CETI project (The whale translation initiative) is an initiative that aims to decode the “language” of sperm whales. Using artificial intelligence, the team behind the project hopes to decode the clicks (or clicks) that sperm whales use to echolocation and communicate with each other. Eventually, they might develop a language to talk to these sea giants.
Clicks are organized into standard sequences called codes. To decode this string of sounds, the researchers plan to exploit automatic natural language processing (natural language processing or NLP) – a subfield of artificial intelligence focused on processing written and spoken human language. The team has already applied the recordings of koda sperm whales to an NLP algorithm, with promising results.
Why sperm whales in particular? Aside from having the largest brains of any species, these animals display traits similar to humans. They are endowed with conscious thought and the ability to plan, speak and feel; They are communal animals that live in groups of 20 to 40 individuals, with strong family ties. Their sophisticated vocal communications are a great starting point for advanced machine learning tools that can be applied to other animals.
Objectives: Gather and contextualize data
The main obstacle to be overcome in this project is sufficient data collection; Machine learning requires a very large set of data for training and model building. The team’s goal is to successfully collect four billion “words” from sperm whales! Initially, for the first proof of concept, he relied on his research Dominica Sperm Whale Project, who have collected just under 100,000 code plus valuable information on the social life and behavior of cetaceans.
Roughly 100,000 kodas may seem like a lot, but they’re actually too little for the task. For comparison, GPT-3 – Predictive Language Model for Deep Learning Developed by OpenAI and released in 2020 – Trained with about 175 billion parameters (ten times more than any language model previously developed)!
Another point of difficulty: putting all the code in context. In human language, words can have different meanings depending on the context, or even have no meaning at all; The same is true for other languages, including the tapping of sperm whales. However, it will undoubtedly take years of research dedicated to studying these cetaceans in their natural habitat to correlate each sound with a specific context.
The CETI project brings together cryptologists, robotics, linguists, AI experts, technologists, and marine biologists from universities around the world to carry out this arduous task. Note that in 2020, the team had organized Decoding workshop Contact in non-human species for Simmons Institute for Computing Theory, where experts who study non-human communication across a variety of species shared their research; It was an opportunity to gather a lot of information on this topic.
Towards a greater respect for the living world
But do animals really have a language? The question is still debated among the scientific community. Many believe that language is exclusive to humans. According to Austrian biologist Konrad Lorenz, one of the pioneers in animal behavior, “ Animals do not have language in the true sense of the term In other words, they communicate, but do not speak. Karsten Brensing, a German marine biologist who specializes in communicating with animals, believes, on the contrary, that the exchange of many animals can be described in languages.
For this, several conditions must be fulfilled according to it: semantics (for meaning), grammar (for sentence construction) and vocabulary learning (for this to be considered as a language, all the sounds that the animal produces do not have to be innate). And some animals (certain species of birds, or even dolphins) have already demonstrated this ability to learn, respecting the other two conditions. Sperm whale clicks seem ideal candidates for trying to decipher their meaning — not least because it’s easier to translate as zeros and one of the continuous sounds that other whale species produce.
If the team achieves their goals, the next step will be to develop an interactive chatbot that attempts to engage in dialogue with sperm whales living in the wild – a feat that could completely change the way humans perceive and interact with nature. The researchers admit that their research may also not reveal anything interesting, in other words, whales can turn out to be incredibly boring. ” But we don’t think that’s the case. In my experience as a biologist, whenever I look at something up close, there has never been a moment when I have been disappointed by the animals. “,” David Gruber said:, Project manager.
The team specifies that CETI data will be published for mutual collaboration. She hopes that the discoveries made on sperm whales will provide a basis for a better understanding of the communication of other animals in the ocean and on land: elephants, birds, gorillas and more. ” If we discover that there is an entire civilization under our noses, it could lead to a change in the way we deal with our environment. This may result in a greater respect for the living world Michael Bronstein, Head of Machine Learning for the CETI Project explains.