers who need tech that’s fast, portable, and waterproof.
What Comes Next?
DolphinGemma could be the first step toward basic two-way communication with dolphins like matching a sound to an object or a behavior. Over time, that could build a kind of shared vocabulary between humans and dolphins.
Even better?
The AI model will be open to the public this summer, meaning scientists all over the world can train it with dolphin sounds or maybe even try it with other whales and porpoises.
This isn’t just about cool tech, it’s about redefining what communication means. If dolphins have a real language, and we can learn it, it could change how we think about animals, intelligence, and even our place on the planet.
Comment below:
🌊 Would you want to talk to a dolphin if you could?
“So how she take know say Nigeria is the light 🤣😭”