Earth Species, with Aza Raskin

What is the Earth Species Project? Can we talk to animals, or at least understand what they are saying to each other?  That’s a question that researchers hope to answer with the help of AI. Earth Species Project, a non-profit that develops sophisticated AI technologies, hopes its software can help. Specifically, they have developed Nature-LM audio which is an audio-language machine learning algorithm with the potential to decipher animal communications. How does it work? By gathering and evaluating huge amounts of audio information from different species, Nature-LM audio can identify  “individuals in recordings”, and evaluate patterns.  For software users, it does not require the user to have any programming skills. Specifically “analyzing animal sounds… [allows for] decoding complex communication and behaviors to monitor the health of entire ecosystems.” This AI model was trained on “bioacoustic archives like Xeno-canto, iNaturalist, the Watkins Marine Mammal Sound Database, and the Animal Sound Archive” along with “general audio, human speech, and music data” while connecting this ”audio encoder to a leading language model.” Benefits of this approach NatureLM “can classify or detect thousands of species across diverse taxa including birds, whales, and aurans–without the need to retrain the model for each task.”. It has other capabilities like “predicting life-stage and simple call-types of birds, and captioning bioacoustic audio” which are useful when trying to analyze the behavior of different species and their associated cues. The software enables evaluation of large amounts of animal sounds and allows evaluation of that data “freely via human language text”. According to a benchmark that they established, called the Beans-zero, which “provides a standardized way to measure… performance across various bioacoustic tasks, enabling consistent comparisons and fostering progress in the field”, NatureLM-audio “achieves state-of-the-art performance on most tasks”. This is especially true in regards to bird and marine mammal sounds, which they are able to identify without fine-tuning–an extremely gruesome task in machine learning to change pre-existing models which better fit your data and train it for specific tasks. Potential Drawbacks Like all AI models, Nature LM-audio could impact employment opportunities, in this case for  animal biologists and researchers, and by using substantial amounts of energy to run the model.  And, like all AI programs, any conclusions and decisions made through the program need to be carefully evaluated.  It will take time and effort to determine how valuable the model is.    Conclusion Raskin believes that the creation of NatureLM has many positive implications because it allows humans to listen to the voices of animals. It gives us an understanding of their behavior to not only learn more about them, but also by giving insights on how to help them with conservation efforts. Moreover, it can alert researchers to what exactly is endangering certain species, prevent these efforts, and create a lot of more data necessary to analyze trends.  About our guest Aza Raskin is a trained mathematician and a dark matter physicist and  Co-Founder/President of the Earth Species Project. ResourcesEarth Species Project, Introducing NatureLM-audio: An Audio-Language Foundation Model for BioacousticsGrantham Research Institute on Climate Change and the Environment, What opportunities and risks does AI present for climate action? Forbes, The 15 Biggest Risks Of Artificial Intelligence Further ReadingEarth Species Project, About UsUniversity of Cincinnati, 9 Benefits of Artificial Intelligence (AI) in 2025 For a transcript, please visit https://climatebreak.org/earth-species-with-aza-raskin