Artificial intelligence may be the missing piece needed to make the longstanding human dream of communicating with animals possible. Earth Species Project researchers have created NatureLM-audio, an AI audio language foundation model that can use animal vocalizations to identify various aspects of communication and behavior. NatureLM-audio is the first large audio-language model designed specifically to analyze animal sounds.
Trained on a curated dataset that includes human language and environmental sounds, the AI model can detect and identify the species of animals producing the sound, classify different types of calls, and predict the animal’s approximate life stage. NatureLM-audio has even shown potential in identifying vocalizations of species it has never encountered before.
This isn’t the first time generative AI has been applied for translation purposes. AI models have successfully translated human languages but have had more difficulty deciphering meaning from an unknown language. This makes translating animal languages trickier, especially since researchers are working with a limited understanding of how animals communicate through sound.
About the Earth Species Project
Earth Species Project is a nonprofit focused on addressing planetary concerns. It recently secured $17 million in grants to further its work using AI to decode animal communication. The organization aims to apply its large language model to improve our understanding of non-human languages, transform our relationship with nature, enhance animal and ecological research, and support more effective animal welfare and conservation outcomes. Advocates say using AI to decode animal communication may provide a compelling case for giving animals broader legal rights.
Applying generative AI to animal communication
In an April 1 post, ElevenLabs—creator of the speech-to-text model Scribe—announced its new AI tool, Text to Bark, suggesting it could help pet lovers enjoy similar AI-powered animal communication tools with their furry companions. The company said its AI-powered TTS model for dogs uses a new AI-powered “Pawdio” engine to support cross-species communication, turning human language into “fluent barking.” While the April Fool’s Day announcement was likely just for laughs, the Earth Species Project is not the only endeavor involving AI for animal communication purposes.
The nonprofit Cetacean Translation Initiative (CETI) is an interdisciplinary scientific and conservation project that applies AI to translating other-species communication. The listening project’s initial phase involves training its AI to decode the communication of sperm whales using a one-of-a-kind large-scale acoustic and behavioral dataset. With advanced machine learning and state-of-the-art robotics, CETI is working to protect the oceans and planet.