AI Translates Whale Communication

For decades, scientists have known that whales differ from other marine life in their complex social structures and vocalizations. However, a recent breakthrough by Project CETI has moved the needle from simple observation to actual decoding. By using advanced machine learning, researchers have identified what they call a “phonetic alphabet” within the clicking sounds of sperm whales. This discovery suggests that these marine giants possess a communication system far more similar to human language than we ever imagined.

The Role of Project CETI and AI

Project CETI (Cetacean Translation Initiative) is a nonprofit organization composed of marine biologists, cryptographers, roboticists, and linguists. Their primary goal is to apply advanced artificial intelligence to listen to and translate the communication of sperm whales.

In May 2024, the team published a landmark study in the journal Nature Communications. The study was led by Pratyusha Sharma, a computer science PhD student at MIT, alongside Shane Gero, the founder of the Dominica Sperm Whale Project.

The team utilized a massive dataset collected by Gero over nearly two decades. This dataset included recordings of about 60 sperm whales from the Eastern Caribbean clan off the coast of Dominica. While humans can hear the clicks, our brains are not fast enough or sensitive enough to detect the subtle micro-variations in timing and pitch. This is where Artificial Intelligence stepped in. The machine learning algorithms analyzed nearly 9,000 distinct codas (click patterns) to find structures that human ears had missed.

Decoding the "Coda"

Sperm whales communicate primarily through “codas,” which are short bursts of clicks that sound somewhat like Morse code. For years, scientists believed these codas were relatively static. They thought a specific pattern, such as “Click-Click-Pause-Click,” was simply a rigid call sign used to identify a whale or a family group.

The AI analysis proved this assumption wrong. The machine learning models revealed that these codas are part of a complex combinatorial system. This means the whales are not just repeating the same sounds; they are combining different auditory building blocks to create new meanings, much like how humans combine phonemes to make words, and words to make sentences.

The Four Elements of Whale Speech

The researchers identified four specific components that whales manipulate to change the meaning or context of a message. These components function remarkably like a phonetic alphabet:

• Rhythm: The core timing pattern of the clicks. This is the base layer of the communication.
• Tempo: The speed at which the clicks are delivered. The AI found that whales would produce the same rhythmic pattern but drastically change the speed to convey different information.
• Rubato: This was one of the most surprising findings. Rubato is a musical term referring to the smooth speeding up or slowing down of tempo. The whales were found to use rubato synchronously. Two whales would speed up or slow down their clicking patterns in perfect unison, suggesting a high level of active listening and coordination.
• Ornamentation: This refers to “extra” clicks added to the end of a standard coda. Researchers compare this to suffixes in human language. Just as adding “-ed” to a verb changes it to past tense, adding an ornamentation click might modify the intent of the whale’s message.

The Significance of "Duality of Patterning"

The identification of these four elements supports the theory that sperm whales utilize “duality of patterning.” This is a linguistic concept that was previously thought to be unique to human language.

Duality of patterning describes a system where a finite number of meaningless units (like letters or sounds) can be combined to create an infinite number of meaningful units (words and sentences). Before this study, animal communication was viewed mostly as non-combinatorial. A dog’s bark does not usually combine with a growl to create a third, distinct definition.

Project CETI’s findings show that sperm whales can combine rhythm, tempo, rubato, and ornamentation to generate a vast array of unique calls. The researchers estimate that these combinations could allow for thousands of distinct messages, far more than the simple emotional cues previously attributed to animal calls.

How the AI Was Trained

The technical execution of this project required a specialized approach to acoustic data. The MIT CSAIL (Computer Science and Artificial Intelligence Laboratory) team did not just feed raw audio into a translator. They visualized the audio data, turning the clicks into visual plots representing time and frequency.

By treating the audio as data points, the algorithms could measure the time between clicks down to the millisecond. This level of precision revealed that what humans thought were identical clicks were actually distinct variations.

The AI detected that the “rubato” effect was not random error or biological inconsistency. Because the whales performed these tempo shifts together during social exchanges, the computer determined the shifts were intentional features of the conversation.

Future Implications for Interspecies Communication

The roadmap for Project CETI extends beyond just analyzing existing tapes. The organization is currently deploying specialized underwater microphones and drones to capture continuous audio and video data in the Caribbean.

The ultimate goal is to move from decoding to interaction. Once the “grammar” is fully mapped, the team hopes to attempt a two-way communication experiment. This would involve broadcasting a synthetically generated sequence of clicks—using the correct tempo and rubato—to a whale to see if it responds in a meaningful way.

If successful, this would represent the first time humans have engaged in a complex, syntax-based dialogue with another species, fundamentally changing our understanding of intelligence on Earth.

Frequently Asked Questions

What is a sperm whale coda? A coda is a pattern of clicking sounds used by sperm whales for communication. It functions similarly to Morse code but with complex variations in speed and rhythm.

Did Project CETI prove whales have a language? The study proves that sperm whales have a phonetic alphabet and a combinatorial communication system, which are key prerequisites for language. However, we do not yet know the specific definitions of the “words” or “sentences” they are creating.

What technology did the researchers use? Researchers used advanced machine learning (AI) algorithms developed by MIT CSAIL to analyze datasets from the Dominica Sperm Whale Project. The AI detected patterns in timing and rhythm that were too subtle for human hearing.

What is the “rubato” finding? Rubato refers to the rhythmic speeding up or slowing down of a click pattern. The study found that whales do this intentionally and often in unison with other whales, indicating it is a deliberate part of their communication style.