Scientists Just Decoded Language of the Whales Using AI... And It's Not What You Think

A breakthrough for scientists after they had a 20-minute conversation with a humpback whale. Beneath the surface of the ocean, an ancient conversation is happening right now. Rapid clicks ripple through the dark water. Messages pass between giants with brains six times the size of our own. Using advanced artificial intelligence, scientists are listening in ways never before possible. And what they have uncovered is something nobody would have predicted. They are finding an entirely new way to understand communication between intelligent life on Earth. The Call of the Deep. Since humanity first ventured onto open water, the ocean has kept its deepest secrets. For centuries, sailors returning from long voyages brought back extraordinary tales of creatures they could barely describe, and of haunting, melodic sounds rising from somewhere beneath the hull. Those sounds were considered one of the sea's great marvels, beautiful and mysterious, but never truly examined. The animals that made them were admired for their size, but for most of human history, no one seriously considered that a whale might have a mind to match its body. For centuries, whales were resources, sources of blubber for lamps, bone for corsets, and little else. That perception held until one unexpected discovery changed everything. In the 1970s, the world was caught up in a new wave of environmental awareness. It was during this period of shifting consciousness that a quiet, scientific revelation would reframe the way we thought about marine life. A team of researchers studying humpback whales set out to sea with recording equipment, listening carefully to the sounds these animals produced. What they captured was astonishing. The moans and groans were not random noise. They were intricate, layered, and evolving, vocalizations that could last several hours and shifted in structure over time. Those recordings were compiled into an album called Songs of the Humpback Whale, produced by biologist Roger Payne and his colleagues. It was the beginning of something enormous. The release caught the public's imagination in a way that pure scientific papers rarely do, and it helped ignite a global conservation movement that pulled humpback whales back from the edge of extinction. The cultural impact was so significant that Whale Song was included on the Voyager Golden Record in 1977. The golden disc launched into space aboard the Voyager spacecraft as a message representing life on Earth to any civilization that might one day find it. Humanity chose whale voices as one of the sounds worth sending to the stars. That choice turned out to be more prophetic than anyone realized at the time. But as scientists went deeper into the field of bioacoustics, the study of sound produced by living creatures, they were left with more questions than answers. Early researchers spent countless hours with basic recording tools, listening and analyzing, trying to understand what they were hearing. Were whale songs mating calls, distress signals, a form of social communication? Nobody could say with confidence. For decades we were limited by the sheer volume of data and the boundaries of human perception. We could hear the sounds. We simply had no way to understand them. That is where a man named Dr. David Gruber arrived with an idea that changed everything. The Dream Team. Dr. David Gruber is an American marine biologist and National Geographic researcher who became convinced that understanding whale communication required a completely different toolkit. He did not look to the ocean for inspiration. He looked to space. Gruber drew on a program called SETI, the Search for Extraterrestrial Intelligence, an initiative built around listening for signals from other civilizations across the cosmos. The concept behind SETI is patient and audacious. If there is intelligent life out there, perhaps it is transmitting and perhaps we can learn to hear it. Gruber asked a question that reframed the entire field. What if alien intelligence was not out there but already here? What if it had been here all along, swimming beneath the surface of every ocean on Earth? The creature he had in mind was the sperm whale. The choice was not arbitrary. The sperm whale possesses the largest brain of any animal known to have ever existed on Earth, roughly eight kilograms, nearly six times the mass of a human brain. These animals can live for 60 to 70 years, dive deeper than two kilometers into the ocean, and maintain complex social bonds across entire lifetimes. They have been on this planet far longer than our species. Whatever intelligence looks like across evolutionary time, the sperm whale has had a great deal of time to develop it. This idea, radical, almost science fiction in flavor, became the foundation of Project CET-1, the Cetacean Translation Initiative. The team Gruber assembled was unlike anything the scientific world had seen. Marine biologists, computer scientists, linguists, and roboticists. On the surface, an unlikely combination. But understanding the voices of another species was not a problem that any single discipline could solve alone. It required everyone. At the heart of the project's biological knowledge was Shane Gero, a marine biologist who had spent more than a decade studying a single family of sperm whales near the Caribbean island of Dominica through his long-running Dominica Sperm Whale Project. This was not a job for Gero. It was a calling. He knew these whales individually. He watched them grow. He had given them names. Sperm whales live in matrilineal societies, tight-knit family groups led by older females, where calves are raised communally and knowledge is passed between generations over decades-long relationships. Gero had documented this social architecture with a depth no researcher had achieved before. That level of intimate, long-term field knowledge became the bedrock that everything else was built on. Equally important was what had recently become possible in technology. Natural language processing, the branch of artificial intelligence that allows machines to understand and interpret human language, had advanced dramatically. Before Project CETY, the idea of translating whale communication was essentially a fantasy. The data was there, but the tools to make sense of it were not. Then, around 2019, the team recognized that the technology needed to build an underwater recording studio had finally arrived. A facility that could, for the first time, realistically begin translating