Major Breakthrough! Google's Model Deciphers Dolphin Language, Making Cross - Species Communication a Reality

Major Breakthrough! Google's Model Deciphers Dolphin Language, Making Cross - Species Communication a Reality

In this world full of wonderful discoveries, yet another astonishing piece of news has emerged! On April 14th, local time, which also happened to be World Dolphin Day, Google's CEO, Sundar Pichai, excitedly announced a groundbreaking achievement to the world - DolphinGemma was born! This model, trained based on years of dolphin sound data, has brought unprecedented hope for humans and dolphins to achieve cross - species communication.

The Magical DolphinGemma: Compact but Powerful

DolphinGemma has a special feature. Its parameters are only 400M, so small that it can directly run on the Google Pixel 9 phone, which is very convenient. Pichai also revealed that it is expected to be open - sourced this summer to promote global scientific researchers to work together to explore the mystery of dolphin language. This news instantly triggered a heated discussion among netizens. Everyone marveled at the wonder of technology, and many people directly stated that they were full of anticipation for this model that can enable humans to "talk" to dolphins. Interestingly, OpenAI had just released GPT 4.1 at that time. Some netizens humorously said that although GPT 4.1 is powerful, DolphinGemma seems to be more intriguing. This inadvertent "comparison" has added some fun to the technology circle.

How Google Cracked the Dolphin's "Language Code"

Google's AI - based model, DolphinGemma, is like a smart language learner. After careful training, it can learn the unique vocal structure of dolphins and generate similar sound sequences. Its principle is a bit like a large - language model predicting the next word. By identifying the input dolphin audio, it predicts the possible subsequent sounds, thus attempting to interpret the hidden meanings behind these sounds.

To make DolphinGemma "learn" the dolphin language, Google used its own unique audio technology. The SoundStream tokenizer is like an interpreter, which can quickly convert dolphin sounds into signals that computers can understand. Subsequently, a model architecture proficient in handling complex sequences takes over for analysis.

Of course, high - quality data is crucial for the model to learn accurately. Google has carried out in - depth cooperation with the WDP (Wild Dolphin Project). The WDP is an "expert" in studying the dolphin society. For decades, it has been focusing on studying a trans - generational community of wild Atlantic spotted dolphins (Stenella frontalis) in the Bahamas. Through this "the world's longest - running underwater dolphin research project", a large number of valuable underwater videos and audio materials have been accumulated, and detailed records have been made of the identity, life history, and observed behaviors of each dolphin.

For example, when a dolphin mother calls her calf, she will emit a signature whistle; when dolphins fight, pulse "squeals" are often heard; when courting or chasing sharks, "buzzing" sounds are commonly used. DolphinGemma is extensively trained based on these rich data, gradually "cracking" the dolphin's "language code".

New Exploration of Human - Machine Interaction: CHAT Builds a Communication Bridge

In addition to studying the communication methods of dolphins, the WDP has also cooperated with the Georgia Institute of Technology to develop an underwater application called CHAT (Cetacean Hearing Augmentation Telemetry). Its goal is not to directly interpret the complex natural language of dolphins, but to establish a simple shared vocabulary to build a communication bridge between humans and dolphins.

CHAT will associate the generated synthetic whistles with specific objects that dolphins like, such as seaweed, seagrass, or the scarves used by researchers. Researchers patiently teach curious dolphins to imitate these whistles to express their needs for these items. As the understanding of dolphin natural sounds deepens, more sounds will be added to this system.

The CHAT device uses underwater speakers and microphones to send and receive sounds. By imitating the whistles made by dolphins, it tries to establish basic communication with them. Currently, the Google Pixel 6 phone can already process high - fidelity dolphin sound analysis in real - time. The Pixel 9 phone, which will be launched this summer, will be further upgraded on this basis. The Pixel 9 integrates speaker and microphone functions and uses the powerful processing power of the phone to run deep - learning models and template - matching algorithms simultaneously, allowing researchers to analyze dolphin sounds more efficiently and interact with dolphins more complexly.

Why Dolphins? The Similarity of Languages is the Key

Some people may be curious why Google chose to study the language of dolphins instead of more common cats and dogs. Although Google did not clearly state the reason this time, from a large number of research materials, it can be found that the "language" of dolphins is highly similar to human language. A study published in the Proceedings of the Royal Society B: Biological Sciences shows that the way dolphins communicate with each other is similar to that of humans. When some dolphins make whistle - like sounds, these sounds are emitted by the vibration of specific tissues, and the operating principle is similar to the vocal cord vibration of humans and many terrestrial organisms.

Peter Madsen, the first author of the paper, once said: "In fact, they produce sounds through the resonance frequency of the connective tissue in the nasal cavity, and they can adjust the muscle tension and the air flow passing through at will. This is exactly the same as what humans do when speaking with their vocal cords."

Similar to dolphins, in fact, Google also launched a whale sound recognition model in September 2024, which can recognize the unique calls of eight species of whales and finely distinguish different vocal types of two of these whale species. Whales, like dolphins, also have similarities in their language to human language. A study published in the journal Science at the beginning of this year surprisingly found that by imitating the process of children learning language, there are statistical similarities between the songs of whales and the languages used by humans.

This discovery not only reveals part of the mystery of whale songs but also provides a new perspective for understanding cross - species communication and even AI language models.

Now, Google has brought humans and dolphins one step closer to communication through DolphinGemma. Perhaps in the near future, we can really have a free - flowing "conversation" with these lovely ocean elves. And Demis Hassabis, the co - founder and CEO of DeepMind, also revealed that maybe dogs will be the next to be studied. This also makes us look forward to the development of cross - species communication in the future with even more anticipation. Let's pay attention to the progress of technology together and witness more wonderful possibilities!