To compete with Elon Musk’s Neuralink, Mark Zuckerberg-owned Meta has introduced a new artificial intelligence (AI) system that can convert brain signals into text in real time without the need for brain surgery. Called Brain2Qwerty v2, the new model will help people who have lost the ability to speak because of health conditions like stroke, brain injuries or neurological disorders.
“Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall 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 US-based tech giant said.
What Is Brain2Qwerty v2?
According to Meta, Brain2Qwerty v2 is company’s most advanced AI system yet that can turn brain signals into text without surgery.
Unlike existing brain-computer interfaces that often require electrodes to be implanted inside the brain, Meta’s latest research relies on magnetoencephalography (MEG), a technology that records brain activity from outside the head. This means the system does not require any surgical procedure.
Meta says that Brain2Qwerty v2 is an upgraded version of the company’s earlier Brain2Qwerty v1 project.
Unlike the earlier version, which needed the timing of every key press to convert brain signals into text, the new Brain2Qwerty v2 can generate complete sentences directly from continuous brain activity.
It first detects letters, then understands words and finally uses a large language model (LLM) to create meaningful sentences.
For the research, Zuckerberg-owned company trained Brain2Qwerty v2 on approximately 22,000 sentences from nine volunteer participants, each recorded for 10 hours wearing a magnetoencephalography (MEG) device while actively typing.
“Instead of relying on hand-crafted pipelines to detect neural events, we use end-to-end deep learning to decode directly from raw brain signals,” the company said.
According to the company, the AI achieved an average word accuracy of 61 per cent, while the best-performing participant reached 78 per cent accuracy.
Meta claims this is a big improvement over earlier non-invasive brain decoding systems, which managed only around 8 per cent word accuracy.
Despite the breakthrough, Meta admits the technology is still far from being ready for everyday use. Meta believes these limitations can be addressed over time. The company says increasing the amount of training data will further improve decoding accuracy.
“Our hope is that this work, done in the open, advances neuroscience to identify, diagnose, and treat neurological disorders faster,” the company noted.