Meta has unveiled Brain2Qwerty v2, an AI research system that translates brain activity into text without requiring surgical implants. Instead of invasive devices inside the brain, the system uses magnetoencephalography (MEG) to record brain signals from outside the head. Meta describes an end-to-end approach in which AI analyzes those MEG recordings and produces written output, aiming to generate coherent sentences.
Multiple outlets report that Meta is presenting the system as an improvement over earlier work, with the best results described as reaching about 78% accuracy for a top-performing participant. Several sources also note a key limitation tied to the training and use process: Brain2Qwerty v2 learns from the typing behavior it is intended to reflect, meaning it is not simply converting brain activity into speech or text without any accompanying task.
Overall, the reporting centers on Brain2Qwerty v2’s use of non-invasive MEG and AI-based neural decoding to support text-based communication research. Meta’s claims focus on accuracy improvements and the goal of reducing or eliminating the need for invasive surgery.
