Hugging Face’s New AI Biology Model Lets You Chat With DNA
wired ·
Hugging Face has released Carbon-X, a new open-source AI model that allows users to interact with DNA through natural language queries. Developed in collaboration with researchers from Zhongguancun Academy and the University of Naples Federico II, the tool is designed to help scientists understand how genetic changes affect gene activity and health. Unlike previous models that output numeric data, Carbon-X provides conversational answers, making genomic analysis accessible to non-technical users. The model was trained on 300 billion tokens combining DNA sequences from various species with English text, while intentionally excluding viral data for safety reasons. With 14 billion parameters, it is small enough to run locally on a laptop, ensuring that sensitive patient data remains private and is not sent to the cloud. Users can query specific DNA segments of about 50,000 base pairs to assess whether mutations are benign or disease-causing. However, developers caution that the model has not yet been validated through wet lab experiments, so its predictive accuracy remains unconfirmed.
wired