Meta unveiled Brain2Qwerty v2, a non-invasive brain-computer interface that decodes typed sentences from raw neural signals in real time. The company claims this is the highest-performing system of its kind. The announcement occurred alongside the publication of original Brain2Qwerty research in Nature Neuroscience.
The system achieves an average word accuracy of 61% across participants using magnetoencephalography (MEG). For the best-performing participant, accuracy reached 78%, with more than half of sentences decoded containing one or fewer word errors.
Brain2Qwerty v2 was trained on approximately 22,000 sentences from nine volunteers, each recorded for 10 hours while using an MEG device. This system employs end-to-end deep learning on raw brain signals combined with fine-tuned large language models. It advances from character-level decoding to decode words and semantics directly.
Meta stated that performance scales log-linearly with data volume, indicating potential for further accuracy improvements with additional training data. The 61% word accuracy is a significant improvement from Brain2Qwerty v1, which had a character error rate of 32%.
Previously, achieving high word-level accuracy in brain decoding required surgical implants, which carry risks like infection and signal degradation. Meta indicated that this research could significantly benefit patients with brain lesions or neurological disorders that hinder communication. The company said, “We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating.”
To support ongoing research, Meta released the full training code for both Brain2Qwerty v1 and v2. Additionally, the Basque Center on Cognition, Brain and Language, a research partner, released the v1 dataset. The research has undergone peer review and was published in Nature Neuroscience.
Public reaction to the announcement was mixed. Some praised the technology for its accessibility, while others expressed distrust regarding Meta’s role in brain-reading technology, citing concerns over the company’s advertising-driven business model.
