By 2025 we will see AI and machine learning help make real progress in understanding animal communication, answering a question that has puzzled humans since ancient times: “What do animals say to each other?” The recent Coller-Dolittle Prize, which offers cash prizes of up to half a million dollars to scientists who “crack the code” is a sign of the high confidence that recent technological advances in machine learning and large-scale linguistic models (LLMs) are placing. the goal we hold.
Many research groups have worked for years on algorithms to make sense of animal sounds. For example, Project Ceti, has been decoding the clicks of sperm whale trains and the songs of humpbacks. These modern machine learning tools require very large amounts of data, and until now, such high-quality and well-defined amounts of data have been lacking.
Consider LLMs like ChatGPT that have training data available that includes all the text available online. Such information to communicate with animals was not available in the past. Not only are human data corpora many orders of magnitude larger than the kind of data we have access to for wild animals: More than 500 GB words were used to train GPT-3, compared to more than 8,000 “codas”. ” (or voices) of Project Ceti’s recent analysis of sperm whale communication.
In addition, when we work in human language, we already know what is being said. We even know what a “name” means, which is a huge advantage in explaining animal communication, where scientists rarely know that a wolf’s cry, for example, means something different from another wolf’s cry, or that wolves consider a cry to be somehow a “name” in human language.
However, 2025 will bring new developments, both in the amount of animal communication data available to scientists, and the types and capabilities of AI algorithms that can be applied to that data. Automated recording of animal sounds has been put within easy reach of all scientific research groups, with cheap recording devices like the AudioMoth exploding in popularity.
Large data sets are now coming online, as recorders can be left in the field, listening to the calls of gibbons in the forest or birds in the forest, 24/7, over long periods of time. There were times when such large data sets could not be handled manually. Now, new automatic detection algorithms based on convolutional neural networks can run through thousands of hours of recordings, select animal sounds and combine them into different types, according to their natural acoustic characteristics.
Once those large animal datasets are available, new analysis algorithms become possible, such as using deep neural networks to discover the hidden structure in animal vocalization sequences, which may resemble the logical structure in human language.
However, an important question that remains unclear is, what exactly do we hope to do with these animal sounds? Other organizations, such as Interspecies.io, stated their mission clearly as, “transferring signals from one species to corresponding signals of another.” In other words, you are translate animal communication in human language. Yet most scientists agree that non-human animals have no real language of their own—at least not in the way that we humans do.
The Coller Dolittle Prize is more complex, seeking a way to “communicate or clarify biological interactions.” Analysis is a less ambitious goal than translation, considering that animals may not, in fact, have a language that can be translated. Today we do not know how much information, or how little, animals transfer between themselves. By 2025, humanity will be able to leapfrog our understanding of not just how much animals say but what they really say to each other.