AI-Guided Robots Are Now Tracking and Translating Sperm Whale Conversations in Real Time

The ocean is no longer silent to artificial intelligence. Deep off the coast of the Caribbean island of Dominica, autonomous underwater gliders are currently tracking sperm whales not by sight, but by listening to their conversations in real time.[1]

For decades, marine biologists knew that sperm whales communicated using rhythmic bursts of clicks known as "codas." However, continuous observation of these highly social mammals was nearly impossible. Sperm whales routinely dive to depths exceeding 1.6 kilometers, remaining submerged in total darkness for up to 50 minutes at a time before briefly surfacing.[1][2][3]

In May 2026, Project CETI (Cetacean Translation Initiative) announced a major breakthrough: the successful deployment of AI-guided robotic gliders that can identify specific whale voices and autonomously steer themselves to follow the animals for months at a time. The technology marks a shift from brief, opportunistic encounters to continuous, sustained observation of marine life.[1][8]

The gliders operate using a "backseat driver" software system. Four onboard hydrophones capture ambient ocean audio, while embedded machine learning models instantly isolate the distinctive codas from the noise of waves and distant ships. The software calculates the origin of the vocalizations and automatically directs the glider's navigation system to change course and follow the pod.[1]

This real-time tracking is paired with a new generative AI model called WhAM (Whale Acoustics Model). Built by theoretical computer scientists and marine biologists, WhAM does not just record the clicks; it predicts the sequence of codas that will follow. The system treats the vocalizations with the same architectural approach used for human language translation, identifying patterns that were previously invisible to human analysts.[3][7]

The deployment of these AI tools follows a series of landmark discoveries published earlier in the spring. In March 2026, researchers detailed a "phonetic alphabet" in whale communication, revealing that the animals use vowel and diphthong-like spectral patterns. The structure of these codas shows close parallels to the phonetics and phonology of human languages, suggesting an independent evolution of complex speech.[2][4]

The technology has already yielded unprecedented biological insights into the lives of the ocean's largest predators. Using high-resolution drone footage, computer vision, and multiscale network analysis, Project CETI captured the first quantitative evidence of cooperative birth assistance among non-primates.[4][5]

The AI-assisted analysis documented female sperm whales from two unrelated family units coming together to support a laboring mother and her newborn calf. Researchers observed coordinated lifting and caregiving behaviors, demonstrating a level of social complexity and cooperation dating back more than 36 million years.[4][6]

Despite these rapid advancements, AI researchers emphasize that translating whale codas presents a unique hurdle known as the "massive domain gap." Unlike translating French to English—where human experiences of the physical world heavily overlap—mapping whale concepts requires the AI to infer meaning from an entirely alien, aquatic environment.[3][7]

Project CETI has set an ambitious goal: to comprehend 20 distinct vocalized expressions—such as commands for diving, hunting, or sleeping—within the next five years. While the current models are focused on listening and pattern recognition, the team hopes to eventually achieve a basic level of two-way communication.[2][8]

As the AI models become more sophisticated, the scientific community is carefully weighing the ethics of "playback" experiments. Broadcasting AI-generated codas back to the whales is ethically fraught, as past acoustic experiments in animal behavior have sometimes caused distress, prompting whales to fall silent or alter their natural migration patterns.[3]

Beyond the philosophical implications of interspecies translation, the real-time AI tracking offers immediate conservation benefits. By continuously mapping how whales respond to human-made noise, the data can inform dynamic, evidence-based policy decisions.[1]

This could include implementing temporary fishing restrictions, rerouting commercial shipping traffic, or mandating speed reductions when gliders detect a socializing pod nearby. As artificial intelligence continues to decode the deep ocean, the gap between human and cetacean intelligence may prove narrower than ever imagined, offering a powerful new tool to protect one of the planet's most enigmatic species.[1][8]