Room-Sized Brain Scanner Types Words From Brain Signals at 61% Accuracy
The system is meant to help patients who cannot speak, and the hardware still fills a shielded room.
Lightspeed Venture Partners · Meta Can Read Your Brain, The AI Science Race & AI Ranking With Ion Stoica | LightworkA room-sized scanner can now type words from brain signals alone with 61% word accuracy.
Meta's research lab built the system, called Brain2Qwerty. A person sits under a helmet-like scanner, known as MEG for the magnetic fields it reads from firing neurons, and the software reconstructs intended typing in real time. It averaged 61% accuracy, and reached 78% for the best participants, up from about 8% for earlier noninvasive methods.
The immediate goal is medical, to help millions of patients with brain damage who are unable to speak. It offers the most noninvasive option so far compared with implants that require drilling to place electrodes or threading sensors through blood vessels.
Do not expect mind-reading glasses soon. Each scanner costs multiple millions of dollars, needs a magnetically shielded room, and sometimes weighs about the same as a compact car.
Questions people ask
How does a MEG scanner actually pick up brain activity?
MEG stands for magnetoencephalography. It uses highly sensitive detectors around the head to pick up the tiny magnetic fields produced when large groups of neurons fire, and software then matches those patterns to words being typed.
Is it really reading thoughts?
Not in the way people fear. It decodes patterns linked to typing or intended language during a controlled task, not free-floating thoughts, memories, or opinions. It also requires a cooperative participant sitting still inside a massive, shielded scanner.
Could this ever shrink to something wearable?
In principle, smaller and cheaper magnetic sensors could one day move brain decoding toward wearables. For now, cost, size, shielding, and the need for extreme sensitivity keep MEG confined to specialized labs and clinics.