Meta's brain-scanning system reads sentences non-invasively, code open source

Meta released Brain2Qwerty v2, a non-invasive brain-scanning system that decodes sentences from brain activity with 61% word accuracy—approaching the performance of surgical implant methods. The system uses AI and magnetoencephalography (MEG) recordings to translate brain signals into text in real-time, and Meta has open-sourced the training code and datasets to accelerate neuroscience research for people with communication disorders. The technology shows potential to bypass the scalability limitations of invasive surgical procedures currently used for brain-computer interfaces.