Meta announced earlierIt's called "Brain2Qwerty", which can interpret brain activity information in a non-invasive wayTechnical Research White Paper, claiming that it can help patients who have lost their communication function due to brain damage regain the ability to communicate with the outside world.
According to the instructions, "Brain2Qwerty" can interpret the sentences conceived in the patient's mind through electroencephalogram (EEG) or magnetoencephalogram (MEG) waveforms and convert them into specific text with an accuracy rate of 80%.
Through this method, Meta claims that more patients who are unable to communicate normally due to brain damage can regain the ability to communicate with the outside world through non-invasive means.
On the other hand, Meta plans to use this technology as a foundation to further enable the brain's thoughts to be converted into verbal content more quickly, and then to convert thoughts into action commands.
Currently, "Brain2Qwerty" analyzes 1000 brain activity images per second to learn how each word, syllable, letter, and other information is conceived in the brain. This allows the neural operation model that produces language in the brain to be used for the development of advanced machine intelligence.
However, this research still has many details to overcome. For example, the current average character error rate is 32%, which means that on average one error occurs for every three characters. At the same time, the current research mainly uses magnetoencephalography content analysis, so it must be carried out in an environment free from other electromagnetic interference, and the subjects must remain as still as possible.
