In 2025, we will see the apply of artificial intelligence and machine learning to make real progress in understanding animal communication, answering a question that has been plaguing humans since our existence: “What do animals say to each other?” Recent Coller-Dolittle Awardoffering cash prizes worth up to half a million dollars to scientists who “crack the code” indicates hopeful confidence that the latest technological advances in machine learning and huge language models (LLM) put this goal within reach.
Many research groups have been working on algorithms for understanding animal sounds for years. For example, Project Ceti decoded the clicking trains of sperm whales and the songs of humpback whales. These state-of-the-art machine learning tools require extremely huge amounts of data, and until now, such amounts of high-quality and well-annotated data have been lacking.
Consider an LLM such as ChatGPT which has training data available covering all text available online. Such information on animal communication has not been available in the past. It’s not just that human data corpora are many orders of magnitude larger than the data we have access to for animals in the wild: over 500 GB of words were used to train GPT-3 compared to just over 8,000 “codes” (or vocalizations) for the recent Ceti Project analysis of sperm whale communication.
Additionally, when working with human language, we already do this know what is being said. We even know what constitutes a “word,” which is a huge advantage over the interpretation of animal communication, where scientists rarely know whether, for example, the howl of a particular wolf means something different than the howl of another wolf, or even whether wolves consider the howl to be somehow analogous to “words” in human language.
Nevertheless, 2025 will bring up-to-date advances, both in the amount of animal communication data available to scientists and in the types and power of artificial intelligence algorithms that can be applied to this data. Automated recording of animal sounds has become within the reach of every research group, and low-cost recording devices such as the AudioMoth have become popular.
Massive datasets are now going online because recorders can be left in the field and listen to the calls of gibbons in the jungle or birds in the forest, 24/7, for long periods of time. There have been cases where such huge data sets could not be managed manually. Now up-to-date automatic detection algorithms based on convolutional neural networks can sift through thousands of hours of recordings, picking out animal sounds and grouping them into different types according to their natural acoustic properties.
Once these huge animal datasets become available, up-to-date analytical algorithms will become possible, such as using deep neural networks to find hidden structure in animal vocalization sequences that may be analogous to meaningful structure in human language.
However, the fundamental question that remains unclear is: what exactly do we hope to do with these animal sounds? Some organizations, such as Interspecies.io, have stated their goal quite clearly as “transforming the signals of one species into coherent signals for another.” In other words, to translate animal communication into human language. However, most scientists agree that non-human animals do not have a language of their own – at least not in the sense that we humans have a language.
Coller’s Dolittle Prize is a bit more sophisticated and looks for a way to “communicate or decode the body’s communication.” Decipherment is a slightly less ambitious goal than translation, given the possibility that animals may not actually have a language that can be translated. Today we do not know how much information, or how little, animals transmit between each other. In 2025, humanity will have the potential to leapfrog our understanding of not only how much animals talk, but also what exactly they say to each other.