Earth Species Project
We are an open-source collaborative and nonprofit dedicated to decoding non-human language.
Two sibling Gelada Monkeys (Theropithecus Gelada) Location: Simien Mountains National Park, Ethiopia Photo: Marc Guitard
We believe that an understanding of non-human languages will change our ecological impact on this planet.
We are inspired by the incredible diversity of communication systems on Earth. From the mycelial networks that connect Earth’s forests like an Internet to the rainbow-rave richness of cuttlefish communication, we are surrounded by messages and meaning.

We are motivated by the recent monumental milestone in machine learning: the invention of techniques that can translate languages without dictionaries.

Like the telescope, these new tools will forever change our perspective.
Our technical roadmap is available on Github.
Annual Migration of Giant Cuttlefish (Sepioteuthis lessoniana)
Location: Point Lowly, South Australia
Photo: Mick Valos

Our Story

Britt Selvitelle and Aza Raskin met in 2007. At the time, Britt was on the founding team at Twitter and Aza was helping start Mozilla Labs. They instantly connected over their shared curiosity and respect for how technology changes the way we relate to ourselves, each other, and the natural world. They’ve stayed close ever since.
In 2013, the seed for the Earth Species Project was planted when Aza heard an NPR story on the uncannily human-like vocalizations of the Gelada monkey. Thore Bergman, the evolutionary biologist at University of Michigan heading the study, described it:
“We noticed sounds like people were talking around you. Most primates only make a few sounds, but geladas produce a complex stream with a rhythm similar to language.”
— Hints of Human Language Heard in Lip-Smacking Monkey Talk, Wired, April 8, 2013
At the same time, a new machine learning technique emerged that could learn a geometric representation of an entire language, like a galaxy where each star is a word and the distance and direction between stars encodes relational meaning.
The geometric shape of English, or at least its top 10,000 most spoken words. This shape is actually in hundreds of dimensions. This structure is known as a “latent space”.
By 2017, we were meeting once a week to figure out how to fit the puzzle pieces together. We discovered two recently published papers demonstrating a machine learning milestone: techniques to translate an unknown language without the help of a single example.

The insight: not only can languages be represented as geometric structures, they can also be translated by matching their structures to each other.
Human languages can be translated without dictionaries or examples of translation by matching their latent geometric representations — their shapes.

To dive deeper, see the 2018 research by M. Artetxe and G. Lample.
Inspired by the work of numerous ethologists and researchers, we left our computers (mostly) to venture into the field and learn firsthand from those with decades of wisdom in bioacoustics and animal communication. 

Since then, we’ve begun work on cetacean and primate communication. Over time our focus will broaden to corvids, and others. Today, we've partnered with over fourty biologists and institutions and made progress in solving some of the cornerstone problems along the path to decoding non-human communication.

We founded the Earth Species Project in the vein of the organizations whose values and impact in the world have inspired us, like the Internet Archive and Mozilla. 

We are a 501(c)3 nonprofit and are committed to releasing all of our work in the open and towards public benefit.

You can read more on the technical details, join the community on Discord, or follow along and contribute on Github.
The Team
We are thinkers, makers, artificial intelligence engineers, conservation technologists, learners, entrepreneurs, and artists who deeply care about the interdependent future of our planet and its species.

Stay up to date

A short email when something is worth sharing.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.