the language of sperm whales. The way Gruber describes it, this moment felt like the moonshot moment, the point at which something impossible suddenly becomes within reach. The island of Dominica became the center of the project's world. Its volcanic geography causes the ocean floor to drop dramatically 40 just off the coast, allowing sperm whales to come remarkably close to land, far closer than anywhere else in their range. A stable population of whale families returns to those waters year after year, and it was already the place where Gero had spent his career. The conditions were perfect. Project CETY officially launched in 2020, receiving 33 million dollars in funding from the TED Audacious Project. That investment allowed the team to construct what they called a 20 kilometer by 20 kilometer underwater listening and recording studio off the coast of Dominica. A system designed not just to hear the whales, but to understand who was speaking, to whom, and in what social context. This was the technological bridge between listening and comprehension. The oceans and the working technology. To capture whale communication at the scale needed for artificial intelligence to learn from it, the team needed something more than a underwater acoustic sensors, arranged in a grid pattern across the ocean floor. Think of a room lined wall to wall with microphones, sensitive enough to register a pin dropping from across the space. These hydrophones did the same thing across the open ocean, recording 24 hours a day, seven days a week, and crucially, tracking the precise direction and position of every sound. It was not enough to hear a whale, they needed to know exactly which whale was making the sound. The hydrophone network made that possible. It is worth pausing here to appreciate what these hydrophones were actually capturing. Sperm whale clicks are the loudest sounds produced by any animal on earth, capable of reaching 230 decibels. Powerful enough to be physically felt by anything in proximity. These clicks serve a dual function. They are used for echolocation, helping the whales navigate and hunt in total darkness at extreme depth, and they are used for communication. The same biological mechanism that guides a sperm whale through two kilometers of dark water is also the instrument through which it speaks to its family. The hydrophone network was built to catch both signals and begin separating one from the other, but the team went further still. Since there was no way to hand a microphone to a sperm whale, they deployed autonomous drones and robotic systems that could get close to the animals and place small non-invasive suction cup tags directly onto the whales' backs. These tags were loaded with sensors. They captured audio. They tracked movement at any depth. They recorded heart rate and logged social interactions with other whales. This combination was essential. The hydrophone network captured what was said and by whom, while the body tags provided the behavioral context. Was this whale calling out while it was hunting? While socializing with family? The behavioral data gave the AI the framework it needed to decode not just the content of the sounds, but their purpose. The scale of what was collected was staggering. Where previous research might have gathered a few hours or a few days of whale vocalizations, Project CETI was accumulating millions, eventually billions, of individual sounds. They were building the largest library of whale communication ever assembled. This was not a luxury. It was a necessity. Machine learning models require vast quantities of data to identify meaningful patterns. The more examples they study, the sharper they become. The scope of this archive was what separated Project CETI from every prior attempt to understand whale language. The data poured in constantly, a river of clicks, whistles, and calls rising from the deep. But raw data, no matter how abundant, is just noise. The hydrophones had listened. The tags had recorded. The language itself still waited. The next step was the one that transformed everything. The AI, breakthrough to a human ear. The underwater world is a blur of overlapping sounds. Clicks, creaks, surges of current, and the resonance of the ocean itself. Finding meaningful patterns in that noise is effectively impossible without help. The first major challenge for the Project CETI team was converting their raw audio into something the artificial intelligence could actually work with. They did this through a process called spectrogram conversion, turning sound waves into visual images. A spectrogram translates audio into a chart, where frequency and intensity over time become visible patterns. By converting whale sounds into images, the AI could literally see the structure of what it was hearing, rather than simply processing raw waveform data. From these images, the AI's first task was to identify the codas, the specific sequences of clicks that sperm whales use to communicate, while filtering out the constant background noise of the ocean. This was just the starting point. The team then turned to deep learning and neural networks, more advanced AI architectures designed to recognize patterns in ways that parallel how human brains learn. Crucially, the researchers applied transformer models, the same underlying architecture that powers large language models like the systems behind modern AI chatbots. These tools were originally developed to find structure in human language. The Project CETI team repurposed them for an entirely different kind of language, feeding them whale codas, the way other researchers feed them text. These systems were not given rules about whale communication. They were simply fed millions of codas and left to find structure on their own. Over time, the AI began identifying which sounds appeared together, how their patterns shifted across different situations, and what those shifts might mean. It was learning the way a child learns a language, through exposure, repetition, and pattern recognition, except it was doing this at a speed and scale no human analyst could approach. What the system found began to surface details invisible to the human ear. Subtle variations in rhythm, tiny differences in tempo, micro changes in the timing of clicks within a coda. None of these were perceptible to a human listener, but the AI could detect and categorize them with precision. The sheer volume of the data set, millions of vocalizations collected across years, gave the AI the raw material to identify patterns that would have remained completely