By Julia Simpson
An international team of roboticists, cryptographers, and biologists, along with experts in machine learning, underwater acoustics, and linguistics, recently assembled to decode messages from the deep1. That may sound like the title-card montage for the next “team of international super spies” blockbuster, but no: this is the diverse, interdisciplinary expertise behind the Cetacean Translation Initiative (CETI). In April 2021, CETI, spearheaded by National Geographic Explorer David Gruber, officially launched operations in Dominica. Their ultimate goal? To decipher the language of sperm whales. Despite CETI being described by National Geographic2 as possibly “the largest interspecies communication effort in history,” key CETI collaborator Shane Gero, founder of the Dominica Sperm Whale Project, has emphasized that CETI’s direct aim is not to facilitate human-whale conversations, but rather to understand what sperm whales say to each other3. This approach – attempting to gain insight into the intricacies of animal communication outside of a human context – is a major shift4 from previous research that has investigated animal cognition and communicative abilities from a strictly anthropocentric perspective (e.g., teaching chimpanzees sign language). At this point, you may be thinking cool, but how? The short answer is, machine learning and natural-language processing, but more on that later, because if you’re like me, the first question that popped in your head after hearing about CETI was, why sperm whales, of all creatures?
To start with, in addition to being the proud bearers of the animal kingdom’s largest brains, sperm whales’ massive noses are also home to “the world’s most powerful biological sonar system4.” They communicate in different combinations of short clicks, called codas, which can be used for short- or long-distance information sharing. Though the bioacoustics initially evolved to facilitate echolocation, they have become adapted into something closer to what humans think of as a language. Sperm whales use codas to communicate information during wide varieties of social situations, and fascinatingly, distinctive coda dialects have been recorded from different sperm whale populations. The whales form clans that can consist of thousands of individuals, and these individuals will segregate themselves socially even when their ecological range overlaps with that of another clan. On a smaller scale, family units are matrilineal and highly cooperative; decisions are made as groups and young are raised communally. Strikingly, sperm whale calves spend their first two years producing “babbles” of codas and over time begin producing more structured codas in the dialect of their family group, a developmental process remarkably akin to that of language acquisition in human children. This extensive evidence of complex society-like organization among sperm whales, as well as intricate communication structures utilized by them, form the basis for CETI’s focus on sperm whales as they embark on their project4.
Now for the “how.” None of this would be possible without recent advancements in the natural language processing sector of machine learning, which now allow models to identify phonemes (sounds that we can put together to form words that carry meaning), morphemes (a language’s “smallest meaning-carrying units”), and grammar of a language, just from independently analyzing massive inputs of recorded language data. The real breakthroughs in cracking the sperm whale code will come when this information, once collected, can be combined further data adding behavioral and social context to this kind of ‘bare-bones’ language architecture4. This will allow the machine learning models to create predictive conversation programs, which CETI will then use to conduct “playback-based validation” experiments. With their predictive programs, CETI will construct a collection of codas with estimated meanings, play them aloud to sperm whales, and assess whether the response aligns with what they predict it will be4. If so, it would confirm the success of their programs in both estimating the meanings of sperm whale language elements, and in the ability to employ them in appropriate contexts (similar to how Alexa hears you when you make song requests, then knows to respond by playing that song2).
The biggest hurdles thus far in efforts to study cetacean communication boil down to how the ocean is big and deep and humans cannot breathe underwater, nor are we physically equipped to withstand great depths. These factors among many others make it incredibly challenging to study sperm whales at any kind of large scale, much less at the scale and level of detail required to gather enough audio and visual recordings to be analyzed by CETI’s natural language processing models. To overcome these historic difficulties, CETI is implementing a multifaceted approach using several cutting-edge technologies. These include establishing buoy arrays to pick up background acoustics in a region; tagging whales to track their movements, behaviors, and associations; and deploying both aquatic and aerial drones to monitor whale populations in an area. Current CETI researcher Daniela Rus is the director of MIT’s Computer Science and Artificial Intelligence Laboratory2, which in 2018, created the first robotic fish that can swim alongside real fish and record visuals for extended time periods at considerable depths5 . This type of surveillance – which would allow ‘incognito’ human observation of sperm whale activities – is supportive of Gruber’s vision for CETI and the application of its non-human-centric perspective on attempting to understand sperm whale communications.
A quote3 from Shane Gero, part of which I paraphrased earlier, kept resonating as I worked on this article. He wonders: “What’s important to whales and what do they say when they’re talking to each other? That really has the power to pivot people’s perspective about our obligation to the natural world.” More than anything, it’s that last part that keeps drawing my attention. I’m admittedly verging into speculative fiction territory here, but indulge me in a thought exercise: how would our collective understanding of our own societies shift if, a few years down the road, we end up decoding the language of sperm whales and it’s discovered that their levels of intelligence, consciousness, and societal structures are comparable to our own? What type of information would be valued during true inter-species communication? What would it mean if we had to reckon with not being the only intelligent species on Earth with legitimate claim to the planet’s resources? How would these realizations affect our subsequent behaviors in regard to ocean-related endeavors like shipping, travel, exploration, pollution, generation of renewable energy?
The interdisciplinary team of scientists behind the groundbreaking Cetacean Translation Initiative is embarking on an exciting mission: to study the behavior and communication patterns of sperm whales in hopes of deciphering their language.
1. CETI P. Team – Project CETI. https://www.projectceti.org/science-team.
2. Welch C. Groundbreaking effort launched to decode whale language. National Geographic. https://www.nationalgeographic.com/animals/article/scientists-plan-to-use-ai-to-try-to-decode-the-language-of-whales?cmpid=org=ngp::mc=social::src=twitter::cmp=editorial::add=tw20210419animals-whalecommunication&sf245094996=1. Published 2021.
3. Crawford B. Carleton marine biologist’s research key to Emmy-winning whale documentary. Ottowa Citizen. https://ottawacitizen.com/news/local-news/carleton-marine-biologists-research-key-to-emmy-winning-whale-documentary. Published 2021.
4. Andreas J, Begus G, Bronstein MM, et al. Cetacean Translation Initiative: a roadmap to deciphering the communication of sperm whales. arXiv. 2021:1-29. https://arxiv.org/abs/2104.08614.
5. Conner-Simons A. Soft robotic fish swims alongside real ones in coral reefs. 2018:1-9. https://news.mit.edu/2018/soft-robotic-fish-swims-alongside-real-ones-coral-reefs-0321#:~:text=In a paper out today,real fish in the ocean.