Tag
This paper presents Brain2Qwerty v2, an AI model that accurately decodes natural sentences from non-invasive magnetoencephalography brain recordings with a 39% word error rate, demonstrating that data scaling can help bridge performance gaps with invasive methods.