Why sperm whales
Sperm whales (Physeter macrocephalus) are unusually good targets for communication research. They have the largest brain of any species, by absolute size. They live in stable matrilineal clans with multi-decade social relationships. They communicate using clicks organized into stereotyped sequences called codas, and the codas vary across clans in patterns that some researchers describe as 'dialect.' The clicks are loud enough to be recorded at long range. The species occupies bounded ocean areas (CETI's primary site is around Dominica in the Caribbean) where the same individuals can be tracked over years. The communication system is structured enough to be plausibly decodable and limited enough in modality (just click sequences) to be tractable for analysis.
Several integrated workstreams.
What CETI is doing
Several integrated workstreams. Bio-logging — attaching suction-cup-mounted recorders to individual whales for short periods, capturing on-animal recordings of vocalization and behavioral context (this is methodologically related to the Demartsev[3] wearable-logger work on carrion crows). Hydrophone array deployment — fixed and floating microphone systems around the Dominica study site that capture the whale-vocal soundscape continuously. AI/machine-learning analysis — applying transformer-based and -based models to the captured audio to find structure in coda sequences, individual-signature patterns, and putative social-context correlations. Behavioral observation — surface-based observation of whale social groupings and behaviors to provide context for the acoustic data.
Who's involved
Shane Gero, a Dominica-based sperm whale researcher who has tracked the same family groups for two decades, anchors the field biology. David Gruber at City University of New York leads the broader project. Roger Payne, the whale-song researcher who recorded the famous humpback whale recordings in the 1970s, was a founding scientific advisor until his death in 2023. Yossi Yovel at Tel Aviv, an animal-communication researcher whose earlier work on bat vocalizations established methodologies CETI now uses, contributes. Earth Species Project[1] collaborates on the AI side. The Audacious Project provided most of the initial funding. The collaboration involves marine biology, signal processing, machine learning, linguistics, and conservation — a deliberately interdisciplinary structure that the field generally lacks.
What has been published
Several papers have come out of CETI's initial years. Identification of more structured combinatorial properties in coda sequences than previously appreciated, including evidence that some coda variations function as rhythmic-tempo modifications analogous to phonemes in human languages. Documentation of clan-specific coda patterns at finer resolution than earlier work showed. Demonstration that machine-learning embeddings of sperm whale codas can recover known clan structure from the audio alone, validating the methodology. These are real, substantial findings. They do not constitute 'translation' in the human-language sense, and CETI's own communication is careful about not claiming translation has occurred — the project is doing empirical structural mapping, not Rosetta Stone decoding.
What CETI realistically might achieve
The optimistic case — within the original 5-year project window — is a substantially improved structural understanding of sperm whale coda repertoire, documented dialect/clan variation, and empirical evidence about which coda patterns correlate with which observed behaviors. This is genuinely transformative for cetacean biology and would be the highest-resolution communication-structure mapping of any non-human species to date. The optimistic case is not 'translation.' Whether the species's communication includes referential or compositional content at all is the empirical question CETI is helping address; the answer might be yes, partially, or essentially no. All three answers are research-worthy outcomes; the field will know more by 2027-2028.
What this tells us about crow possibilities
Several lessons. First, the resource scale needed for serious animal-communication AI work is large — millions of dollars, multi-year commitments, dedicated species expertise, dedicated AI expertise, dedicated fieldwork. CrowLingo's atlas, at the scale of a single small project, is doing reference and education work; it is not doing original AI-bioacoustic research at the CETI scale. Second, the realistic outcome of even a well-funded program like CETI is structural mapping, not human-style translation. Third, the species-specific traits that make sperm whales tractable (bounded ocean populations, long-term identification, distinct communication units) are mostly absent for American crows (geographically dispersed, harder long-term tracking of individuals, less-discrete communication units), which means a CETI-equivalent for crows would face additional methodological challenges. The crow research that could happen exists as a real possibility; the funding and infrastructure required to make it happen don't exist yet.