One afternoon in January 2024, off the island of Dominica in the Atlantic Ocean, a sperm whale named Pinchy and I drifted calmly at the surface of the water, examining each other.
After several minutes, Pinchy took a long breath, pointed her head downward and slowly began her descent into the ocean. As her 40-foot-long silhouette disappeared, I considered the distances in our worlds. We are both mammals whose time underwater relies on borrowed breaths of air from the surface. But Pinchy had a life that existed a mile underwater; while I lived in a metropolis where buildings stretched toward the sky. I wondered what Pinchy might tell me if I could understand what she was saying.
As a marine biologist studying how whales communicate, my dream is to one day answer that question. My team and I have been able to discern that sperm whales have their own alphabet and that this alphabet seems to be pillared by their own version of vowels. I've learned that humans are far from the only species intelligent and complex enough to develop a form of language and culture. At a time when living in a technology-fueled civilization may make us humans feel more distant from the natural world, this discovery helps me feel more connected to it.
Paradoxically, that's been made possible thanks to the use of technology itself, especially artificial intelligence. What I've come to know is that technology and nature do not exist in a zero-sum universe where the ascendance of one side is the downfall of the other. Instead, these tools can give humans an opportunity to feel more tethered to the flora and fauna that surround us.
That revelation has been many years in the making. I've spent my career designing technologies that are meant to see and hear from the perspectives of marine creatures, which led me to sperm whales. They are capable of making some of the loudest and most complex sounds in the animal kingdom. When socializing, they emit a series of clicks, called codas.
Several years ago, A.I. technologies like large language models (the systems that power tools like ChatGPT) began to demonstrate an ability to predict word patterns and formulate new sentences on their own. I started to connect with computer science colleagues who wondered if these models could be applied to sperm whale codas. The whale biologist Shane Gero, who has been studying sperm whale families for two decades in the eastern Caribbean, had an annotated sperm whale data set featuring thousands of codas. In a pilot study we ran, A.I. that was fed the data was able to accurately predict the type of coda, the whale's vocal clan and the individual whale with over 90 percent accuracy.