hidden to any previous form of analysis. And then came the discovery that shattered an assumption nobody had seriously questioned. In findings published in Nature Communications in 2024, the team revealed something that rewrote what scientists thought was possible in animal communication. Sperm whales are using a communication system that functions like a phonetic alphabet. Not letters, but something structurally equivalent. The variables include rhythm, tempo, and what the scientists call ornamentation. Extra clicks added at the beginning or end of a coda that alter its meaning the way a suffix or inflection changes a word. A slow, steady rhythm might carry one meaning. Speed it up slightly or add an extra click at the end and the meaning changes entirely. The number of unique combinations this system allows is enormous. What the AI had uncovered was not a simple set of animal signals. It was the architecture of a language, complex, structured, and capable of expressing far more than anyone had imagined. For the first time, the team could see the building blocks of whale communication. They understood the syntax, the grammar. What they still did not know was what it all meant. Unveiling the whale's worldview. For a long time, the working assumption was that if whales communicated at all, their messages would be basic. Survival commands, warnings, simple needs. Generations of research had built toward expecting something straightforward. What the AI revealed was the opposite of straightforward. The patterns the system uncovered were not tied to simple actions. They were tied to social context. The whales were not just exchanging functional signals. They were having conversations. Complex, sometimes personal ones reflecting relationships and social bonds between individuals. It was not a code. It was a culture. The AI continued to dig deeper and the detail only increased. Within the clicks, there were subtle variations acting as functional equivalents to vowels in human speech. The team identified the use of rubato, fine-grained, deliberate changes in rhythm that shifted the meaning of a coda without altering its basic structure. A pause, a quickening, an added ornament at the end. All of these shaped what was being communicated. It was as if the basic units of their language could be bent and molded to convey emotion, intention, or nuance. A level of linguistic sophistication that science had previously assumed was uniquely human. The AI also revealed something that no researcher had predicted. The whales use dialects. Among the sperm whale families off Dominica alone, researchers have identified approximately 21 distinct coda types used by eastern Caribbean populations, each one a recognizable unit, like a word in a shared regional vocabulary. But different family clans adapt and vary these codas in ways that act as acoustic signatures of group identity. Variations subtle enough to be invisible to a human listener, but distinct enough that the AI could identify which family group was speaking based on the acoustic fingerprint alone. The same way a person from one region of a country speaks differently from someone raised elsewhere, whale families carry their own specific way of speaking. This language is not a universal code shared identically by all sperm whales. It is a living part of their culture, passed down across generations that can span 60 or 70 years, as distinctive and personal as a family name. A single coda could carry multiple meanings depending on its rhythm, its context, and the dialect it was spoken in. The word was not the message. The way it was spoken was. With AI as the translator, scientists were not just identifying sounds. They were gaining access to a culture that no one had known existed. Studies and research programs launched to examine the social connections, individual identities, and shared histories embedded in whale communication. Our understanding shifted from simple behavioral observation to something closer to anthropology, an attempt to understand how these communities work from the inside. But what could all of this mean for what comes next? The goal of two-way communication. For decades, science had been content to listen. Researchers deployed hydrophones across the oceans, recorded millions of whale clicks, and patiently tried to decode patterns hidden within the noise. But now, for the first time, listening was no longer enough. Having built an increasingly sophisticated understanding of how sperm whale communication is structured, the Project SETI team set their sights on something that would have seemed impossible only a few years ago, talking back, not in human language, in theirs. The team is developing a real-time communication system, what some researchers have compared to an underwater chatbot. The objective is not to teach whales English or force them to adapt to us. Instead, the goal is to use the patterns, rhythms, and structures that AI has uncovered to generate simple messages in sperm whale language itself. These messages are transmitted through specially designed underwater speakers, capable of reproducing the precise acoustic characteristics of natural whale codas. The researchers are deliberately starting small. Rather than attempting complex conversations, they are focusing on the simplest possible exchanges. A greeting, a response to a familiar coda, or an interaction tied to an object or event in the surrounding environment. If those basic exchanges can be reliably understood, they could become the foundation for increasingly sophisticated communication in the future. The ambition is unmistakable. Establishing the first genuine two-way dialogue between humans and another highly intelligent species. Achieving that goal requires more than technological breakthroughs. It demands extraordinary caution. When you begin building a bridge between two worlds, you also have to consider what might cross it. Should humans initiate communication with another species at all? Could repeated interactions unintentionally disrupt whale societies, alter natural behaviors, or even influence the evolution of their communication over generations? Unlike human languages, whale cultures have developed over millions of years without our interference. Any attempt to enter that world carries enormous responsibility. Project CETI has placed these ethical questions at the center of its mission. Marine biologists, linguists, ethicists, and conservation experts all contribute to developing strict guidelines for how and when playback experiments should occur. Any communication, if it