Sperm whales are among the loudest living animals on the planet, producing crackles, bangs and jerky clicks to communicate with other whales that are within meters or even hundreds of kilometers.
This symphony of patterned clicks, known as codas, might be sophisticated enough to be considered a language in its own right. But will humans ever understand what these cetaceans are saying?
The answer may be, but early researchers must collect and analyze an unprecedented number of sperm whale communications, the researchers told 45Seconds.fr.
Related: Tale of 2 Tails: Why do sharks and whales swim so differently?
With a brain six times the size of ours, sperm whales (Macrocephalic physeter) have complex social structures and spend much of their time socializing and trading codas. These messages can be as short as 10 seconds or last longer than half an hour. In fact, “the complexity and duration of whale vocalizations suggests that they are at least in principle capable of exhibiting more complex grammar” than other non-human animals, according to an April 2021 article on sperm whales published on the preprint server. arXiv.org.
This article, carried out by an interdisciplinary project known as CETI (Cetacean Translation Initiative), presents a plan to decode sperm whale vocalizations, first by collecting recordings of sperm whales and then using machine learning to try to decode the click sequences of the latter. other mammals use it to communicate. CETI chose to study sperm whales in relation to other whales because their clicks have a structure close to Morse code, which artificial intelligence (AI) could have easier to analyze.
Breaking the surface
What little humans know about sperm whales has been learned fairly recently. It wasn’t until the 1950s that we noticed that they produced sounds, and it was not known that they used these sounds to communicate until the 1970s, according to new research published by CETI.
This click appears to serve a dual purpose. Sperm whales can dive to depths of 4,000 feet (1,200 meters), three times more than nuclear submarines, according to the Woods Holes Oceanographic Institution. Because it is dark at these depths, they evolved to search for squid and other sea creatures using clicks for echolocation, a type of sonar. This same click mechanism is also used in their social vocalizations, although the communication clicks are tighter, according to the CETI paper.
Understanding this has been a challenge, as sperm whales have “been so difficult for humans to study for so many years,” David Gruber, marine biologist and CETI project manager, told 45Seconds.fr. But now, “we actually have the tools to be able to look at this more in depth in a way that we haven’t been able to do before.” These tools include AI, robotics and drones, he said.
Pratyusha Sharma, data science researcher for CETI and doctoral student at MIT’s Computing and Artificial Intelligence Laboratory, explained to 45Seconds.fr recent developments in artificial intelligence and language models, such as GPT- 3, which uses deep learning to build human-like texts or stories on command, and the past year has taken the AI community by storm. Scientists are hoping these same methods could be applied to sperm whale vocalizations, she said. The only problem: these methods have a voracious appetite for data.
The CETI project currently has records of around 100,000 sperm whale clicks, painstakingly collected by marine biologists over many years, but machine learning algorithms may need around 4 billion. To fill this gap, CETI is setting up numerous automated channels for collecting sperm whale records. These include underwater microphones placed in waters frequented by sperm whales, microphones that can be dropped by eagle-eyed aerial drones as soon as they spot a group of sperm whales gathering on the surface, and even robotic fish that can follow and listen to whales discreetly from a distance.
But even with all this data, will we be able to decipher it? Many machine learning algorithms have found audio more difficult to analyze than text. For example, it can be difficult to analyze where a word begins and ends. As Sharma explained, “Suppose there is a word” umbrella “. Is that “uh” the word or is it “umbrella” or “umbrella”? The barriers between spoken words are more ambiguous and less regular, so patterns may require more data to be discovered.
This is not the only difficulty that CETI will encounter. “Whether someone is from, say, Japan or the United States or elsewhere, the worlds we’re talking about are very similar; we talk about the people, we talk about their actions, ”said Sharma. “But the worlds these whales live in are very different, aren’t they? And the behaviors are very different. “
Additionally, sperm whales are known to have dialects, according to a 2016 study in the journal Royal Society Open Science, who analyzed the codas of nine groups of sperm whales in the Caribbean for six years.
But these difficulties are also what makes the project so interesting. What exactly one sperm whale says to another remains as dark and murky as the waters in which they swim, but this mystery makes all the answers CETI find all the more intriguing. As Gruber said, “We learn so much when we try to see the world from another’s point of view. “
Originally posted on 45Seconds.fr.
45seconds is a new medium, do not hesitate to share our article on social networks to give us a solid boost. 🙂