Meta has released Brain2Qwerty v2, a non-invasive brain-to-text system that decodes typed sentences from brain activity captured via a magnetoencephalography (MEG) scanner. The system achieved 61% average word accuracy (up to 78% for the best participant), a significant jump from earlier non-invasive systems that managed single digits. It uses an LLM pipeline — similar to ChatGPT — to reconstruct sentences from noisy brain signals, and Meta has open-sourced both the code and dataset. However, major limitations remain: the MEG scanner is room-sized and expensive, the system cannot operate in real time, and critically, it requires users to actually type to train the model — making it unsuitable for fully paralyzed patients. Invasive implants still lead in accuracy, with surgical systems reaching 92% sentence-level accuracy.
759 Impressions1 Comment