happens, must happen on the whales' terms, in their language at their pace, and only when researchers believe the interactions are unlikely to cause harm. In 2023, the team demonstrated just how close this future may be. During a field expedition off the coast of Dominica, researchers used a specially calibrated underwater speaker to playback a sperm whale coda. A nearby whale responded with the same coda, creating a brief but remarkable back-and-forth exchange. While scientists are careful not to describe this as a full conversation, the response strongly suggested that the playback was recognized as meaningful communication rather than random underwater noise. For the researchers who had spent years building the world's largest collection of whale recordings and training AI to decode them, the moment was extraordinary. It represented one of the first real tests of whether humans could participate, however modestly, in the communication system of another species. It was only a single exchange, but sometimes history changes with a single exchange. That brief interaction suggested that the team's models were capturing genuine features of sperm whale communication. It marked an early step toward a future where communication between species may move beyond observation into interaction, a milestone that could reshape biology, linguistics, and our understanding of intelligence itself. The work of Project SETI is rewriting what we thought was possible in marine biology, cognitive science, artificial intelligence, and the study of intelligence itself. For most of human history, understanding an animal's inner life meant observing it from the outside. Scientists could document behavior, measure movements, record sounders, and make educated inferences about what those behaviors might mean, but the actual content of animal communication remained hidden behind an invisible wall. We could hear the sounds, but we could not understand the message. Artificial intelligence is beginning to change that. Instead of simply tracking where whales swim or how often they dive, researchers are now uncovering the structure of their social world. AI is helping reveal which whales maintain lifelong family relationships, how different clans preserve unique cultural traditions, how information may spread through communities, and how vocal patterns vary depending on context. What seemed scientifically impossible just a decade ago is becoming an achievable research program. The implications reach far beyond whales. The discovery that sperm whales possess a highly structured communication system, one containing rhythm, hierarchy, dialects, cultural transmission, and statistical properties resembling elements of human language, challenges one of humanity's oldest assumptions that complex language is uniquely human. Scientists are careful not to claim that whale communication is equivalent to human language, but it increasingly appears that sophisticated symbolic communication may have evolved independently in more than one branch of life on Earth. If that is true, intelligence itself may be far more diverse than we once imagined. The connection to space is impossible to ignore. The name CETI deliberately echoes SETI, the Search for Extraterrestrial Intelligence. For more than half a century, scientists have pointed radio telescopes toward the stars, searching for signals from civilizations beyond Earth. But detecting a signal is only the first challenge. The far greater challenge would be understanding it. That is exactly the problem Project CETI is solving. The machine learning systems being developed to identify structure within whale communication are designed to recognize meaningful patterns in data that humans cannot interpret directly. Those same approaches could one day help scientists analyze entirely unfamiliar forms of communication, whether from another species on Earth or, someday, from an extraterrestrial civilization whose methods of communication bear no resemblance to our own. In that sense, Project CETI is building more than a translator for whales. It is creating a framework for decoding intelligence wherever it may be found. Perhaps the most immediate impact, however, is much closer to home. For centuries, whales were viewed primarily as resources to exploit and, more recently, as endangered animals to protect. Even conservation efforts often focused on populations rather than individuals. But discovering that sperm whales maintain lifelong family bonds, inherit cultural traditions across generations, recognize one another through distinct vocal dialects, and engage in rich social interactions changes that perspective fundamentally. When we begin to understand that these animals are not simply making sounds but exchanging meaningful information with members of their families and communities, the ocean becomes populated not just by wildlife but by societies. That realization changes conservation from an abstract environmental issue into something deeply personal. Protecting whales is no longer just about preserving biodiversity. It becomes about safeguarding cultures, relationships, and forms of intelligence that evolved independently from our own over millions of years. And Project CETI is only the beginning. The team continues to expand its AI models with ever larger data sets while refining methods for interpreting increasingly complex communication patterns. Researchers also hope to extend similar techniques to other highly social cetaceans, including orcas, pilot whales, and dolphins. Around the world, a growing number of scientific collaborations are using artificial intelligence to decode the communication systems of birds, elephants, bats, primates, and countless other species. Each new discovery suggests that the natural world may be far richer in language and intelligence than humans ever realized. The picture of life on Earth is becoming larger, stranger, and more extraordinary with every passing year. The whales have been talking for millions of years. We are finally, perhaps for the very first time, beginning to understand what they have been saying all along. So what do you think? Is reaching across the species barrier one of the most important scientific achievements humanity could pursue? Or does attempting to communicate with another intelligent species raise ethical questions we are only beginning to understand? Share your thoughts in the comments below. And if you enjoy exploring discoveries that challenge how we see our world, subscribe and hit the notification bell so you never miss what we uncover next.