A Bold Scientific Frontier

For decades, researchers have recorded, catalogued, and marveled at the complexity of whale vocalizations. Now, a new generation of scientists is asking a more audacious question: could we actually decode what whales are communicating — and perhaps one day respond in kind? Thanks to advances in artificial intelligence, natural language processing, and underwater recording technology, this question has moved from science fiction to active research agenda.

The CETI Project: Listening at Scale

One of the most ambitious efforts in this space is Project CETI (Cetacean Translation Initiative), a multidisciplinary research collaboration launched in 2021. CETI's goal is to record billions of sperm whale clicks (codas) using an array of underwater hydrophones, biologging tags on individual whales, and eventually underwater robotic systems — and then apply the same large-scale machine learning techniques that enabled breakthroughs in human language AI to find patterns, structure, and potentially meaning in whale communication.

The core hypothesis is that sperm whale codas may encode far more information than previously recognized — potentially including identity, emotional state, social context, and behavioral intent. By training AI models on massive datasets of annotated coda recordings, researchers hope to build a kind of "rosetta stone" for cetacean communication.

What Makes Sperm Whale Codas Interesting for AI Analysis

Sperm whale codas have several properties that make them well-suited to computational analysis:

  • They are discrete and rhythmic — composed of identifiable click patterns with countable elements, unlike the continuous tonal songs of humpbacks.
  • They are produced in social contexts — exchanged between known individuals within documented social structures, allowing researchers to link acoustic data to behavioral observations.
  • They show cultural variation — different clan groups share distinct coda repertoires, suggesting a learned, socially transmitted communication system rather than purely innate signals.
  • There is a growing archive of recordings — decades of field work by researchers including Shane Gero (the Dominica Sperm Whale Project) have produced annotated datasets that can serve as training data.

The Challenges of Non-Human Language

Translating whale communication is not simply a matter of applying existing AI tools. Several fundamental challenges complicate the effort:

  1. No ground truth — With human language AI, models can be trained on text with known meanings. With whale communication, researchers don't know in advance what any given coda "means" — the meaning itself must be inferred from patterns and context.
  2. Different cognitive architecture — Whale "language," if it exists in a meaningful sense, almost certainly doesn't map neatly onto human linguistic concepts like nouns, verbs, or declarative statements.
  3. Data collection difficulty — Recording individual known whales in their natural social context at sufficient quality and scale is extraordinarily challenging logistically.
  4. Signal complexity — Sperm whales can produce multiple simultaneous click trains, and codas overlap with echolocation clicks in ways that are difficult to separate.

What Would "Translation" Even Mean?

Many researchers are careful to temper expectations. The goal of projects like CETI is less likely to produce a phrase-by-phrase translation and more likely to reveal the structural complexity of whale communication — whether it has the combinatorial richness, context-dependence, and cultural variation that characterize human language. Even establishing that whale communication is "language-like" in these senses would be a landmark scientific finding.

Some researchers go further, envisioning eventually playing synthesized codas back to whales and observing responses — a rudimentary form of two-way interaction that could help validate interpretations of what specific patterns communicate.

Why It Matters Beyond the Science

The broader significance of this research extends beyond pure curiosity. Demonstrating that whales possess genuinely complex, culturally transmitted communication — possibly approaching language — would have profound implications for how we think about cetacean intelligence, moral status, and the urgency of conservation. If we are, in some meaningful sense, neighbors in a shared world of communicating minds, the case for protecting them becomes not just ecological but ethical.

The whales have been singing for millions of years. We're only just beginning to learn how to listen.