Hugging Face releases Carbon-X, an open AI model that lets researchers 'chat with DNA'
The 14-billion-parameter model answers plain-language questions about DNA and runs on a laptop, but it has not yet been tested in the lab.
The AI platform Hugging Face on Thursday released Carbon-X, an open model for understanding DNA that can run locally on a laptop, Wired reported. It was developed with researchers at Zhongguancun Academy in China and the University of Naples Federico II in Italy.
The model was trained on both genetic sequences and ordinary text, so users can ask how changes in DNA affect gene activity and health and get conversational answers. "We've unlocked a new capability in language models, which is fundamentally about being able to chat with DNA," said Lewis Tunstall, a senior research scientist at Hugging Face. He said it could help researchers design drugs and genetic therapies, for example by judging whether a mutation is harmless or causes disease.
The team trained it on 300 billion tokens combining DNA from mammals, other animals, plants, fungi and protozoa with English text. Viruses were left out on purpose to reduce the risk of dangerous uses. It can read stretches of about 50,000 DNA letters at a time, so users cannot upload a whole human genome.
With 14 billion parameters, it is small enough to run on a user's own hardware, so no data is sent to Hugging Face or any other company, which matters for sensitive patient information. Its developers caution that they have not yet checked its predictions in real laboratory experiments, so they cannot say how accurate it is.