On June 29, Meta announced the revolutionary Brain2Qwerty v2 system, capable of reconstructing typed text solely from brain activity recordings — all without any surgical intervention. According to the company, the average accuracy at the individual word level reaches an impressive 61%, while previous non-invasive methods barely achieved 8%.
How It Works
The model was trained on a dataset of approximately 22,000 sentences. Nine healthy volunteers participated in the experiment, each spending 10 hours in a magnetoencephalography scanner — a device that measures magnetic fields generated by neural activity. Participants typed heard phrases on a keyboard, and the AI learned to match brain signals with the text.
Instead of manually extracting neural events, Brain2Qwerty v2 uses end-to-end deep learning, decoding data directly from raw brain signals. The system processes continuous activity recordings and reconstructs the entire sentence, using a fine-tuned language model that accounts for semantic context to refine the result.
Results and Limitations
The best-performing participant demonstrated 78% word-level accuracy. More than half of their sentences were decoded with one error or none at all. However, the development is still far from clinical or everyday use. Testing was conducted exclusively on healthy volunteers who actually typed text. For people who have lost speech or the ability to move, this approach requires separate validation.
Additionally, the system does not function as real-time text input. The model processes the entire sentence, so the user does not see each word immediately after typing it — a significant drawback for practical use.
Context and Prospects
The project is part of the Digital Brain Project — a $5 million initiative aimed at creating open neuroscience datasets. Recall that in September 2022, Meta published research on a "brain decoder" that used AI to convert thoughts into speech. The current version is a significant step forward, but commercialization is still far off.
My expert opinion: Achieving 61% accuracy on non-invasive data is a breakthrough that changes perceptions of neurointerface capabilities. However, the key challenge remains: adapting the technology for real patients with speech impairments. For now, Brain2Qwerty v2 is a powerful research tool, not a ready-made market solution. The BCI (brain-computer interface) market is valued at $2 billion and growing at 15% annually, and such developments could catalyze explosive growth in the next 3-5 years.