Article: Can We Finally Learn What Animals Are Saying?

Photo Source: PickPik

People talk a lot. We talk in real time. We converse. We also talk in a delayed fashion. We listen to podcasts and the radio, watch TV, read, and write. Our languages help us connect, explain how we feel, describe our world, and cooperate to solve problems. Language represents an essential element of the human experience and has been found to have varied expressions around the world in different societies. Some languages have many words, and others much less. The Oxford English Dictionary has 273,000 headwords, while the Swedish dictionary has about 20,000 words.

People have studied human language for centuries, cataloging words in dictionaries and mapping the syntax or rules governing the order of words to make sentences. Languages in different societies differ in the sounds of words and the structure of sentences. Still, the commonality of words in sentences in human language has made the remarkable capabilities of deep learning systems like Google Translate large language models such as ChatGPT and Gemini possible. The predictable order of words in different languages provides the structure for computer models to learn languages and use that knowledge to translate one language into another. According to Weglot, a web translator, ChatGPT beat all the other computer translators for accuracy, including Google Translate and Gemini, in various languages, including French, Korean, and Tagalog, which is spoken in the Philippines. (weglot.com)

Animals talk a lot, too. Chirps, barks, clicks, and songs are among the many sounds that one can hear from birds, mammals, frogs, and insects during a walk outside, especially from spring through fall. Researchers, linguists, and other academics have argued whether these animal sounds constitute a language with words and syntax. Researchers in decoding animal communications have worked for years to try and map certain sounds to meanings. For example, dolphins communicate in part with a series of clicks, and researchers tried to determine if dolphins were conversing when working together to solve a puzzle. In the experiment, two bottlenose dolphins received a puzzle consisting of a plastic tube with lids on each end, and the lids had ropes attached to them. The tube contained dolphin treats such as fish or gelatin cubes. To free the treats, the dolphins needed to work together, pulling on a rope in opposite directions. The researchers recorded the sounds of the dolphins communicating with each other throughout the experiment. The researchers led by Holi Eskelinen found the cooperating dolphins produced more chatter while cooperating to solve the tasty puzzle. (springer.com) Although the content of their communication remains a mystery, the researchers concluded that the communication had different qualities compared to social chatter or communication when foraging. 

The success of large language models in human language translation has inspired animal communication researchers to try to apply the same models to animal communication. One such group called Earth Species Project has embarked on a mission to try and translate the communications of forty different animals, from crows to beluga whales. (earthspecies.org) It remains a challenge to truly understand any non-human animal communication. Still, these efforts to look at animal sounds with artificial intelligence open up the possibility of understanding the vocal communications of certain animals. However, with humans and our ongoing vocal communication, researcher Albert Mehrabian found that human communication consists of 55%  nonverbal, 38% vocal, such as tone and volume, and only 7% words. (utpb.edu) We should be very cautious when trying to understand animal communication only as words and syntax, for there may be vocabulary that transcends words that we need to know to truly understand animal languages.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.