Meta unveils brain-to-text AI that decodes thoughts without surgery

Meta ↘ 4.90% on Sunday unveiled Brain2Qwerty v2, a non-invasive brain-computer interface that decodes typed sentences from raw neural signals in real time, marking what the company calls the highest-performing system of its kind. The announcement coincided with the publication of the original Brain2Qwerty research in Nature Neuroscience.digg+1

The system achieves an average word accuracy of 61% across participants using magnetoencephalography (MEG), a non-invasive technique that measures magnetic fields generated by brain activity. For the top-performing participant, accuracy reached 78%, with more than half of sentences decoded with one or fewer word errors.linkedin+1

How It Works

Brain2Qwerty v2 was trained on approximately 22,000 sentences from nine volunteers, each recorded for 10 hours while wearing an MEG device and typing. The pipeline uses end-to-end deep learning on raw brain signals combined with fine-tuned large language models to bridge what Meta described as "the gap between noisy neural data and coherent language".digg+1

The system advances beyond the character-level decoding of its predecessor to decode words and semantics directly, and Meta says performance scales log-linearly with data volume — suggesting further gains are achievable with more training data.linkedin

Clinical Promise and Context

The 61% word accuracy represents a dramatic leap from previous non-invasive approaches. Brain2Qwerty v1, published the same day in Nature Neuroscience, had achieved a character error rate of 32% using MEG. Until now, high word-level accuracy in brain decoding had been available only through surgical implants, which carry risks including infection and signal degradation over time.dallasexpress+2

Meta framed the research as a potential lifeline for patients with brain lesions or neurological disorders that prevent communication. "We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating," the company said.linkedin

Open Science, Mixed Reception

To accelerate further work, Meta released the full training code for both v1 and v2, while its research partner, the Basque Center on Cognition, Brain and Language, released the v1 dataset. Jean-Rémi King, a researcher involved in the project, clarified on social media that the peer-reviewed paper was published in Nature Neuroscience.digg+1

Public reaction was divided. While some praised the advance for accessibility, others expressed distrust of Meta's involvement in brain-reading technology given the company's advertising-driven business model.digg