- Google and the Wild Dolphin project have developed an AI model formed to understand the vocalizations of Dolphins
- Dolphingemma can operate directly on Pixel smartphones
- It will be open source this summer
For most of human history, our relationship with dolphins was a unilateral conversation: we speak, they grine and we hochons their heads as we understand each other before throwing them a fish. But now Google has a plan to use AI to fill this division. In collaboration with Georgia Tech and the Wild Dolphin Project (WDP), Google has created Dolphingemma, a new model of AI trained to understand and even generate chatter of dolphins.
The WDP has been collecting data on a specific group of spotted dolphins from the wild Atlantic since 1985. The Bahamas based pod has provided enormous amounts of audio, video and behavioral notes as the researchers have observed them, documenting each Squawk and Buzz and trying to collect what it means. This audio treasure is now introduced into Dolphingemma, which is based on the family of Google open gemma models. Dolphingemma takes dolphin sounds as a starter, treats them using audio tokenizers like Soundstream and predicts what vocalization could then come. Imagine self-automatic, but for dolphins.
The model is very thin and can work on a Google pixel. The WDP is already deploying Dolphingemma on the ground this summer, using 9 pixels in waterproof platforms. These models will listen to, identify the voice models and help researchers report significant sequences in real time.
To watch
Flipper speaks
But the ultimate goal here is not only a passive listening. WDP and Georgia Tech also work on a system called chat (abbreviation of hearing aid telemetry cetacean), which is essentially a bidirectional communication system for humans and dolphins. Cat allows researchers to attribute synthetic whistles to objects such as dolphins, including sea herbaries and floating scarves, then wait to see if the dolphins imitate these sounds to ask them. It’s a bit like inventing a shared language, except with underwater microphones instead of flash cards.
Dolphingemma not only analyzes the sounds of dolphin afterwards; It helps to anticipate the sounds that could happen, which allows you to make faster response times and smoother interactions. Essentially, it is like a predictive keyboard for dolphins. The entire project is still at an early stage, but Google plans to open Dolphingemma later this year to accelerate progress.
The initial model is formed on the vocalizations of the Spotted Dolphins of the Atlantic, but it could theoretically be adapted to other species with an adjustment. The idea is to give other researchers the keys to AI so that they can apply it to their own acoustic data sets. Of course, it is still far from chatting with the dolphins of philosophy or their favorite snacks. There is no guarantee that the vocalizations of dolphins perfectly map human language. But Dolphingemma will help pass over the audio years for significant models.
Dolphins are not the only animals that humans can use AI to communicate. Another group of scientists has developed an AI algorithm to decode the emotions of pigs according to their grunts, cries and smoking to help farmers understand their emotional and physical health. However, dolphins are undeniably more charismatic. Who knows, maybe one day you can ask a dolphin for directions while you are sailing, at least if you don’t let your phone fall in the water.
