Meta's Brain2Qwerty v2 decodes typed sentences from brain signals with 61% accuracy

Brain2Qwerty v2 achieves 61% word accuracy decoding natural sentences from non-invasive MEG brain recordings. Meta open-sourced the training code and the v1 dataset; the system was trained on 22,000 sentences from nine participants, each recorded for 10 hours. The best participant reached 78% word accuracy.