How AI helped scientists find an alphabet in sperm whale sounds
Scientists at the Massachusetts Institute of Technology (MIT) are using machine learning to research sperm whale communication habits -- and they're learning unprecedented things. Their new study, titled "Contextual and Combinatorial Structure in Sperm Whale Vocalisations," shines some light on the language sperm whales use to communicate.
The study focuses on whale vocalizations known as codas, or short bursts of clicks with various inter-click intervals. Instead of the more familiar music-like droning of the humpback whale, sperm whales communicate via clicks that sound similar to Morse Code.
Also: AI breakthrough enables scientists to read Roman scrolls once buried by Mount Vesuvius
Scientists have been analyzing sperm whale language for decades in search of some sort of pattern, but haven't learned much. According to the paper, past research has identified about 150 of these codas, but that's about the extent of humanity's knowledge in the area, until now.
The new study reveals that those clicks aren't arbitrary and that the language of sperm whales is much more complex than previously believed.
After using machine learning to analyze nearly 9,000 different codas from Eastern Caribbean whales, researchers discovered a "sperm whale phonetic alphabet" of sorts. When combined, those clicks essentially form words, the research paper explains, where things like rhythm and tempo give the clicks different meanings.
This shows, the study says, that context-sensitive sounds and words aren't just limited to humans.
Also: You can make big money from AI - but only if people trust your data
Of course, figuring out what that context means isn't easy. Let's say you happened to be analyzing human communication at a dentist's office. As an NPR article on sperm whale communication points out, you'd probably conclude that a word like "molar" was a frequent and important part of our discourse.
While we're not close to understanding or translating sperm whale speak, the discovery of these alphabet-like language building blocks shows just how useful artificial intelligence can be, especially in the scientific field.