Meta's non-invasive AI converts brain signals into typed text without surgery

Meta's FAIR research team has developed Brain2Qwerty v2, a system that reads magnetic signals from outside the skull to reconstruct sentences a person is typing — with no implants or surgery required. Accuracy continues to improve with additional recordings, partly because AI agents rewrote parts of the optimisation code autonomously. Clinical use for paralysed patients remains a distant goal, but the system is narrowing the gap with invasive surgical implants.

More on this topic

💻 Technology Meta's brain-scanning AI reads sentences non-invasively, code now open source Hacker News · 27d ago 💻 Technology Meta's Brain2Qwerty v2 hits 61% word accuracy without brain surgery The Register · 28d ago 💻 Technology Meta's Brain2Qwerty translates brain activity into text without surgery Decrypt · 29d ago

More on this topic

💻 Technology Meta's brain-scanning AI reads sentences non-invasively, code now open source Hacker News · 27d ago 💻 Technology Meta's Brain2Qwerty v2 hits 61% word accuracy without brain surgery The Register · 28d ago 💻 Technology Meta's Brain2Qwerty translates brain activity into text without surgery Decrypt · 29d ago