What do sperm whales “talk” about? Artificial intelligence helps decipher their voices
Sperm whales have the largest brains in the animal kingdom and complex social behavior in which members of families and cultural groups exchange a series of clicks to communicate and locate their prey. So far, some of these clicks have been shown to convey their own personality, as if each one had its own name, but Pratyusha Sharma’s team’s discovery goes much further: artificial intelligence has allowed them to demonstrate that these animals’ vocalizations are more complex than thought earlier, and contain structures similar to human phonetics.
For work published on Tuesday in the journal Natural communicationsSharma and his team from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Whale Translation Initiative (CETI) analyzed the vocalizations of about 60 cetaceans living in groups in the eastern Caribbean using machine learning systems. The authors claim to have deciphered for the first time what they call the “sperm whale phonetic alphabet”, which allows them to begin deciphering their messages, although its contents are still unknown.
“The short series of clicks that these cetaceans use to communicate with each other have both a combinatorial structure and context-sensitive call modulation,” Sharma explains to elDiario.es. “Combinatorial vocalization systems are rare, and their presence indicates that in principle this species is capable of representing a broader message space.”
These cetaceans communicate using codes, which helps us understand how they can encode information in their calls.
The researchers analyzed up to nine thousand codes collected from families of Eastern Caribbean sperm whales monitored by the Dominica Sperm Whale Project, using a combination of pattern recognition and classification algorithms and body recording equipment. And it turned out that sperm whale communications were not random or simplified, but were structured in a complex and combinatorial manner.
Codes that change meaning
The basis of the communication system is what scientists called codes, short series of clicks, the structure of which varies significantly depending on the context of the conversation. “Sperm whales communicate using click sounds, and their communication signals are produced by stringing together a short burst of clicks,” explains the lead researcher. “Understanding what aspects of these codes they can control and change helps us understand how they can encode information in their calls.”
The result, the lead author adds, shows that these codes are not random, but are created by carefully combining four characteristics: rhythm, tempo, rubato (a gradual change in the same coda) and ornamentation (an “extra click” added to the end of the code in a group shorter signals). “We found that sperm whales can modulate one characteristic (overall duration) of their call while keeping another (rhythm) constant,” says Sharma. “They have extra clicks that we call decorations that sit on top of different codes. They most often occur towards the beginning and end of a course of exchange, acting as markers of discourse.”
Individual nuances and variations
Experiments were conducted using acoustic biological tags to reveal the intricate details of these animals’ vocal patterns. By developing new data analysis and visualization techniques, CSAIL researchers discovered that individual sperm whales can emit multiple code patterns over long-term exchanges, rather than simply repeating the same code. These patterns, the study authors say, are nuanced and include detailed variations that other sperm whales also produce and recognize.
We venture into the unknown to unravel the mysteries of sperm whale communication without any real pre-existing data.
The similarity to complex human language, Sharma explains, comes from the way they use two properties: context and combinatorics. The first manifests itself in testing that some aspects of sperm whale calls become apparent only in a broader communicative context. “In humans, speech tone has contextual variations,” he says. “In sperm whales, it is rubato and ornamentation that become apparent in a broader communicative context.”
In terms of combinatorics, he continues, people combine sounds like “æ,” “p,” “l,” and “É™,” which don’t mean anything on their own, to create words like “Manzana(apple) that convey meaning. From now on we can arrange words to create sentences, for example: Red apple on the tableto convey more complex values. “In sperm whales we see a combination of two levels of characteristics, which we call rhythm, tempo, rubato and ornamentation, forming codes that are then ordered in exchange,” he says. “We still don’t know at what level and what functions they are going to carry out or even what they are talking about. However, we see that they have a combinatorial mechanism by which they collect calls and are in principle capable of expanding the message space.”
Specific communicative intentions
The use of machine learning was necessary to determine the characteristics of their communication and try to predict what they would say next. “We are venturing into the unknown to unravel the mysteries of sperm whale communication without any real pre-existing data,” says Daniela Rus, director of MIT CSAIL and co-author of the paper. “Our results indicate the presence of structured information content and also challenge the prevailing belief among many linguists that complex communication is unique to humans.”
According to Roos, this is a step toward demonstrating that other species have levels of communicative complexity that have not yet been identified and are deeply linked to behavior. “Our next steps are to decipher the meaning of these messages and explore correlations at the social level between what is said and group actions,” he notes.
The authors admit they are still largely unaware of what sperm whales communicate, but they believe this line of research may one day hold the key to understanding it. “We are studying how to interpret the basic meaning units of their communication,” defends Sharma. “Essentially, our work could lay the groundwork for figuring out how an ‘alien civilization’ might communicate, providing insight into creating algorithms or systems for understanding completely unknown forms of communication.”
Step towards decrypting your messages?
Antonio Jose Osuna MascaroDoctor of Biological Sciences, an expert in animal behavior, finds the study very interesting, although he asks to evaluate it in context. “It is important to clarify that we still know very little about the semantics associated with cetacean communication in general,” he warns. According to him, with the proposal of a sperm whale phonetic alphabet that would be in some way similar to the phonemes that humans use to construct our language, we now know that they have the potential to create messages rich in information, but we still don’t know , the extent to which what they communicate exploits this potential. “This is very important for science,” he concludes, “because we know almost nothing about the semantics of cetacean communication in general, and if we want to confront this problem and one day solve it, we need to know that they are able to communicate.”
This obviously opens up the possibility for animals to encode information by combining elements, but equating this to human language is extremely speculative.
According to Javier Almunia, an expert in cetacean bioacoustics, if confirmed in follow-up studies with other clans with different codes, it would be a significant advance in the study of communication in this species. “The result points to a new level of detail in encoding information using codes known since the mid-1960s,” he says. If these combinations encode information, he believes, then the very difficult work now lies ahead: analyzing the different codes and seeing their effect on the animals’ behavior, which is especially difficult as they descend to depths of more than 2,000 meters for periods of more than one hour.
From their point of view, what is much more speculative in this work is the constant comparison with human language, which they repeat several times throughout the article. “Obviously, this opens up the ability for animals to encode information by combining the elements they describe that are associated with rhythm and pattern code,” he says. “But equating it with human language is extremely speculative.”
Michel Andre, one of the world’s leading experts in bioacoustics, based at the Polytechnic University of Catalonia (UPC) and whose work on sperm whale communication paved the way in 2000, agrees with the same sentiment. We published 20 years ago: “When.” you study the content of information in the sounds of these sperm whales, clicks, you understand that they differ greatly from the moment they are published to the moment they are received, they reach a distance of up to 15 km, which is the limit of communication with other groups,” he recalls. “So we came to the conclusion that the only constant that holds regardless of distance and depth is the time intervals between sounds, and that’s where the meanings might be.” In other words, if there is encoding of information, then it must be in rhythms that are not disturbed during transmission.
As for this new result, he finds it very interesting, although he thinks it would be a mistake to try to make these codes look like our own language. “We recognize that oral communication exists because there is strong coordination among members of a social group,” he says. “But trying to understand the communication mechanism of an animal whose brain has evolved around acoustic communication over 30 million years by comparing it to our way of producing sounds and communicating is likely to fall into an anthropomorphism that will distance us from reality.” .