ARION exists to accelerate the development of interspecies translation by bridging frontier language model research and bioacoustics.
We advocate for a simple, concrete action: include tokenized non-human phonetic datasets in the pretraining corpora of frontier language models. The tools exist. The data exists. The theoretical foundations are published and peer-reviewed. We work to connect these pieces and track progress as the field develops.
ARION takes its name from the Greek poet Arion of Methymna (circa 625 BCE), whose music was understood by dolphins. His story — recorded by Herodotus — is the oldest known tale of acoustic communication bridging human and cetacean worlds.
Read the full story →ARION was founded alongside the publication of the founding paper, which synthesizes published research from Project CETI, Google DeepMind, and the NLP community into a concrete, actionable proposal for interspecies language alignment.
Read the paper →ARION builds on the published work of researchers at Project CETI (MIT CSAIL, UC Berkeley, Carleton University, CUNY), Google DeepMind, the Wild Dolphin Project, and the broader NLP community. We are grateful for their foundational contributions to cetacean communication research and cross-lingual alignment theory.