Since the release of the revolutionary album Humpback whale songs, a vinyl that hit 62 on the charts, featured in National Geographic, and even went to space on Voyager, humans (and maybe aliens) have been obsessed with whale song. It’s easy to see why when you listen to the haunting moans and clicks echoing through the water, even though we have no idea what they are saying.
But what if we could translate those clicks and whistles and moans, and – this is where it gets really wild – what if we could send messages back? Communicating with whales may seem out of reach (and like the start of the greatest disaster movie of all time), but it’s actually something that comes closer and closer to reality.
The interspecies conversation is being led by the Cetacean Translation Initiative (CETI) project, which started its ambitious project in March 2020, according to Hakai Magazine. The goal: to decode the songs of the whales, to establish their “language” and, hopefully, to reply. One can only hope that someone captures a video of the reaction of the first whale to find itself in conversation with an ROV.
The crazy or brilliant idea, depending on your desires to establish a story with a species whose intelligence could challenge ours (when was the last time you heard of a whale putting 15 soft-boiled eggs in its rectum? ?), was born out of a series of chance conversations (with humans, not whales).
It all started with computer scientist Shafi Goldwasser and marine biologist David Gruber as they discussed the similarities between sperm whale clicks and Morse code. They teamed up with computer scientist Michael Bronstein who argued that AI could be used to analyze mountains of sperm whale recordings to find patterns comparable to speech.
Research questions the headache-inducing concept that communication constitutes language and whether it actually exists outside of humans. In a recent interview with IFLScience, Dr Valerie Vergara spoke a lot about communicating with cetaceans, having devoted much of her career to listening to talkative belugas, known as “sea canaries” for their noisy nature.
As Vergara explained, it is known that young whales exhibit their own beluga “chatter” when trying to learn vocalizations from their parents and the larger pod. An emerging avenue of Vergara’s work focuses on recognizing unique voice signatures in beluga whales, the decoding of which could be essential in establishing whether they are speaking and whether these communications constitute conversation.
Taking the cat out of cetaceans and turning it into something we can analyze requires a lot of data processing, much more than human researchers, that’s where AI comes in. Language models like GPT-3 can effectively finish an unfinished sentence (or title) by learning what comes next, much like AutoCorrect. That said, GPT-3 still sometimes gets catastrophically wrong.
So we have a model and we have whales – what’s missing? This type of technology required around 175 billion words to function for human language, while the current bank of sperm whale “codas”, the term for the sperm whale equivalent of a word, stands at 100,000 relatively meager. So the next step is to dramatically increase the number of sperm whale records so that we can properly train the neural network of an AI.
Even once done, however, one wonders how the technology could be received by unsuspecting whales. “Maybe they would just answer, ‘Stop talking nonsense! “, Said Bronstein.
[H/T: Hakai Magazine]