AI will decode sounds and create a model that communicates with them.
NEW YORK (PNN) - October 28, 2021 - A team of international scientists recently launched an ambitious project to listen to, contextualize and translate the communication of sperm whales, with a goal of talking to the majestic marine animals.
The initiative, called Project CETI (Cetacean Translation Initiative), is harnessing the power of artificial intelligence to interpret clicking sounds, or “codas” sperm whales make to communicate with one another.
Researchers are using natural language processing or NLP - a subfield of artificial intelligence focused on processing written and spoken human language - which will be trained four billion sperm whale codas.
The plan is to have the AI correlate each sound with a specific context - a feat that will take at least five years, according to the researchers.
If the team achieves these goals, the next step would be to develop and deploy an interactive chatbot that engages in dialogue with Sperm whales living in the wild.
Michael Bronstein, the lead of machine learning for Project CETI, told Hakai, “If we discover that there is an entire civilization basically under our nose - maybe it will result in some shift in the way that we treat our environment; and maybe it will result in more respect for the living world.”
The sperm whale has a brain that is five times heavier than a humans and is also deemed the largest on Earth.
Although the brains might be different, both the whales and humans have a complex communication system and live in family groups.
“These [mammals] make a clicking sound at varying frequencies when they are in the company of other whales. The question is, is this just a simple code or a true language?” said Professor Dan Tchernov of University of Haifa's Leon H. Charney School of Marine Sciences who is one of the leaders of the project.
“Right now, our database is not comprehensive enough to know the answer to this question. However, with the advancement of machine learning, and advanced linguistics, we realized that if we gathered enough data about their voices, the context in which these sounds are employed and understood and their behavior and motivation behind these sounds, we can then develop an algorithm which will determine whether they have an authentic language.”
“CETI has laid out its five-year plan, which includes deploying dozens of state-of-the-art devices to listen to sperm whale communication, interpret their voices, and then attempt to communicate back,” Harvard John A. Paulson School of Engineering and Applied Science, which is part of Project CETI, shared in a statement.
“The interdisciplinary team of researchers will need to develop the most delicate robotic technologies to date to listen to and contextualize whale sounds; design a hydrophone array to study a population of whales; and build a big-data pipeline to examine the recorded data and decode it using advanced machine learning, natural language processing, and data science.”
The project, which started in 2020, includes experts from Harvard, MIT, Imperial College London, City University of New York, University of Haifa, UC Berkeley, and the Institute for Scientific Interchange.
Researchers have already started their work by developing non-invasive soft devices to attach to the whales and collect sounds and other data.
The tags will be unique in that must stick to the whale's skin without causing damage and have to withstand speeds up to 30 miles per hour, depths up to 6,561 feet and temperatures as low as a few degrees Fahrenheit.
A prototype is currently being tests, that uses suction cups inspired by octopus and squid - the whale's natural prey.
Using such designs has led to cups that can gently but firm latch on the whale.
Testing has only been conducted in a lab setting, which is promising, but the next step is to take the devices to whales in the open ocean.
The prototype houses up to three hydrophones that can provide context on the directionality of the communication between whales.
Once the team collects its four billion codas, the NLP will then develop 'a system analogous to human language models that generates grammatically correct whale utterances' that will be uploaded to the chatbot set for the open ocean.
However, the CETI researchers admit that their search for meaning in whale codas might not turn up anything interesting - but know the feat is worth exploring.
“We understand that one of our greatest risks is that the whales could be incredibly boring,” David Gruber, a marine biologist at City University of New York and program lead, told Hakai Magazine. “But we don't think this is the case. In my experience as a biologist, whenever I really looked at something closely, there has never been a time when I've been underwhelmed by
animals.”
